IBM PowerPC Announcement

[MacRumors]

This Forbes article (http://www.forbes.com/technology/newswire/2002/10/13/rtr749520.html) (courtesy of MacSurfer (http://www.macsurfer.com)) brings us the first news of IBM's new PowerPC Chip that has been speculated about for the past few weeks. According to the article (which references a Monday announcement), the new chip is due in late 2003 and will start at 1.8GHz.

Meanwhile, rumors of Apple involvement continues... but no official stance just yet:


An industry source said Cupertino, California-based Apple would use the chip in its Macintosh computers.

[edesignuk]

This is good news and all, but late 2003 and we're still stuck with just 1.8Ghz (I know it says starting at, so maybe a top speed of 2.2Ghz, if we're lucky).
Don't flame me for this, but, Intel and AMD will probably have introduced brand new lines of processors by then running at 4Ghz+...somethings not quite right there, we need more POWER :(

[bobindashadows]

is this story based on fact, or speculation, is what i'm wondering. 1.8 Ghz, not bad compared to the 1.4 stuff some people speculated. The gap will have greatly widened by then, hopefully there will be some secrets up apple's sleeves. (if it is based on official information)

On another note, I got the first comment! (bahhh!!!!)

[edesignuk]

Originally posted by bobindashadows

On another note, I got the first comment! (bahhh!!!!)
No u didn't! :D :p

[Over Achiever]

Late next year...sounds reasonable. I'd rather get a PowerPC type chip...ah well, I'll buy a powerbook soon, get the aging G4 :(

[e-coli]

lovely.

the holy grail "G5" we've all been waiting for...

but still sucking wind at a meager 1.8 GHz. And not for another year. By then you should be able to get a cheap 5 GHz box.

:rolleyes:

[vniow]

Originally posted by edesignuk
This is good news and all, but late 2003 and we're still stuck with just 1.8Ghz (I know it says starting at, so maybe a top speed of 2.2Ghz, if we're lucky).
Don't flame me for this, but, Intel and AMD will probably have introduced brand new lines of processors by then running at 4Ghz+...somethings not quite right there, we need more POWER :(


Probably not 4Ghz, but at least 3.7 Ghz I'm guessing, AMD's Clawhammer will be a 3400+ (about 3Ghz I think) so Intel would hve at least caught up to that since the AMD chip is going to be released in the first quarter.
A 1.8 Ghz Power4mini-me will likely kick some serious ass, but it just doesn't look that fast, Intel and AMD are way past 1.8 Ghz and will be even farther by the time this gets released.

[arn]

It's not all about Mhz though... we'll see...

arn

[Aciddan]

[quote] "This is the time to introduce a 64-bit machine capable of being used on a desktop,"

As much as most posters would want to see this "lite" chip in the near future, I think IBM has positioned itself well against the 64bit competition from Intel and AMD. The 1.8 GHz clock speed is seriously impressive, seeing as the hammer is around the 800Mhz mark (there was a review on anandtech a while back). Even though MHz ain't everything :)

I'm guessing this Power 970 (speculation about single-core looks correct) will supercede some motorolla 85xx chip (85xx possibly in January?) since the release date of the Power chip is late 2003

Guess we'll get some more tidbits on the chip come the demos/papers/speeches of the microprocessor forum...

-- Dan =)

[ffakr]

Originally posted by e-coli
lovely.
but still sucking wind at a meager 1.8 GHz. And not for another year. By then you should be able to get a cheap 5 GHz box.

:rolleyes:

hahaha. No, you won't be able to get a 5GHz box by the end of next year, but there will be a P4 at 4ghz. With the new hyperthreading support, don't expect P4 to average over 2 ops per cycle, probably around 1.5.
If the word so far is correct, the "meager" IBM chip will have a max of 8 ops per cycle... figure an average of around 4 ops per cycle if you look at the PPCs history.
I expect it will easily do twice the work per cycle as a next-gen P4. Try doing high precision math and it will pull way ahead.

1.8GHz, if the other rumors are correct, will be very very powerful. 2+ is always better, but a 1.8 GHz 64bit chip is pretty sweet. Just wait till the Mac clusters start posting scores with this chip...

[nixd2001]

Originally posted by edesignuk
This is good news and all, but late 2003 and we're still stuck with just 1.8Ghz (I know it says starting at, so maybe a top speed of 2.2Ghz, if we're lucky).
Don't flame me for this, but, Intel and AMD will probably have introduced brand new lines of processors by then running at 4Ghz+...somethings not quite right there, we need more POWER :(

Although it goes against the grain of having discussion groups based on rumors, wait until the week has played out and some technical details have become available.

We need throughput (power if you wish to call it that). But we probably need memory bandwidth before we need Mhz (as Arn sez later as well). a g4 with an internal clock rate of 2Ghz would probably not be much faster than a 1.25Ghz G4.

[wumpus]

This PowerPC 970 sounds very impressive...Shame it is coming so late, and it still is unclear if it supports AltiVec (which has kept us shackled to the lead turkey G4 and Motorola), which is crucial.

So, what can Apple do for the next year? More overclocking/hacking of the embedded/telecoms focused irrelevance that is the 'Motorola MPC 7455 Host Processor'? Look at Moto's own official documents: The chip does NOT support DDR in any form, or a 167mhz FSB, or an official speed of 1.25ghz, or modern features like out-of-order processing (where instructions are executed in the fastest way possible, rather than just in the order they were issued)...That 7lb heat sink in the latest Macs gives it away, the latest G4s have been overclocked and hacked to the gills by Apple, as they have no choice but to squeeze tiny improvements out of an expensive, out of date CPU that was never designed with Macs in mind.

So what now? Until the 970 arrives, will Apple have no choice but to hack and overclock some more, maybe using exotic cooling methods to force 1.33 or 1.4ghz, and a 200mhzish FSB? Or will Apple come up with another hack (like their 'DDR' implimentation), and shove 4 of these lead balloons into a Mac (never mind they only support 2-way SMP)? Combine this with Apple's 10.2 and .mac rip-offs, forced X booting, underspeccing of machines, etc. and we will lose yet more market share.

If Moto's rumoured '7470' comes along, built on an 0.13 micron process and modernised a bit (real DDR333, 266mhz FSB, 512k L2, 4mb L3, and speeds of 1.6ghz+) we might get some breathing space. If the rumoured '7500' or '8500' (there is already a 'G5'-class CPU, Motos impressive MPC8540, but it is a System-on-a-chip with no AltiVec and is an 'Integrated Communications Processor') with the enhanced e500 core, DDR400 with fast on-chip memory controller, and RapidIO 500mhz FSB then we will be given a new lease on life until IBM comes around.

Dont get me wrong, the 970 is great news, and it is the first real desktop-focused PowerPC since the original G3. The POWER4 is the most powerful CPU available anywhere, and even a 'lite' version will be great, as will be a closer relationship with IBM. I am going to hang on to my maxxed out G3/400 and maybe stick a 7455 in it when i can and wait for Macs that can truly compete and compell again....

[Nipsy]

If this is a Power4 Lite with Altivec as we've been speculating, the MHz myth will finally be over.

If any of you have seen or used a 1GHz Power4 in workstation guise (lowered L2 from server), you know it is prolly the fastest chip you can get at a desk.

I'm sad to hear that it is a year away, but I can tell you that Forbes has better intel (pun intended) than we do.

[daRAT]

Ya know, geek.com had a few months back an article, (which I ca n not find now :rolleyes: About IBM announcing in thier R&D division a 250 or so GHZ cpu, no that anit an error :] The article went on to say 300 ghz was doable and if they wanted, it could be on the self by end of next year.

Granted, this would be probably a 100,000$ chip, but it makes me wonder if Intel. AMD and IBM already have 5 + ghz pretty much ready to go.

I know they couldn't release them (the market wouldn't like that), who would buy a new compter a year later? Why would you need too with that speed?

While I half agree with Arn re; mhz anit evertything, it is sounding a mantra to sooth all dissapointments in the lackluster cpu speed from Apple. I don't blame Apple, I think they are caught in the middle, switch to IBM or stick with Moto, or *gasp* go with Intel or AMD, (my opinion would be AMD, while I am not crazy about thier chips, I do think they would be more agressive and inovative than Intel).

Thanks for listening :]

*kicks the soapbox to the next macrumors poster...*

[AmigaMac]

Neither will AMDs ClawHammer nor Intel's Itanium will be over 3 GHz next year... you're getting Athlon and P4 clockspeeds mixed up with their 64bit variants which wont be clocked as high! So quit sweating the MHz myth now will ya!!!

[FattyMembrane]

i hate to be another pessimist, but 1.8 ghz in a year is pathetic. there are lots of people who've been waiting for the g5 since about '98 and the best we're going to get is a 1.8ghz pushing the begining of 2004? i certainly hope that those of you with more faith, who have stated that this is very sufficient, are right, because unless this chip can process at lightspeed, the future of apple looks rather dismal if they go ahead with the power4. if moto came out of leftfield with some badass chip in 6 months, i'd be absolutely shocked, but very happy.

[rice_web]

The one thing that I am hopeful for is the bus speed. IBM's G5 has been rumored to support 6.4GBs of throughput, while their G3 supports 3.2GBs. Using some math....

If 3.2GBs = 200MHz x 2 system bus (200MHz with DDR) = 400MHz

Then 6.4GBs = 200 x 2 x 2 (200Mhz with DDR and double-pumped) = 800MHz

[GeeYouEye]

It should be noted that there was an article a while back about the "Pentium 5", and how it *gasp* only clocks at 1.2-1.4 GHz, with almost identical performance to a 1.25GHz G4. It would, of course be suicide for Intel to release it, and marketing is probably why they haven't yet. 2003 and 2004 are going to be some interesting years for the computer industry.

[Aciddan]

Originally posted by daRAT
Ya know, geek.com had a few months back an article, (which I ca n not find now :rolleyes: About IBM announcing in thier R&D division a 250 or so GHZ cpu, no that anit an error :] The article went on to say 300 ghz was doable and if they wanted, it could be on the self by end of next year.

The last stuff I read on this was from a defense journal (after being told about it by an old lecturer of mine) - it was a 50Ghz chip (a few years ago now), BUT these are small micro-controllers not chips that are intended for use as CPUs. No doubt used for very specific tasks too...

I can't remember any more details, but I'm sure some of the guys here could shed some more light on this...

Summary - 300Ghz is a no-go...

-- Dan,

[mozez]

i hate to tell you guys this but SPEC came out with there new stats on floating point and other used processing methods for chips and the power 4 at 1.3 ghz got spanked by the 32 bit 2.8 ghz p4, even the 2600 from amd beat it flat out. this was posted in another thread just days ago, that the power 4 will not be faster than a new p4 or amd. mhz is not everything, it means nothing, stats and bench marks do. and all mac benchmarks against pcs, always have to use like a horrible bad pc to make it even look close, like taking away all pc technology so that it is closer to the mac and thus more equal. a 1.8 power 4 lite, will not be able to beat whatever is out on a pc, i wonder if it would even beat what is out now, forget next year. i truly hope that 1.8 is the low end and like 2.5 or dual 2.2 is the high cause they will need it to even compete. especially since pcs today that are custom built and i stress custom built, are as stable or more stable than macs. just an opinion, not based on how awesome osx is, cause that's what i use and i love it.

[scem0]

How can they already know what the lowest speed will be.... My guess is that they don't know what speeds they will be at, and they will know early next year. But not now. And some people are forgetting that 1.8 is the lowest speed - according to the article. A ~2.2GHz will perform as fast as anyone needs it to. I am more concerned about the pricing, and not the speed. Apple's machines are waaaaay overpriced right now, if they can lower their prices then their market share will go up a bunch...

[ddtlm]

mozez:

The 1.3ghz Power4 still leads the P4 and Athlon nicely in floating point SPEC. Proof? Look here:

http://www.aceshardware.com/read_news.jsp?id=60000436

It doesn't do too bad in integer SPEC either, but does loose to the top x86 chips there.

[beatle888]

Originally posted by edesignuk
This is good news and all, but late 2003 and we're still stuck with just 1.8Ghz :(

just calm your noodle ed...this is gonna rock.

[arn]

Originally posted by wumpus

So, what can Apple do for the next year? More overclocking/hacking of the embedded/telecoms focused irrelevance that is the 'Motorola MPC 7455 Host Processor'? Look at Moto's own official documents: The chip does NOT support DDR in any form, or a 167mhz FSB, or an official speed of 1.25ghz

The 1.25GHz Motorola Processor is NOT overclocked. It's officially a 1.25GHZ chip.

http://forums.macrumors.com/showthread.php?threadid=12115
arn

[scem0]

Why is everyone freaking out about it only being 1.8 GHz.... That is very fast - you have to look past that number that comes right before the GHz....

[JW Pepper]

Remember you can get 4 x as much water through a 2" pipe as as you can through 1" pipe. I suspect that a similar situation exists with 64bit processors, so I wouldn't compare the mghz.

[arn]

Originally posted by FattyMembrane
i hate to be another pessimist, but 1.8 ghz in a year is pathetic. there are lots of people who've been waiting for the g5 since about '98 and the best we're going to get is a 1.8ghz pushing the begining of 2004?

Apples and Oranges people...

1.8GHz of the new chip DOES NOT EQUAL a 1.8 GHz G4.

If you must focus on numbers, see this link (http://www.aceshardware.com/read_news.jsp?id=60000436) ddtlm posted. And look at the Spec scores of a 1.3ghz Power4 compared to a 2.8ghz Pentium.

arn

[ddtlm]

scem0:

Yeah, 1.8ghz is a nice increase over 1.3ghz (38%)... for the P4 to keep up that pace it has to make it to 3.8ghz or so by the PPC-970 launch day. Intel can probably do that, but will it do more than that?

(Of course I just made the assemption that the PPC-970 will perform the same per clock as the Power4.)

[scem0]

my guess is no. Intel is in trouble right now. I think they have dug themselves into a hole kinda... And now with all the legal stuff. I think this chip will be more then enough.

[Aciddan]

Originally posted by scem0
Why is everyone freaking out about it only being 1.8 GHz.... That is very fast - you have to look past that number that comes right before the GHz....

(shameless me too post)

Have a look at the clock speeds on other 64 bit chips:

http://techupdate.zdnet.com/techupdate/stories/main/0,14179,2870870,00.html

Itanium 2 at 1 Ghz and AMD at 2 (and as I said earlier, I saw a review with an 800MHz hammer) - when comparing Apples with Apples (Pun intended :P) you can see that the IBM chip is extremely competitive. Even intel has backed themselves into a corner with the Itanium because of MHz marketing...

the end of 2003 is going to be *very* cool. In the short term, I'm more itnerested in rumors of new Moto chips (remember the rumor about IBM and Moto being used in high end machines in the future?) - I'm speculating that this will mean a new moto chip in January (either a funkified G4 or an 85xx) then at the end of the year the high end professional chips will more to IBM - and the motos will filter down into the consumer products...

-- Dan =)

[alex_ant]

Originally posted by FattyMembrane i hate to be another pessimist, but 1.8 ghz in a year is pathetic. there are lots of people who've been waiting for the g5 since about '98 and the best we're going to get is a 1.8ghz pushing the begining of 2004? i certainly hope that those of you with more faith, who have stated that this is very sufficient, are right, because unless this chip can process at lightspeed, the future of apple looks rather dismal if they go ahead with the power4.
Taking into account the rumored specifications of this chip, then at 1.8GHz, this chip should be able to pretty much stomp on any Intel or AMD product on the horizon before 2004.

6.4GB/s memory bandwidth, enhanced double-precision AltiVec-compatible vector unit, 8 operations/cycle... it will be very very nice. IBM wouldn't be building it if they knew it wouldn't be competitive.

[scem0]

That is what I am hoping....

[FattyMembrane]

Originally posted by ddtlm
The 1.3ghz Power4 still leads the P4 and Athlon nicely in floating point SPEC. Proof? Look here...

damn, apple needs to start using SPARCs :D . the 1ghz itanium is nipping at the heels of the power4 and i certainly doubt that if apple rolls out 64 bit computers, intel will just leave pc manufacturers with the p4.

[alex_ant]

Originally posted by rice_web
The one thing that I am hopeful for is the bus speed. IBM's G5 has been rumored to support 6.4GBs of throughput, while their G3 supports 3.2GBs. Using some math....

If 3.2GBs = 200MHz x 2 system bus (200MHz with DDR) = 400MHz

Then 6.4GBs = 200 x 2 x 2 (200Mhz with DDR and double-pumped) = 800MHz
Who cares about the bus clock? If it can push 6.4GB/s at low latency with a 25MHz bus, I won't be complaining.

[alex_ant]

Originally posted by mozez
i hate to tell you guys this but SPEC came out with there new stats on floating point and other used processing methods for chips and the power 4 at 1.3 ghz got spanked by the 32 bit 2.8 ghz p4, even the 2600 from amd beat it flat out. this was posted in another thread just days ago, that the power 4 will not be faster than a new p4 or amd.
Which new stats are you referring to? I know the P4 wins in SPEC_int2000, but it's close in fp. I'm going to assume that this chip is not strictly a stripped down POWER4, that it is a stripped down POWER4 with various enhancements. I would imagine this is what will account for its delayed arrival.

[nixd2001]

Originally posted by alex_ant
6.4GB/s memory bandwidth, enhanced double-precision AltiVec-compatible vector unit, 8 operations/cycle... it will be very very nice. IBM wouldn't be building it if they knew it wouldn't be competitive.

Woooahh - did I miss something or is the "double precision" bit of Altivec something you're our first info source on?

[applemacdude]

The P4 is just overclocked. Thats why the mac is hot and pc are very hot. it(p4) produces a lot of heat specially in the laptops

[alex_ant]

Originally posted by nixd2001
Woooahh - did I miss something or is the "double precision" bit of Altivec something you're our first info source on?
Good question - I thought I read about this new chip having double-precision VMX, but I can't remember where from. Oh well, I can't imagine that it wouldn't... :)

[nixd2001]

Originally posted by alex_ant

Good question - I thought I read about this new chip having double-precision VMX, but I can't remember where from. Oh well, I can't imagine that it wouldn't... :)

It would certainly make sense, given performance seems to be a major driver ;)

[MacCoaster]

Originally posted by applemacdude
The P4 is just overclocked. Thats why the mac is hot and pc are very hot. it(p4) produces a lot of heat specially in the laptops
Pentium 4s are not overclocked at all. They've passed factory tests to run at their respective rating. I think you misunderstand what overclocking is. Also, Intel is planning a 3.06GHz this month, well, at least some time soon later this year, as far as I know.

Last time I used a PowerBook G4 800MHz, the keyboard got very hot, if not hotter than an AMD Athlon Mobile XP laptop I used. (Athlons are much hotter than P4s, by the way)

You might want to research more indepth before posting outlandish fallacies.

[robguz]

Whoopee, the mac world will finally have 1.8Ghz by 2004 (c'mon we all know that late 2003 means 2004 at the earliest-otherwise we'd all be using the real Apollo chips running at 1.6Ghz for the past 6 months). Meanwhile the Wintel world will be at 4Ghz plus. In 2004 we'll get to party like it's 2000!

Take all this with a huge grain of salt folks and pray for compatability.

[ibjoshua]

okay, let's talk about perceived performance.

i'm sitting at my work pc. it's a beige p4 1.6ghz with XP pro. it's full of cheap nasty components including a built in video card.

a lot of the time it is a dog! the start menu sometimes takes 10 to 30 seconds to render. applications regularly crash and killing errant processes is much harder than it was back in NT4 days.

do i know why? no. could be the virus software, could be IO could be the cpu, could be the motherboard, could be the ram. it could just be XP. what i do know is a lot of the time i'd rather be using my 500mhz iBook with 10.2.

it's as simple as that. i would guess most consumers and plenty of business workers have dull clones.

macs will (hopefully) always kick cloned arses and that's why we love them.

BTW, why do people keep talkin about the itanium? how many desktops have you seen with an itanium cpu in them?

and for the mhz believers here is an experiment for you:

1. open a text editor
2. hold down any key for roughly 60 seconds.
3. open another file
4. type "hi world"
5. copy it
6. press command-v (ctrl-v on PC) repeatedly for roughly 60 seconds.
7. count the number of characters in each file.

which file had the greater number of characters in it?
i'll wager even though each command-v took longer you still did more work by printing 8 characters at a time for the same period.

i love my ibook but i have a feeling my next computer will have an even beefier IBM cpu. i can wait.

i_b_joshua

[arn]

Originally posted by robguz
Whoopee, the mac world will finally have 1.8Ghz by 2004 (c'mon we all know that late 2003 means 2004 at the earliest-otherwise we'd all be using the real Apollo chips running at 1.6Ghz for the past 6 months). Meanwhile the Wintel world will be at 4Ghz plus. In 2004 we'll get to party like it's 2000!

Please read the rest of this thread.

Quick question: Would you rather have a higher MHz number or Faster Performace?

Remember, the IBM PowerPC is a different architecture... you can't compare the new chip to current Motorola G4's. You certainly can't extrapolate Mhz.

The 1.3 GHz IBM Power4 benchs close to the 2.8GHz Pentium according to numbers cited in this thread.

arn

[ddtlm]

robguz:

Read the thread, your post is ignorant. Based on SPEC results found here:

http://www.aceshardware.com/read_news.jsp?id=60000436

It is apparent that Intel needs to make it to 3.86ghz or so for the Pentium4 to keep the same performance that they have relative to a 1.3ghz Power4 when compared to a 1.8ghz PPC-970 (assuming that the PPC-970 has the same IPC as a Power4).

[vniow]

Originally posted by arn


Please read the rest of this thread.

Quick question: Would you rather have a higher MHz number or Faster Performace?

Remember, the IBM PowerPC is a different architecture... you can't compare the new chip to current Motorola G4's. You certainly can't extrapolate Mhz.

The 1.3 GHz IBM Power4 benchs close to the 2.8GHz Pentium according to numbers cited in this thread.

arn


There is about zero question that a 1.3 Ghz power4mini-me will kick any Intel's ass, that's not in dispute, what people are ponting out is numbers, plain and simple.
3Ghz looks a helluva lot faster than 1.8 ghz, even if it's really not.
The PC world is hooked on speed and it's going to stay that way intil we get a low clock speed kick ass CPU, and hopefully it will be this chip.

[robguz]

Originally posted by ddtlm
robguz:

Read the thread, your post is ignorant. Based on SPEC results found here:

http://www.aceshardware.com/read_news.jsp?id=60000436

It is apparent that Intel needs to make it to 3.86ghz or so for the Pentium4 to keep the same performance that they have relative to a 1.3ghz Power4 when compared to a 1.8ghz PPC-970 (assuming that the PPC-970 has the same IPC as a Power4).

Uh, no, you're the ignorant one for assuming that this new chip will be exactly the same performance wise as the Power4, that Spec benchmarks will be exactly the same running OSX, that Apple will use this chip for certain, that IBM will really ship this within a year, that Apple will start using the chip the day IBM releases it. Don't give me benchmarks on a totally different chip running a totally different OS. Can OSX even run on this unannounced chip? Who knows. Maybe it will take Apple another year to accomplish that. It took them 6 years to get a usable version of OSX out the door even though there were already PowerPC builds of the NextStep before Apple purchased Next.

I read the thread. Basically the same as every other thread in the past 5 years, e.g. new chip is coming and it will be great/we'll blow away Wintel/naysayers like me who say I'll believe it when I see it and if Apple's past history is any iindication, don't hold your breath.

[arn]

Originally posted by edvniow



There is about zero question that a 1.3 Ghz power4mini-me will kick any Intel's ass, that's not in dispute, what people are ponting out is numbers, plain and simple.
3Ghz looks a helluva lot faster than 1.8 ghz, even if it's really not.
The PC world is hooked on speed and it's going to stay that way intil we get a low clock speed kick ass CPU, and hopefully it will be this chip.

I don't think so... I think people are seeing 1.8GHz and comparing it to 1.25 GHz and complaining without thinking of the potential advantages beyond the #s. Of course, this is all speculation at this point... but it's not far-out speculation as it is a Power4 based chip.

The thing about Mhz not being the end all is going to be a point that I think more consumers will learn... not just because of Apple... but also AMD and Intel's going to have to sell it as well... their Itanium processors start at 800mhz-1ghz... so they're a bit behind the 8ball as well trying to market those as "powerful" processors.

arn

[arn]

Originally posted by robguz


I read the thread. Basically the same as every other thread in the past 5 years, e.g. new chip is coming and it will be great/we'll blow away Wintel/naysayers like me who say I'll believe it when I see it and if Apple's past history is any iindication, don't hold your breath.

There are a lot of assumptions floating around... but we have to assume some things in order to even discuss this...

Point being, there is a 1.8GHz IBM Power4-based PowerPC due in Late 2003.

Based on your original post, you seemed to not understand that there were architectual differences that potentially make a 1.8GHz chip compare favorably to what intel's got on the market.

I'm not saying you have to accept it - but your post would be the same kinda of post someone would write if Motorola announced a 1.8GHz G4 in late 2003. There is a big [potential] difference here... and I was pointing that out.

arn

[vniow]

Originally posted by arn


I don't think so... I think people are seeing 1.8GHz and comparing it to 1.25 GHz and complaining without thinking of the potential advantages beyond the #s. Of course, this is all speculation at this point... but it's not far-out speculation as it is a Power4 based chip.

The thing about Mhz not being the end all is going to be a point that I think more consumers will learn... not just because of Apple... but also AMD and Intel's going to have to sell it as well... their Itanium processors start at 800mhz-1ghz... so they're a bit behind the 8ball as well trying to market those as "powerful" processors.

arn


I think you're right to a point, they are starting to see that speed istn't everything, Intel's a more recent ads celebrate multimedia, not Ghz, but I'm a PC user and while I can see through the Mhz myth, I know of a lot of people that don't, they tend to look at numbers rather than actual performance.
It is changing and I think Apple has had at least some part in it but it's a slow change.
I think a chip like this may prove that Mhz doesn't really matter as much as other things, but I'm not so sure it will be a quick change.

[ddtlm]

robguz:

Uh, no, you're the ignorant one for assuming that this new chip will be exactly the same performance wise as the Power4, that Spec benchmarks will be exactly the same running OSX, that Apple will use this chip for certain, that IBM will really ship this within a year, that Apple will start using the chip the day IBM releases it. Don't give me benchmarks on a totally different chip running a totally different OS.
Haha! Not only did I list the IPC as an assumption which clearly means I am aware that it was one, but I didn't say a damn thing about OS's or about Apple. Woot! You lose! :)

[Postal]

Assuming that the G4 as we know it is somewhat faster than an equivalently-clocked Athlon or P4 (which should generally be true), and the hints have been that the single-core PPC 970 may be about twice as quick as a G4... that's fast. :)

For reference, I believe the first 64-bit Athlons are expected to launch at 1.6 GHz (or a 3400+ rating - the 800 MHz part was an early sample), so assuming IBM matched the performance ratio exactly - and they may well come close or exceed it - the PPC 970 would be rated as the equivalent of an "Athlon 3800+" chip.

Depending on the actual specs the chip may be useful more to keep Apple genuinely competitive (rather than the dramatic gap we see now), but for many Mac users or potential converts, that's enough!

[alex_ant]

Originally posted by robguz
Don't give me benchmarks on a totally different chip running a totally different OS.
The OS doesn't really matter in the SPEC benchmarks - what matters is the processor in question and the compiler.
Can OSX even run on this unannounced chip? Who knows. Maybe it will take Apple another year to accomplish that.
Apple needing a year to get OS X to run on this new chip would be like Microsoft needing a year to get Windows 95 running on anything faster than a 486. This 970 will be PPC-compatible; it will probably necessitate an OS upgrade, but I would consider that more or less a triviality.
It took them 6 years to get a usable version of OSX out the door even though there were already PowerPC builds of the NextStep before Apple purchased Next.
We both know that moving a highly portable OS to a CPU that is more or less binary compatible with existing PowerPCs cannot be compared to re-engineering and programming major portions of a brand new operating system.
I read the thread. Basically the same as every other thread in the past 5 years, e.g. new chip is coming and it will be great/we'll blow away Wintel/naysayers like me who say I'll believe it when I see it and if Apple's past history is any iindication, don't hold your breath.
I understand how you feel, but we (Mac users) should be happy to hear that finally we have some real, solid evidence of a way out of this long, dark tunnel, and don't have to believe in vaporous Motorola G5 conspiracies any longer.

Alex

[ddtlm]

alex_ant:

The OS doesn't really matter in the SPEC benchmarks - what matters is the processor in question and the compiler.
That is actually not true... OSX has a high system overhead because of the microkernel. I would expect Linux on a Mac to deliver better SPEC scores than OSX on a Mac.

[alex_ant]

Originally posted by Postal
Assuming that the G4 as we know it is somewhat faster than an equivalently-clocked Athlon or P4 (which should generally be true)
If you don't take AltiVec into account, the G4 is on par with a Pentium III clock-for-clock in SPEC, and somewhat slower (within 100 points give or take, depending on the particular model of processor) than both the Athlon and the P4 clock-for-clock. I think the notion that the G4 is more efficient than competing x86 chips is a popular misconception. The G4 has architecturally never been significantly updated except to make way for higher clock speed.

I agree with your conclusion though, that this new chip looks nice and that it will be way faster than the G4.

[alex_ant]

Originally posted by ddtlm
alex_ant:

That is actually not true... OSX has a high system overhead because of the microkernel. I would expect Linux on a Mac to deliver better SPEC scores than OSX on a Mac.
I thought what OS X had was a cross between a microkernel and a monolithic kernel. You would know better than I, though.

I would expect Linux on a Mac to deliver better scores than OS X on a Mac in ANYTHING. :)

[ddtlm]

alex_ant:

I thought what OS X had was a cross between a microkernel and a monolithic kernel.
Whatever it is, I don't like it as much as the Linux kernel.

If you don't take AltiVec into account, the G4 is on par with a Pentium III clock-for-clock in SPEC
This appears to only be true for integer SPEC. The G4 appears to be well behind per-clock in float SPEC, when not using AltiVec (I do not have any idea how much AltiVec will help, because it is the nature of SPEC that all such things must be done by the compiler and not by hand).

Anyway, I once thought the G4 had a nice IPC advantage but I no longer believe that. I was especially dissapointed when I first realized that the G4 is an in-order-execution CPU, whereas Athlons and P4s can execute instructions out of order. (And so can the Power4.) AFAIK, the most powerful in-order CPU in the Sun UltraSparc, which was listed in those SPEC results I've linked. Note that it was not esecially good. Note also that the other SPARC CPU was by Fujitsu, and it does out-of-order execution and is much more powerful overall.

[ibjoshua]

Originally posted by alex_ant
I would expect Linux on a Mac to deliver better scores than OS X on a Mac in ANYTHING. :)

i take issue with that!

i would expect OS X to deliver better scores in usability and style
:)

i_b_joshua

[beatle888]

Originally posted by MacCoaster

You might want to research more indepth before posting outlandish fallacies.

jeeze what a drama queen.

[Nipsy]

Originally posted by alex_ant

If you don't take AltiVec into account, the G4 is on par with a Pentium III clock-for-clock in SPEC, and somewhat slower (within 100 points give or take, depending on the particular model of processor) than both the Athlon and the P4 clock-for-clock. [/B]

The PIV, clock for clock, is slower than then the PIII, so how could a G4 be on par with the PIII, and slower than the slower than PIII PIV?

[Choppaface]

wait so everything's going 64 bit? I don't get this...apple's going to have two gigantic changes in less than..what 5 years? isn't the jump to a 64 bit platform going to be equal to, if not greater, than the jump from OS9 to OSX?

[nixd2001]

Originally posted by Choppaface
wait so everything's going 64 bit? I don't get this...apple's going to have two gigantic changes in less than..what 5 years? isn't the jump to a 64 bit platform going to be equal to, if not greater, than the jump from OS9 to OSX?

Nope. If not for Apple Marketing, you might even miss it.

[scem0]

You wouldn't know about the change, but you would definitely notice the speed difference.

[robguz]

Originally posted by ddtlm
robguz:


Haha! Not only did I list the IPC as an assumption which clearly means I am aware that it was one, but I didn't say a damn thing about OS's or about Apple. Woot! You lose! :)

Actually, you did implicitly by relying on Spec benchmarks. And of course you're igoring the other assumptions you made. Woot? Shouldn't you be going back to your 4th grade classroom now?

[Mr. Anderson]

Ok, so I missed this this weekend, but this morning trying to read the article all I got was a bunch of ads and no article - anyone else with this problem?

[Telomar]

Originally posted by dukestreet
Ok, so I missed this this weekend, but this morning trying to read the article all I got was a bunch of ads and no article - anyone else with this problem?

Still works for me and I'm not sure if this is somewhere earlier in the thread but here is the article just for you.

IBM server chip seen slimmed down for Apple Macs
Reuters, 10.13.02, 3:49 PM ET

(TRANSMISSION EMBARGO UNTIL 12:01 am EDT/0401 GMT)

ARMONK, N.Y. (Reuters) - International Business Machines Corp. Monday announced a microchip for personal computers that will crunch data in chunks twice as big as the current standard and is expected by industry watchers to be used by Apple Computer Inc.

Apple was not available to comment, and IBM declined to comment on which PC makers would use the chip, but its plans would mark a change for the industry, which has emphasized the importance of the speed of a chip rather than its ability to handle heavy workloads.

IBM said its new PowerPC chip would go into production late next year and process 64 bits of data at a time at 1.8 Gigahertz, or 1.8 billion cycles per second.

The microchip is the brain of a computer, although personal computer chips now come only in 32-bit flavors, which are tuned to do light workloads fast but cannot handle heavy duty server chores.

Chekib Akrout, vice president of IBM microprocessor development, said big databases and the Internet challenged PCs: "This is the time to introduce a 64-bit machine capable of being used on a desktop," he said in a telephone interview.

An industry source said Cupertino, California-based Apple would use the chip in its Macintosh computers.

That could catapult Apple, long dogged by the belief its chips are slower than those produced by Intel Corp., to the technological head of the class and put to rest speculation it was considering moving soon to the Intel platform.

The fastest of the current generation of PowerPC chips in Macintosh computers runs at 1.25 Gigahertz, while the top Intel Pentium is 2.8 Gigahertz. Apple says its machines are already more efficient than Intel-based ones.

Advanced Micro Devices Inc., the main competitor to Intel, is also developing a 64-bit chip code-named Hammer that is expected early next year, although it is planned primarily for servers rather than personal computers.


SERVER LITE

IBM said the new PowerPC 970 microchip is a "lite" version of its Power4 chip, which it launched last fall in its sophisticated computer server, code-named "Regatta."

The PowerPC can run 32-bit applications as well as 64-bit ones and is tuned for graphics, like some Intel chips, IBM said. However, it is not designed to run programs written for Intel chips.

The PowerPC has the same energy-saving attributes as the Power4 but uses only one central processing unit, not two.

The chip will be available in the second half of 2003 and be built in IBM's East Fishkill, New York, chip plant, a new facility that is currently doing test-runs and aims to ramp up into production on other chips later this year.

One analyst said the chip's attributes mean it would work well in the professional publishing sector, for high-end graphics and other media-intense tasks.

"This processor would be a great processor for a Macintosh," said Tom Halfhill, an analyst with San Jose, California-based In-Stat/MDR.

IBM, the world's largest computer company, has seen revenue from its microelectronics division dwindle in recent quarters as the semiconductor sector has been hit by a sharp downturn in demand. It has restructured the unit, selling off some assets.

IBM announced earlier this week more details on its new microelectronics design unit, where it will design chips and other electronics for customers, including Sony Corp.

Copyright 2002, Reuters News Service

[alex_ant]

Originally posted by Nipsy
The PIV, clock for clock, is slower than then the PIII, so how could a G4 be on par with the PIII, and slower than the slower than PIII PIV?
I lied (oops), and the G4 is also slower than the PIII clock-for-clock (AltiVec aside), at least according to the SPEC results available which used GCC 2.whatever for the tests. So the G4 (my guess) is probably a bit behind the PIII in integer, and still way behind in fp.

[GPTurismo]

You Morons need to listen to Arn. You can't compare the chips processors speeds. It's all in archetecture, component speeds, system arch, OS, and what the machine is doing.

A) I am running OS X 10.2 on a g3 400 with 256 ram and it is running very damn smooth. Try running XP on a 400 p3 with 256 and see how nice it is.

B) It's not all about megahertz. It's about system optimization, pipes, channels all of that are major factors.

C) Arn is absolutely correct. You can't compare it at all, and you haven't even sat at a mchine using it. I beleive even if you did you would still whine about the megahertz. If you are so worried about megahertz envy jump platforms. I don't care. I am happy with apples equipment.

D) Just shut up your whining. Sheesh. Apple is looking to be the first true 64 bit OS + System out there. and you guys still whine. Even intel and MS are having problems entering the market. The Itanium was horrible, the itanium 2 looks to be flat on it's face since you can get Power4's and sparcs at a fraction of the price....

GPT

[peterjhill]

Sun's premier workstation (http://www.sun.com/desktop/sunblade2000/) is running a 1 GHz 64 bit UltraSparc processor. It is not slow compared to a Pentium either. So, if Apple releases a machine with this processor, you have to ask yourself (in Clint Eastwood accent) will Windows be running on 64 bit processors on the everymans desktop? Will it?

I don't think so. Mac OS X will be 64 bit ready far before windows will be. I would like to see a dual or quad processor server with this cpu.

Oh, another thing to look at, Intel's new server is really awesome. It is like a mainframe in 8 rack units (with two servers linked together), the x series 440 (http://www.pc.ibm.com/us/eserver/xseries/x440.html) .

You can have 16 processors linked together and actually segment it into multiple servers with redundant failover CPUs, 16 mem slots and you can mirror the memory. It takes up to 64 GB of memory in a single 4U chassis.

This is a market I want to see apple enter.

[OSeXy!]

I think the whole MHz (GHz) issue could have a shake-out in the next year.

If the Power4-lite performs as we hope (pray!) it will, it could be that excessive clock speeds will then be seen as a problem rather than a virtue. Who wants a current-hungry, battery-draining, noisy, crash-prone 4GHz chip, when a better designed one at half the speed can do more work more reliably?

Marketing is key here: If a better benchmarking system ever becomes popular, people will know to respect the chip which achieves the highest power at the lowest clockrate.

In general, I hope apple sticks to building its machines around the most efficient chips - not the fastest clockrates. Particularly as the markets for servers, on the one hand, and laptops, on the other, continue to increase...

[kenohki]

Sun's premier workstation (http://www.sun.com/desktop/sunblade2000/) is running a 1 GHz 64 bit UltraSparc processor.
Now now, it's 1050MHz so that's actually 1.05GHz for the UltraSPARC III Cu. ;) Sorry, everyone else was being so damn nit-pickey that I had to have a little fun. Seriously though, it's a nice chip and they're working with TI to have it at 1.2 GHz soon.


I thought what OS X had was a cross between a microkernel and a monolithic kernel.

Yes, Apple has done some work to make it more monolithic in design than a true MACH microkernel. It cuts down on some of the message passing between traditional kernel functions that are split off into "user level" processes in MACH. Because we're working with small system images for the Macintosh, MACH tends to be a little too chatty.


Okay, so now for my 2 cents worth...

One thing I think people are overlooking in this discussion is that this processor will have a SIMD unit. We don't know how that is going to affect performance. IBM stated that this processor is 8-way superscalar. Does that include instructions for the VMX unit or no? Do you basically get a POWER4 core plus VMX?! (Imagine the performance coming off that with Apple's Velocity Engine optimizations along with keeping VMX fed with a 6.4GB/s bus.) I don't think we'll really know the answers to some of these questions until the microprocessor forum but even so, the chip should be very impressive.

[ewinemiller]

Originally posted by Nipsy


The PIV, clock for clock, is slower than then the PIII, so how could a G4 be on par with the PIII, and slower than the slower than PIII PIV?

That's not quite true anymore, at the time the P4 was released the current PIII did out run it, but now it depends on the model. Of my three PCs generally slowest to fastest clock for clock are PIII Coppermine (the ~733s to 1ghz), P4 northwood 533mhz bus, PIII Tualatin (newer PIII server or mobile processors in the 1.0 to 1.4 ghz range). So you see they can be faster and slower, depends on the model and/or what task your currently working on.

[chibianh]

well, in case anyone is interested, IBM's press release of the chip.

http://www-3.ibm.com/chips/news/2002/1014_powerpc.html

[Gus]

Ok, IBM is set to announce this chip today, and the "real" media (Reuters, AP) have published articles speculating that Apple will use this chip and as such, AAPL stock is up .50 this morning. What kind of a drop do you think we'll see if it is revealed that Apple will NOT use this chip. I think the stock will drop back down into the low $13 range. I hate how Wall Street works. Great company+Great products+pretty consistent profit=downgrade in stock every quarter?

Gus

[ryan]

Originally posted by chibianh
well, in case anyone is interested, IBM's press release of the chip.

http://www-3.ibm.com/chips/news/2002/1014_powerpc.html
This is what I like from the announcement:

The chip incorporates an innovative communications link, or "bus," specially developed to speed information between the processor and memory. Running at a speed of up to 900 megahertz, the bus can deliver information to the processor at up to 6.4 gigabytes per second, to help ensure that the high-performance processor is fed data at sufficient speeds.

The problem of slower buses are painfully evident in the Mac and are also felt in your everyday PC. One of the reasons "true" workstations form a company like Sun feel so fast is due to the incredible amount of data their buses can move between the processor and memory.

[Gus]

"The chip incorporates an innovative communications link, or "bus," specially developed to speed information between the processor and memory. Running at a speed of up to 900 megahertz, the bus can deliver information to the processor at up to 6.4 gigabytes per second, to help ensure that the high-performance processor is fed data at sufficient speeds."

THAT sounds promising!

Gus

[OSeXy!]

Also sounds good:

"While supporting 64-bit computing for emerging applications, the PowerPC 970 also provides native support for traditional 32-bit applications, which can help preserve users’ and developers’ software investments..."

Sure sounds like it's 'Made for Apple'. Wish they had talked more about its vector unit, though.

[ogun7]

This chip has to have VMX (or some facsimilie thereof) in order for Apple to use it, right?

[DharvaBinky]

I find it amusing that so many people are, here of all places, are completely obsessed with MHz... Very funny. I look at it like this:

Which would be more efficient at getting work done?

Emptying a pool, 1 bucket at a time, but do the buckets very fast? Or emptying a pool, 1 40-gallon drum at a time, but done slightly slower.

Obviously, the drum will empty the pool faster. A well designed 64-bit CPU at a lower speed will do much better than anything Intel has for the desktop *or* server arena.

AMD and Apple really need to team up and come up with a meaningful speed rating (not AMD's PR "QuantiSpeed" bull...). I can't remember the name of the thingee, but there was a processor I recall hearing about that ran at 100MHz, but was inexpensive and massively parallel, and went really *really* fast.

I think, in the end, Apple would do well to even *reduce* clock speeds if possible. Don't even *try* to compete on that basis. Maybe even put out promo literature that refers to P4s as incredibly expensive EZBake Ovens... If they note "real world" performance specs, not stuff Harry Homeowner doesn't use like Photoshop UnSharp filters, and talk about how their "advanced design" gives them better performance at lower clock, rather than just generate lots of heat, they'd be better off. But Apple is the passive/aggressive co-dependant marketeer... <shrug>

Name names, bust heads, fire whatstheirnamewhowrotetheswitchads. Make funny ads. Hire guys that write beer commercials to do Apple marketing.

"Apple Macintosh. Faster, Sexier, and Easier. Unlike Girls. True."

;)

Binky

[DaveGee]

Originally posted by ogun7
This chip has to have VMX (or some facsimilie thereof) in order for Apple to use it, right?

Correct answers would be NO

But this new CPU does have VMX so the first part of the question isn't really relevant and as for the 2nd part... Apple's software OS X etc runs on G3 class processors right? So AltiVec isn't required either.

All of that being said...

I have a feeling none of this really matters because the VMX unit in the GPUL will play just fine with AltiVec code. From everything I've read...

Dave

[DharvaBinky]

Originally posted by ogun7
This chip has to have VMX (or some facsimilie thereof) in order for Apple to use it, right?

I think the reuter's line that says "optimized for graphics" imples a vector processor. That would be a way for non-techs to comprehend what SIMD is...

<grin>

Binky

[ogun7]

What's really sad about this whole MHz matter is that most of us REALLY care what the PC weenies think about Macs. I mean, when you see the reaction on most people's faces when you boot up X on a Powerbook, they're already hooked. so the "elegant" interface is what really matters most. Speed is only important really when it comes to people like me who study animation and visual effects and 99% of the folks in school look at you like you need to be committed when you say you want to use a Mac for 3D. ( The IT department knows better, though...)

[Jookbox]

Originally posted by DharvaBinky


I think, in the end, Apple would do well to even *reduce* clock speeds if possible. Don't even *try* to compete on that basis. Maybe even put out promo literature that refers to P4s as incredibly expensive EZBake Ovens... If they note "real world" performance specs, not stuff Harry Homeowner doesn't use like Photoshop UnSharp filters, and talk about how their "advanced design" gives them better performance at lower clock, rather than just generate lots of heat, they'd be better off. But Apple is the passive/aggressive co-dependant marketeer... <shrug>


Binky

no offense but i think you're nuts. you're thinking of ways for apple to dupe the public. you would rather have them spend their time doing that then actually make their computers faster? ghz is not a myth. the upcoming 3ghz p3 will be faster than the dual 1.25 mac. some of you people are coming up with very creative ideas on being complacent.

[Jookbox]

Originally posted by ogun7
What's really sad about this whole MHz matter is that most of us REALLY care what the PC weenies think about Macs.

so what's the difference between a pc weenie and a mac weenie? :rolleyes:

[ogun7]

Originally posted by Jookbox


no offense but i think you're nuts. you're thinking of ways for apple to dupe the public. you would rather have them spend their time doing that then actually make their computers faster? ghz is not a myth. the upcoming 3ghz p3 will be faster than the dual 1.25 mac. some of you people are coming up with very creative ideas on being complacent.

I think he's nuts too, but it's not being complacent to want Apple to market the 'Megahertz Myth' harder. I want and need faster chips for 3D and digital video, but I stay Mac for the efficiency of the OS

[kenohki]

Originally posted by Jookbox


no offense but i think you're nuts. you're thinking of ways for apple to dupe the public. you would rather have them spend their time doing that then actually make their computers faster? ghz is not a myth. the upcoming 3ghz p3 will be faster than the dual 1.25 mac. some of you people are coming up with very creative ideas on being complacent.

So what. A 900MHz Itanium 2 would wipe the floor with both of those chips. People who focus on an oscillation frequency rather than the bigger picture of system design are just playing into the marketing department's hands. If you think Apple would be duping the public by doing something like that, just wait till Intel starts their marketing spin for Itanium or a slower clocked P5. They'll have a lot of explaining to do to Joe Consumer.

End game is that GHz is a piss poor rating to put on a box as a performance indicator.

[ktlx]

Originally posted by DharvaBinky
I find it amusing that so many people are, here of all places, are completely obsessed with MHz... Very funny. I look at it like this:

Which would be more efficient at getting work done?

Emptying a pool, 1 bucket at a time, but do the buckets very fast? Or emptying a pool, 1 40-gallon drum at a time, but done slightly slower.

Obviously, the drum will empty the pool faster. A well designed 64-bit CPU at a lower speed will do much better than anything Intel has for the desktop *or* server arena.

Your analogy does not work correctly for most applications. It assumes that the work is nearly continuous and is not made up of certain logical atomic units.

If you are primarily working with image manipulation on 32-bit pixels, a 64-bit processor does not bring anything to the table other than the ability to manipulate images larger than 4GB more efficiently. A 64-bit processor does not perform two 32-bit arithmetic operations per clock cycle.

A 64-bit processor does bring a lot to the table for anything that consumes large amounts of memory or lots and lots of long integer or floating point calculations. For chunking on a 300MB TIFF 8-bit color image or running iMovie, you probably are not going to see any improvements beyond those provided by the large cache and newer design.

I imagine the 3D rendering guys and movie producers have to be eagerly anticipating Apples based upon this chip. For most of us using PowerMac G4s today, the Pentium 4 will still offer more useable horsepower.

[backspinner]

"Running at a speed of up to 900 megahertz, the bus can deliver information to the processor at up to 6.4 gigabytes per second"

talking about FSB! that number sounds super cool and 6.4 G bytes comes indeed down to 64 bit wide at 860 MHz...

[Roger1]

I didn't see this link posted. Probably just a copy of the same announcement. Enjoy
http://reuters.com/news_article.jhtml?type=technologynews&StoryID=1571983

[porovaara]

Originally posted by GPTurismo
A) I am running OS X 10.2 on a g3 400 with 256 ram and it is running very damn smooth. Try running XP on a 400 p3 with 256 and see how nice it is.
D) Just shut up your whining. Sheesh. Apple is looking to be the first true 64 bit OS + System out there. and you guys still whine.

I do A all the time at work, it is fine. And much faster in day to day activities than the cube I use at the same time. Sorry it is just a fact.

As to B linux and FreeBSD will both have full 64 bit version on Itanium and Hammer at least 9 months before this hits the streets.

[ddtlm]

kenohki:

So what. A 900MHz Itanium 2 would wipe the floor with both of those chips. People who focus on an oscillation frequency rather than the bigger picture of system design are just playing into the marketing department's hands. If you think Apple would be duping the public by doing something like that, just wait till Intel starts their marketing spin for Itanium or a slower clocked P5. They'll have a lot of explaining to do to Joe Consumer.
Since when will Intel be marketing a Itanium II system starting at like $7000 to $10000 to "Joe Consumer"? And why does everyone here think that the P5 will be clocked lower, when all indications are that the "P5" will simply be a modified P4 on 90nm process tech? I swear, everything about P4's and Intel is distorted to suit people's wildest fantasies and their deep insecurities about current Mac hardware.

[ddtlm]

Oh, and for those people thinking Apple is pioneering this whole 64-bit thing, or even those that think it will be the first 64-bit desktop, you should look at this Sun computer:

http://www.sun.com/desktop/sunblade150/

Similar models have been available for a year or two, starting at about $1000. And lets not forget what porovaara said about 64-bit Linux on Hammers and Itaniums... quite true.

[MacBandit]

There are way too many assumptions going on here. The facts as I see them are.

IBM PowerPC 970 at up to 1.8Ghz by second half of next year.
It will support up to 6.4GB/s throughput on the bus.
It is a 64bit processor with backwards 32bit support.
It does have VMX coprocessor on board.
We do not know if Apple even plans on using this processor.

What needs to be said is we do not know how fast this processor is. We do know that clock cycles are not the end all speed factor of any cpu. Also to anyone claiming how fast this processor is, how do you know? This is not a Power4 it is based on it but it is different.

What is also fact is in the next year or so cpu speeds are going to take a tremendous plunge and 1.8Ghz is not that slow as compared to what is going to be on the cutting edge in a year or so. Intel will be marketing a faster processor then the current 3Ghz P4s but the same processor will also have a much lower frequency.

I also believe that Motorola has a new cpu hidden up there sleeves that will be released early next year. This is in my oppinion the processor that Apple will be using come July.

[ddtlm]

MacBandit, others:

Where in the heck is this low-clocked Intel chip rumor coming from? I know that there is going to be a new chip for latops that clocks lower, but it is not the Pentium 5.

Someone needs to provide linkage to back up this whole low-clock Pentium 5 thing.

[alex_ant]

Originally posted by MacBandit
We do not know if Apple even plans on using this processor.
We do now (http://forums.macrumors.com/showthread.php?s=&postid=170196#post170196)

[MOM]

Question to somebody who may know: If Apple wanted to boost Xserve for a more demanding market (Hollywood-remember shake etc), could they use the Power4 chip (not the light version) with little trouble? MOM

[alex_ant]

Originally posted by MOM
Question to somebody who may know: If Apple wanted to boost Xserve for a more demanding market (Hollywood-remember shake etc), could they use the Power4 chip (not the light version) with little trouble? MOM
They'd need to completely redesign the motherboard, and the cost would go way, way up, and they wouldn't be able to fit the Xserve in a 1U form factor anymore, and the Power4 doesn't have a SIMD unit, so... not really. :)

[ewinemiller]

Originally posted by GPTurismo
A) I am running OS X 10.2 on a g3 400 with 256 ram and it is running very damn smooth. Try running XP on a 400 p3 with 256 and see how nice it is.

Originally posted by porovaara
I do A all the time at work, it is fine. And much faster in day to day activities than the cube I use at the same time. Sorry it is just a fact.


And I'm running XP on a 366 PII, w/ 256meg, runs fine, actually a little snappier than windows 2k was.

[Rocketman]

Originally posted by Roger1
I didn't see this link posted. Probably just a copy of the same announcement. Enjoy
http://reuters.com/news_article.jhtml?type=technologynews&StoryID=1571983

With all this banter over how painfl it is to wait, and how Intel is stomping all over Motorols'a Mhz, and how people are talking anbout running out and buying a pee cee and all, I thought a brief and likely ignored bit of insight would be helpful.

People buy macs for reasons mainly OTHER than pure speed. If the computer does what YOU want it to do, the slight premium you pay compared to those chop shiops that make up pee cee kits to order, is more than made up by system reliability, component parts quality, and more than double the service life.

As with any computer product, there will always be a twice as good computer a short year away. But it is a year away. To deal with reality you have to buy the best computer you can afford when it is time for YOU to buy one.

Rocketman

[Todd H]

Looks like it's time to finally start saving for a Mac. 64-bit OSX. Mmmmmmmm sweet!

I've been using a Windows PC for a while now. Can't say I've been too impressed, especially with Microsoft. I've been wanting to get a Mac, but the slowness of the processor has made me think twice. No longer. Hopefully this time next year I'll be the proud owner of a new Mac.

[Akira]

I don't know if someone has posted this (didn't feel like reading 4 pages) but Apple will use the IBM processors.
http://www.ananova.com/news/story/sm_689753.html?menu=news.technology

Anyway, I hope the PPC 970 will live up to the expectations, I mean 1.8 Ghz (probably with larger pipelines -> less power in one clockcycle) in late 2003 isn't really mindblowing. At the end of 2003, Intel's chip will probably run at 4-5 Ghz.

[CRMarvin42]

Through out this post people have been refering back to benchmarks of the chip and inane analogies when refereing to MHz comparisons.

It really doesn't matter which "chip" is faster. It's the entire experience that matters. You don't buy a mac and run windows or by a Dell and run Mac OSX. The system is a package deal.

If I could accomplish everything I need to on a PC without pulling my hair out I would still use PC's but the fact is that PC's with higher clock speeds waste a lot of that processing power on an operating system that needs to be completely overhauled.

I guess i'm trying to say that our chips may not be the fastest on the market, and depending on your measurement we may not even be in the same ball park. But, in the end that really doesn't matter yet for the customers that apple is courting at this point in time

[DharvaBinky]

http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=01M98653

Does this page, directly from motorola, not say that the G4 7455 is a 133MHz FSB 1GHz chip? Everytime anyone says that the 1.25s from Apple are overclocked, they get their ***** jumped by people insisting that it isn't...

<shrug> Just adding fuel to the fire... ;)

Binky

[Akira]

Apple does need a faster processor, Video apps, 3D apps, they all need raw speed. Apple's looking for a new processor at IBM, because they need a faster one, and IBM has a solution.

Still, they have quite a large gap to fill.

I know Mhz comparison don't mean *****, but 1.8 vs 5, you can't just wave that away by saying Mhz don't mean *****.

[dethl]

From: http://www.sfgate.com/cgi-bin/article.cgi?f=/news/archive/2002/10/14/financial0309EDT0011.DTL


In fact, Microsoft Corp. has yet to release a 64-bit version of Windows that will run on AMD's Hammer chips


ROFLMAO!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!


I can't wait for this to hit the fan along with the megahertz myth!

[locovaca]

Originally posted by DharvaBinky
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=01M98653

Does this page, directly from motorola, not say that the G4 7455 is a 133MHz FSB 1GHz chip? Everytime anyone says that the 1.25s from Apple are overclocked, they get their ***** jumped by people insisting that it isn't...

<shrug> Just adding fuel to the fire... ;)

Binky

You have proved exactly what you were trying to disprove. The 1.25's operate on a 166 mhz bus, which is 33mhz above the spec. 166x7.5 ~ 1.25 GHz. Even if it isn't a total clock speed overclock, it is a FSB overclock.

[eric_n_dfw]

Originally posted by locovaca
You have proved exactly what you were trying to disprove. The 1.25's operate on a 166 mhz bus, which is 33mhz above the spec. 166x7.5 ~ 1.25 GHz. Even if it isn't a total clock speed overclock, it is a FSB overclock.
Acutually, I think they were trying to say that the 1.25's are overclocked.

My question is, who the heck cares? If it's waranted by Apple. Then what's the difference?

[scem0]

I don't think Intel will be at 4-5 GHz a year from now. I am thinking 3.8-4.3 GHz. We are getting a really fast processor a year from now, so who cares about Intel. Yeah, their chip might be faster, but nobody should care because macs have everything else going for them, and Intel processors will probably be only a tiny bit faster if faster at all.

[beatle888]

Originally posted by Todd H
Looks like it's time to finally start saving for a Mac. 64-bit OSX. Mmmmmmmm sweet!

I've been using a Windows PC for a while now. Can't say I've been too impressed, especially with Microsoft. I've been wanting to get a Mac, but the slowness of the processor has made me think twice. No longer. Hopefully this time next year I'll be the proud owner of a new Mac.


have you tried the new duals? a dual with
1gig of ram would be awesome in my mind.
that would be great for photoshop especially
with a RAID for the photoshop scratch disk.

[beatle888]

Originally posted by Akira

I know Mhz comparison don't mean *****, but 1.8 vs 5, you can't just wave that away by saying Mhz don't mean *****.

just read your own statement...you said it
yourself..."I know Mhz comparison don't mean *****,"

mhz is an inefficiant way of getting something
done fast. it's now time to look at the processor
in other ways....we know how to push a
processor to go really fast with pure clock
cycles...now it's time to see how a processor
can be more efficient by doing more with
slower clock cycles.

i see high mhz processors as something thats
working extrememly hard to process data.
i would rather see a processor work easier
and processing the same amount of data
or more.

[kenohki]

Originally posted by ddtlm
kenohki:


Since when will Intel be marketing a Itanium II system starting at like $7000 to $10000 to "Joe Consumer"?

Intel has said Itanium is their next generation platform for the next decade or some catchy buzzphrase like that. Eventually (though not in the next year or two, but eventually) Intel is going to get consumers onto IA-64. They've implicitly stated x86 is done by their actions. It's a drain to have to maintain two desktop/server architectures as far as R&D goes. Besides, they just charge that much because they CAN. Put the economies of scale of the x86 line to the Itanium and they'll come down in price. It's just a matter of time to allow adoption. (HP sure is betting on it.)

[kenohki]

Originally posted by ddtlm
Oh, and for those people thinking Apple is pioneering this whole 64-bit thing, or even those that think it will be the first 64-bit desktop, you should look at this Sun computer:

http://www.sun.com/desktop/sunblade150/

Similar models have been available for a year or two, starting at about $1000. And lets not forget what porovaara said about 64-bit Linux on Hammers and Itaniums... quite true.

Oh please. That's not a desktop. How many housewives are sitting at home running Solaris with their HotJava browser and zilch consumer software support? They could probably hardly handle the installation of Solaris much less figure out CDE. That's a low end workstation for developers who don't need lots of horsepower but need to work in the Solaris environment.

Barring some of the stuff used in game consoles, the PowerPC 970 and Athlon branded Hammers are going to be the first 64-bit home/desktop targeted chips.

[MacBandit]

Originally posted by ddtlm
MacBandit, others:

Where in the heck is this low-clocked Intel chip rumor coming from? I know that there is going to be a new chip for latops that clocks lower, but it is not the Pentium 5.

Someone needs to provide linkage to back up this whole low-clock Pentium 5 thing.

It has nothing to do with the Pentium line of chips. It's the Intel Itanium that has been held back. It has barely been able to achieve of 1Ghz. It may be approaching 2Ghz now though. I haven't read about it in about 6months.

[MacBandit]

Originally posted by alex_ant

We do now (http://forums.macrumors.com/showthread.php?s=&postid=170196#post170196)


Does anyone actually know what quantity they are buying these chips in. Are they going to buy them for production or just testing. Also these chips will not even be produced until late next year. So the question is still there will Apple use them. Just because they are buying them doesn't mean they plan on going in to full production with them. They could just be being used for testing of possibilities of what they can do if there current plan fails completely. Remember Apple has dozens of different test mules running all the time and sometimes the final product doesn't even come from them.

Still massive assumptions going on building false hype and hope.

[scem0]

But Intel has had some 'difficulties' as of late, so I think that will make apple have a better chance of having a faster processor.

[MacBandit]

Originally posted by DharvaBinky
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=01M98653

Does this page, directly from motorola, not say that the G4 7455 is a 133MHz FSB 1GHz chip? Everytime anyone says that the 1.25s from Apple are overclocked, they get their ***** jumped by people insisting that it isn't...

<shrug> Just adding fuel to the fire... ;)

Binky


So here's the the problem.

A) How old is this page and does it correctly represent the current production?

B) The reason people get there **** jumped on is if you refer to overclocking in reference to a manufacturor you sound like an idiot.

Overclocking is a term used to describe end user actions on a processor and can not and should not be used in describing a manufacturor. If a manufacturor is capable of getting 3Ghz out of a 500Mhz chip using refrigeration then the chip tests good at 3Ghz and thusly is a 3Ghz chip. The speed that a processor is rated at has nothing to do with manufacturoring its simply the speed the chip test good at with the motherboards cooling properties.

[jettredmont]

Originally posted by arn


If you must focus on numbers, see this link (http://www.aceshardware.com/read_news.jsp?id=60000436) ddtlm posted. And look at the Spec scores of a 1.3ghz Power4 compared to a 2.8ghz Pentium.

arn

A few notes about the SPEC numbers:

1) While the SPECint numbers are quite a bit higher for Dell's Intel P4 implementation than any of the 64-bit brethren, note that the FP numbers for all Intel (and, inexplicably, AMD ... some say this has to do with the Intel-provided compiler used to compile the SPEC benchmark for the AMD chip ...) are significantly lower than any of the 64bit chips enjoy.

2) Note that only a single core is tested by SPEC tests. The dual cores of the POWER4 are not used. Likewise the pseudo-dual-cored P4 3.0GHz will not be measured with SPEC for future reference. Benchmarks tend to be single-threaded scripted entities, as multi-threaded benchmarks are inherently more complex to create and analyze. Applications and users, however, tend to be multi-threading banshees.

3) While previous versions of SPEC could easily be manipulated by adding L2/L3 cache (Spec92 coulde fit entirely in 4MB cache ...), SPEC2000 is supposedly less manipulatable. However, one should always assume that if the benchmark has a weakness then at least some manufacturers have found it and are exploiting that weakness. For this reason, one must look to "odd" jumps in performance (say, a 10% jump in CPU speed engendering a >10% jump in marks. Also, one must recognize that while SPEC* are aimed at the CPU alone, they are affected by the rest of the system (which is why Dell can get much better numbers than HP or even Intel can from their configurations), a fact which is highlighted by the fact that a large gain in CPU speed base rate with no fundamental rearchitecting engenders a fairly small SPECint rating increase (say, from the Dell 2.4GHz to the Dell 2.8GHz).

4) While summaries are good, I'd prefer to look at the "raw" spec numbers. You can see the per-test numbers for the various platforms below by clicking on the "Disclosures" links to their right:

SPECint:

http://www.specbench.org/osg/cpu2000/results/cint2000.html

SPECfp:

http://www.specbench.org/osg/cpu2000/results/cfp2000.html

Finally, let me just say that a 1.8GHz 64-bit processor in late 2003 is right in the middle/low end of the "GHz rating" game, both on the server and on the desktop. On the other hand, I'm not sure that's particularly good for the "educate consumers on the meaningless of MHz across platforms" goal ... We will, however, have Mac performance (meaning, user-visible performance) much more in line with P4 speed at that time, especially with the vaguaries of ~AltiVec vs SSE instruction optimizations thrown in the mix.

[jettredmont]

Originally posted by Choppaface
isn't the jump to a 64 bit platform going to be equal to, if not greater, than the jump from OS9 to OSX?

It shouldn't have to be. The Power4-lite (970) from IBM will run existing 32-bit code, and I believe it will do so "seamlessly" alongside 64-bit code although I haven't seen that specifically stated yet. This means that OS X and the various applications you have now will run just fine on a 64-bit 970. Upgrading to the "64-bit" version of each component might give you a slight increase in raw speed (a major bump in some cases).

We haven't seen any comparison between Instructions Per Cycle numbers for 32-bit mode of the 970 vs, say, the G3, so I can't say how well 32-bit code will run on a new 1.8GHz 970 versus a 1.8GHz G3/G4 processor. However, note that the real bottleneck of Apples these days tends to be the CPU-to-System Controller bus, and the 970 pops that particular bus way up from what any G4 now supports (IBM G3s support more, but not G4s).

I see people in general maybe upgrading OS X and their "multimedia" apps very soon after getting a 970. I suspect that most other apps will just get upgraded in the "normal" cycle of upgrades.

One can point to the Win95 16-32 bit upgrade debacle and say that this might be as horribly executed. However, I should hope that Apple is smarter than MS was (and all history indicates that Apple does such rollovers far more smoothly than MS). I would be surprised if running a 32-bit app in 64-bit OSX involved expensive mode switching with every shared library or Quartz call, as was the case with Windows.

[astrocity20]

I've always thought that Apple could get the upper hand by being one of the first to go all out 64bit. And since Apple really doesn't have anything to lose on the "mega-hertz myth" it works to their advantage. It'll just look like Macs are getting fast while AMD and Intel PCs are getting slower, and that of course doesn't bode well for them. That along with Microsoft's whole Paladium thing, Apple could be surely on their way up. But sheer speed isn't going to save them, dedicated developer support will, rather than Mac being that other computer we may or may not [insert app or game here] develop for it has to be the main priority. Once that happens Apple will be made.

[locovaca]

Originally posted by astrocity20
Once that happens Apple will be made.


This is the statement that worries me the most. So IBM will start to PRODUCE these late next year. Assuming that no problems occur with the developement of the chip, that the yields are good, and that the other hardware is in place, we can expect to see these in January 2004 hypothetically. What is Apple going to do in the interim? Sure, they'll release slightly faster G4's (I'd guess they'll get up to 1.4 or 1.5), but will people buy them? The danger of this announcement is that now the IBM chips are the next best thing, so why buy a 32 bit processor now when in less than a year you can get a 64 bit one that will be tons faster? Is there a compelling reason to buy a G4 now? Yes, it will most of the software coming in the next 3-4 years, but why pay $3500 for a top of the line 32 bit Powermac when in 1 year you can get a top of the line 64 bit Powermac for probably about the same? Granted, this is the same problem that has always existed for those purchasing a new computer, but with this flury of press releases and the promises of 64 bit, this is much bigger than just a clock speed bump.

While I believe that the new chip will be nothing short of amazing, I want to remind you that, when the Pentium 4 was in developement, it was an amazing chip. It was insanely fast, had tons of cache, and ripped apart any processor that existed (which would've been Athlons and P3s) at the time. However, it was way too big, way too power hungry, put off way too much heat, and in order to make it a consumer item, they had to effectively cripple the things that made it good in order to make it a viable consumer processor. That meants stripping out FPUs, lots of the cache, and other components that left it to be the CPU we know and chastise today. I'm not saying IBM will do this, but we should be wary about setting to high of performance expectations out of this chip. The Power4 we know is not a consumer chip- it's a server chip. To make it viable for the consumer market lots of things will have to go, and we don't know what exactly they will cut. Our best bet would be to not make claims of how much faster it will be, but rather be happy with the fact that it will be faster.

[jettredmont]

Originally posted by ktlx


If you are primarily working with image manipulation on 32-bit pixels, a 64-bit processor does not bring anything to the table other than the ability to manipulate images larger than 4GB more efficiently. A 64-bit processor does not perform two 32-bit arithmetic operations per clock cycle.


Correct. 64-bit is not twice as fast as 32-bit unless you are specifically using it as such. 2+2=4 is calculated just as fast on a 32-bit processor as it is on a 64-bit processor. In fact, it is processed just as fast on an 8-bit processor.

A 64-bit processor #1 gives you the ability to easily perform straight math on 64-bit wide integers (which allows greater unit precision but less range than operating in 64-bit floating point). While 64-bit math in a 32-bit processor generally takes 4+ operations, the same procedure would take one operation on a 64-bit CPU. But then, how often do we deal with 64-bit integers?

However, if you are dealing with masking operations, a 64-bit processor will be twice as fast as a 32-bit processor (4x as fast as a 16-bit processor, etc) because a single instruction will mask off 64 bits (8 sets of 8-bit data, 4 sets of 16-bit, or 2 sets of 32-bit) instead of just 32.

Of course, you also have 64-bit memory access thrown in in case 32-bit access just doesn't give you enough GB of memory at your direct-access disposal. The directly-accessible range of memory for the processor jumps from 4.3 GB to 18 Billion GB (18 Giga-Giga-Bytes?). Which, one would hope, should carry desktop computing through the next decade fairly easily. This doesn't mean that the rest of the system will support 18GGB, just as you still see P4 and Athlon systems out there with 1GB max memory. However, it opens the door for the other system components to step through.


A 64-bit processor does bring a lot to the table for anything that consumes large amounts of memory or lots and lots of long integer or floating point calculations. For chunking on a 300MB TIFF 8-bit color image or running iMovie, you probably are not going to see any improvements beyond those provided by the large cache and newer design.


Well, not sure what you mean by "chunking on a 300MB TIFF", but if there is non-arithmetic data manipulation involved, 64-bit operations would likely be helpful. Running iMovie: no, probably won't help much there, but only because iMovie should already be usint AltiVec for its DCT and vectoring masks on a 64-bit register won't be any faster than what AltiVec already provides.

I imagine the 3D rendering guys and movie producers have to be eagerly anticipating Apples based upon this chip. For most of us using PowerMac G4s today, the Pentium 4 will still offer more useable horsepower.

Well, 64-bit processing power allows most typical server uses to run faster. That is, by-the-buttload data manipulations. No, Mozilla won't load a page faster and Word won't make you type faster because there is a 64-bit processor underneath. However, for many of the tasks where one would want a "faster" computer in general, 64-bit computing will help matters.

[MacBandit]

Originally posted by locovaca
The Power4 we know is not a consumer chip- it's a server chip. To make it viable for the consumer market lots of things will have to go, and we don't know what exactly they will cut. Our best bet would be to not make claims of how much faster it will be, but rather be happy with the fact that it will be faster.

This is the biggest problem as I see it. People are getting way too hyped on this. There is no guarantee Apple will ever sell a product with this chip. There is also no way of knowing the performance of it.

On another note I do believe we will be pleasantly surprised by a new chip from Motorola in a tower by July of next year.

[jettredmont]

Originally posted by ddtlm
kenohki:


Since when will Intel be marketing a Itanium II system starting at like $7000 to $10000 to "Joe Consumer"? And why does everyone here think that the P5 will be clocked lower, when all indications are that the "P5" will simply be a modified P4 on 90nm process tech? I swear, everything about P4's and Intel is distorted to suit people's wildest fantasies and their deep insecurities about current Mac hardware.

P5 aka "Prescott" is not a P4 with a smaller die. It is a rearchitected core, possibly with additional instructions (and some suspect that it will take on Yamhill 64-bit extensions to IA-32). It does, however, look to be designed around a 0.09nm process. You should educate yourself before blathering on. I can see how you would make this mistake, however, as many rumor sites called the P4's 0.13nm core "Prescott". This appears to either have been an error, or Intel has gone to naming all their projects with the same code name. I suspect the former.

Also, note that when Prescott comes out, "Deerfield", Intel's "Desktop 64-bit" processor hopeful, should be debuting at around 1.0 GHz. UltraSPARQ V should be at around 1.8-2.1GHz as well. Which is why I say that IBM 970 should be right in the middle of the 64-bit GHz ratings.

[jettredmont]

Originally posted by DharvaBinky
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7455&nodeId=01M98653

Does this page, directly from motorola, not say that the G4 7455 is a 133MHz FSB 1GHz chip? Everytime anyone says that the 1.25s from Apple are overclocked, they get their ***** jumped by people insisting that it isn't...

<shrug> Just adding fuel to the fire... ;)

Binky

Searchecd for "overclock" on that page. Nothing. It notes that the FSB is 133MHz, but that does not equal "overclocked".

People get jumped all over for saying a processor that the CPU manufacturer has RATED at a specific frequency is overclocked because frankly it is logically impossible. Overclocking means exceeding the manufacturer's rated frequency (by whatever means you choose to employ ... modifying the FSB, changing clock multipliers, spilling liquid caffeine on the motherboard ...). A manufacturer, by definition, can not RATE a processor as "above the manufacturer's rating".

I mean, that's just as logically inconstent as claiming that I might ever be wrong!

[jettredmont]

Originally posted by kenohki


Intel has said Itanium is their next generation platform for the next decade or some catchy buzzphrase like that. Eventually (though not in the next year or two, but eventually) Intel is going to get consumers onto IA-64. They've implicitly stated x86 is done by their actions. It's a drain to have to maintain two desktop/server architectures as far as R&D goes. Besides, they just charge that much because they CAN. Put the economies of scale of the x86 line to the Itanium and they'll come down in price. It's just a matter of time to allow adoption. (HP sure is betting on it.)

"Deerfield" (Q3 2003) is Intel's projected 64-bit desktop PC. Some suspect that the P5 may have 64-bit extensions akin to AMD's Hammer, but that is unconfirmed (and, if unleashed, would radically alter Intels forward-looking CPU map).

So, using many of the same techniques IBM is applying to POWER4, Intel will have an "Itanium 3 lite" out at about the same time. However, it will be running at ~1GHz, which isn't much faster than the current Itanium line (although I3 might be more efficient than I2 so you never know).

IBM 970 is not a "home run" "game winner" for Apple, but it sure puts us back in the performance game.

[deepkid]

I know that many of us have been championing for a more powerful line of powermacs, but what if this chip is destined for a completely new power product other than the powermacs or X-serve? (Assuming that sources are correct that Apple will use the new PowerPC 970 chips in the first place.)

Think about it. Apple has bought some powerful and well-respected software/companies in the last year or so, not limited to Shake, Nothing Real and Emagic.

What if they are going to create some kick ass pro solution that has nothing to do with the powermacs?

A sort of Borg of pro macs...

If that's the case, I doubt that the price tag would be within comfortable reach of the typical prosumer who buys powermacs.

I'm waiting to buy a powermac, but we could be thinking in the wrong direction.

[kenohki]

Originally posted by jettredmont


"Deerfield" (Q3 2003) is Intel's projected 64-bit desktop PC. Some suspect that the P5 may have 64-bit extensions akin to AMD's Hammer, but that is unconfirmed (and, if unleashed, would radically alter Intels forward-looking CPU map).

So, using many of the same techniques IBM is applying to POWER4, Intel will have an "Itanium 3 lite" out at about the same time. However, it will be running at ~1GHz, which isn't much faster than the current Itanium line (although I3 might be more efficient than I2 so you never know).

IBM 970 is not a "home run" "game winner" for Apple, but it sure puts us back in the performance game.

Yeah, that was my original point of the post in that Intel marketing may have some consumer education to do with a desktop Itanium debuting to the public at 1GHz. They've let the MHz myth run rampant, to their gain of course. They may have to backpedal a bit.

I agree that Yamhill would dramatically alter things. So far though, Intel has been denying any intention to use it. Guess that will depend on if Hammer puts the hurt on them.

[MacBandit]

Originally posted by deepkid
I know that many of us have been championing for a more powerful line of powermacs, but what if this chip is destined for a completely new power product other than the powermacs or X-serve? (Assuming that sources are correct that Apple will use the new PowerPC 970 chips in the first place.)

Think about it. Apple has bought some powerful and well-respected software/companies in the last year or so, not limited to Shake, Nothing Real and Emagic.

What if they are going to create some kick ass pro solution that has nothing to do with the powermacs?

A sort of Borg of pro macs...

If that's the case, I doubt that the price tag would be within comfortable reach of the typical prosumer who buys powermacs.

I'm waiting to buy a powermac, but we could be thinking in the wrong direction.

I've heard of this and this could be a very real possibility. Apple doesn't have any contenders in the super high end user market where 4Gigs of Ram is not enough.

Has anyone else thought about them marketing a computer based on the X-Serve that you can cluster. This was discussed at the latest MacWorld in France. Imagine putting 40 or more X-Serves in a server tower and having the computing power of all of them combined. Actually there is no limitation to how many except for money and space. Which for the kind of people who would need it would not be a problem for either.

Also if Apple does promote clustering what's preventing a technology that allows the buses in each computer from communicating with each other at full speed over short distances say inside a server tower. The computing power of something like this would be truly awesome approaching that of almost any SuperComputer in the world today for a lot less money.

[solvs]

Anyone who says that Apple and Motorola are overclocking the G4, obviously don't understand what the term "overclock" means, and are therefore to be thought of as idiots.

The 1.25s are 7455A chips, that SAY ON THEM 1.25 GHz. They are rated to run at 1.25 GHz, and therefore ARE NOT overclocked. Duh. The G4 has long been able to run on a 166.67 MHz FSB, Apple just hadn't implemented it yet.

If you are basing the OC'ing assumption on the fact that Moto hasn't yet updated their fact pages, you are a dumbass and have no idea how this stuff really works. Go read up some more before posting. We who actually know this stuff are tired of your idiocy and mis-use of terms.

Besides, it would be illegal for Apple to OC the G4 and not say anything. And as someone else posted, if Moto says it's a 1.25, it's a 1.25. That's how fast it was rated to run, that's how fast it runs. So shut-up, and drop it already, or you will be mercilessly made fun of (so I ended this sentence on a preposition, sue me).

Anyway, back to the rediculously long thread (at least there aren't a hundred of them). Speed is good. Clock speed doesn't mean much, but more speed is always better. IBM good, bring 'em on.

[jettredmont]

Originally posted by solvs
So shut-up, and drop it already, or you will be mercilessly made fun of (so I ended this sentence on a preposition, sue me).


The correct english, of course, would end "or you will have fun made of you mercilessly." ...

Somehow I don't think slang was ever meant to follow conventional standard English rules ... :)

[solvs]

Originally posted by jettredmont


The correct english, of course, would end "or you will have fun made of you mercilessly." ...

Somehow I don't think slang was ever meant to follow conventional standard English rules ... :)

My way sounded better anyway. Proper English be darned. :D

But I think I made my point (ha, I did it again. Started a sentence with BUT. And I started the last sentence without saying I. And these 2 with AND. I can't be stopped with my bad English).

This thread is just gonna get longer and longer isn't it?

[alex_ant]

Originally posted by MacBandit
I've heard of this and this could be a very real possibility. Apple doesn't have any contenders in the super high end user market where 4Gigs of Ram is not enough.
...SGI...IBM...Linux...
Has anyone else thought about them marketing a computer based on the X-Serve that you can cluster.
A computer based on the Xserve that you can cluster... sounds good! They could call it the Xserve. (Clustering is a capability that's coming in OS X Server which will require no new hardware)
Also if Apple does promote clustering what's preventing a technology that allows the buses in each computer from communicating with each other at full speed over short distances say inside a server tower. The computing power of something like this would be truly awesome approaching that of almost any SuperComputer in the world today for a lot less money.
Interesting, although not as interesting as it would be if you knew what you were talking about. "Our computer's got lotsa buses! Let's link 'em up! Blip, bleep, bloop, done! If it was a supercomputer before, it's a super-duper computer now!"

[guenesis]

n/t

[alex_ant]

Originally posted by deepkid
I know that many of us have been championing for a more powerful line of powermacs, but what if this chip is destined for a completely new power product other than the powermacs or X-serve? (Assuming that sources are correct that Apple will use the new PowerPC 970 chips in the first place.)
I don't believe an ultra-high-end Apple computer is in the cards for several reasons:

- If this chip were going in a new, high-end Apple product, the Power Mac and Xserve without any new chip of their own would shrivel up and die. Apple needs a faster CPU in these products more than they need a high-high-end computer.
- The demands of the markets in which Apple is trying to establish itself would, for the most part, rather have the flexibility and scalability that a clustering-capable solution like the Xserve provides (or will provide in future releases of OS X Server) than have to fork over a huge sum for a relatively non-upgradeable machine.
- A quad-processor Xserve or other 4-CPU rackmount computer would have little advantage over the dual-CPU Xserve if it wouldn't fit in a 1U form factor. And fitting 4 CPUs inside a 1U form factor would be a trick, to say the least.
- This new CPU will be suited to the Power Macs and Xserve. Small die size, probably cheap to manufacture, probably not a ton of cache. This is a CPU for desktop computers, workstations, and low-end servers, as IBM itself states.
- Steve Jobs has historically favored simplified product lines, and has never catered to the highest of high end customers.
Think about it. Apple has bought some powerful and well-respected software/companies in the last year or so, not limited to Shake, Nothing Real and Emagic.

What if they are going to create some kick ass pro solution that has nothing to do with the powermacs?
Why not just make the Power Mac the kick ass solution?

[alex_ant]

Originally posted by jettredmont


The correct english, of course, would end "or you will have fun made of you mercilessly." ...

Somehow I don't think slang was ever meant to follow conventional standard English rules ... :)
As long as we're being picky, it would have been "you will be the one of whom fun will mercilessly be made," or something. :D

[kenohki]

Originally posted by MacBandit

Also if Apple does promote clustering what's preventing a technology that allows the buses in each computer from communicating with each other at full speed over short distances say inside a server tower. The computing power of something like this would be truly awesome approaching that of almost any SuperComputer in the world today for a lot less money.

Sounds like the SGI Origin 3000. http://www.sgi.com/origin/3000/ Just as a side note you may want to check it out. They use what SGI calls NUMAflex. They use separate components called bricks which are connected together in a cabinet to create a nice customized solution. There are processor bricks, i/o bricks, storage bricks, etc. The processors are all connected via switches (much more efficient than busses). Of course, it's a ccNUMA architecture (cache coherent Non Uniform Memory Access) which makes memory management for local memory to each processor a bit easier to deal with. Anyway, maybe give it a look. Sounds similar to what you're talking about.

[MacBandit]

Originally posted by alex_ant

...SGI...IBM...Linux...
[/b]
A computer based on the Xserve that you can cluster... sounds good! They could call it the Xserve. (Clustering is a capability that's coming in OS X Server which will require no new hardware)

Interesting, although not as interesting as it would be if you knew what you were talking about. "Our computer's got lotsa buses! Let's link 'em up! Blip, bleep, bloop, done! If it was a supercomputer before, it's a super-duper computer now!" [/B]

Okay what I meant about competitors is that Apple doesn't compete in that market just poor wording.

OSX server does not allow clustering currently but it's been mentioned that it may be added next year at some point.

What I was refering to with the buses was the main system bus allowing each cpu maximum throuput as if they were each in the same case on the same board.

[Catfish_Man]

Originally posted by alex_ant

I don't believe an ultra-high-end Apple computer is in the cards for several reasons:

- If this chip were going in a new, high-end Apple product, the Power Mac and Xserve without any new chip of their own would shrivel up and die. Apple needs a faster CPU in these products more than they need a high-high-end computer.
- The demands of the markets in which Apple is trying to establish itself would, for the most part, rather have the flexibility and scalability that a clustering-capable solution like the Xserve provides (or will provide in future releases of OS X Server) than have to fork over a huge sum for a relatively non-upgradeable machine.
- A quad-processor Xserve or other 4-CPU rackmount computer would have little advantage over the dual-CPU Xserve if it wouldn't fit in a 1U form factor. And fitting 4 CPUs inside a 1U form factor would be a trick, to say the least.
- This new CPU will be suited to the Power Macs and Xserve. Small die size, probably cheap to manufacture, probably not a ton of cache. This is a CPU for desktop computers, workstations, and low-end servers, as IBM itself states.
- Steve Jobs has historically favored simplified product lines, and has never catered to the highest of high end customers.
[/b]
Why not just make the Power Mac the kick ass solution? [/B]

I'm not so sure that this chip is good for the XServe. There's a reason why x86 1Us are still using P3s. The PPC970 is NOT being designed for low power consumption (according to anything I've heard), so for now it's only going to be useful in the PowerMacs. A little later when they transition it to .09 micron... we might start seeing TiBooks or XServes with it (they're pretty similar, actually, and have pretty similar requirements).

[jadam]

and who reported this story to you MR ARN!$@?????

[Telomar]

Originally posted by Catfish_Man


I'm not so sure that this chip is good for the XServe. There's a reason why x86 1Us are still using P3s. The PPC970 is NOT being designed for low power consumption (according to anything I've heard), so for now it's only going to be useful in the PowerMacs. A little later when they transition it to .09 micron... we might start seeing TiBooks or XServes with it (they're pretty similar, actually, and have pretty similar requirements). You're assuming Apple has no plans to release different servers. Not every rack mountable server comes in a 1U format.

[ddtlm]

Catfish_Man:

You can find 1U dual processor Athlons and Pentium 4 (under the Xeon name) here:

http://www.appro.com/1224X.html
http://www.appro.com/1124.html

Because a single 180nm Power4 core (2 per die) at 1.3ghz produces more or less the same heat (60 something watts) as these chips, it is reasonable to speculate that two 130nm PPC-970's could work in a 1U rack.

[moki]

Originally posted by robguz
Whoopee, the mac world will finally have 1.8Ghz by 2004 (c'mon we all know that late 2003 means 2004 at the earliest-otherwise we'd all be using the real Apollo chips running at 1.6Ghz for the past 6 months). Meanwhile the Wintel world will be at 4Ghz plus. In 2004 we'll get to party like it's 2000!

Actually, no. If you assume that the Wintel world will be moving over to the Itanium, it might interest you to know that the current speed of the Itanium 2 -- which scores quite well on the benchmarks, btw -- is a palty 1ghz.

The people in the Wintel world who have been using ghz ratings as a measure of their ***** size are going to be in for a rude awakening when the Itanium is pushed as mainstream. They very well may be forced to admit that there is something to this whole "mhz myth" after all.

One other thing to keep in mind is that the GP-UL is not just a stripped down, single-core POWER4 processor -- it isn't valid to assume that the GP-UL will simply be a scaled down POWER4 from earlier this year, with no enhancements whatsoever.

Wait until you see the benchmarks from this processor; I think you may be quite impressed.

[MacBandit]

Originally posted by moki
One other thing to keep in mind is that the GP-UL is not just a stripped down, single-core POWER4 processor -- it isn't valid to assume that the GP-UL will simply be a scaled down POWER4 from earlier this year, with no enhancements whatsoever.

Wait until you see the benchmarks from this processor; I think you may be quite impressed.

I also think it would be a big assumption to think that the PowerPC970 is any more then simply losely based on the Power4. I think it's basic architecture might be the same but the overal features will be quite different.

[moki]

Also, if you're interested on the POWER4 technology that the GP-UL is based on, check out this rather in-depth review (it is a bit old, but quite good):

http://www.digit-life.com/articles/ibmpower4/

[mangis]

1.8 x dual = 3.6

very near the P4 in a year's time.

Sounds good to me

[ipiloot]

Mybe it's a wishful thinking. Just maybe. But the way yhe things show out makes me thinking that Power 4 is just an replacement for G4 that Apple will throw in to the low-end machines while high-end machines will be using Moto's G5.

[pretentious]

ipiloot
Mybe it's a wishful thinking. Just maybe. But the way yhe things show out makes me thinking that Power 4 is just an replacement for G4 that Apple will throw in to the low-end machines while high-end machines will be using Moto's G5.

What Moto G5??? Next you are probably going to tell me that they’re more than 12 people working on the G4 at Moto, at least that is how it seems like it w/ how slow they are bringing faster chips.
Give me some references that are better than anything that we have for the Power4 GPUL. I would be amazed if Apple didn't use the Power4 in its high-end machines.

My question is this: Assuming that the rumors are true, that Apple is planning on using this chip, and the rumor OS X has already shown that it can boot on this proccesser, and if it can run well;
Given that, w/ all this hoopla (TV, web) over the possible outcome of such a chip, could IBM up the production for a sooner unveiling of a computer running this? Maybe by MWSF? Apple could toss this into an Xserve, the perfect whipping boy for a bringing out of new 64-bit proccesser for low-end servers and desktops.
The late 2003 might just mean that is when they are planning to go into full production, possibly when Apple decides to bring the Power Macs to 64, w/ a small market outcropping of delivering the Xserve to the high-end of Apple’s market, this could surprise all the nay-sayers enough to consider Apple’s even more

[ibjoshua]

Originally posted by pretentious
Given that, w/ all this hoopla (TV, web) over the possible outcome of such a chip, could IBM up the production for a sooner unveiling of a computer running this? Maybe by MWSF? Apple could toss this into an Xserve, the perfect whipping boy for a bringing out of new 64-bit proccesser for low-end servers and desktops.
The late 2003 might just mean that is when they are planning to go into full production, possibly when Apple decides to bring the Power Macs to 64, w/ a small market outcropping of delivering the Xserve to the high-end of Apple’s market, this could surprise all the nay-sayers enough to consider Apple’s even more
you been sniffing petrol (gasoline) again?

;)

i_b_joshua

[locovaca]

Originally posted by mangis
1.8 x dual = 3.6

very near the P4 in a year's time.

Sounds good to me

I love Dual Processors. After owning a dual system (I have three, not all mac, but three duals), I will never go back to a single cpu system. However, I take very much offense to this idea that you all the sudden have twice the performance if you double the CPUs. THIS IS NOT HOW IT WORKS! ALL A DUAL DOES IS WORKS ON TWO THREADS AT ONCE, IT DOES NOT DOUBLE THE WORK SPEED ON ONE THREAD!

Now that that is done, 2x1.8 is just 1.8. You can't put just a "number" on a SMP system. It is something you have to experience to really enjoy.

And anyway, in case you have forgotten, the P4 can go dual (with the Xeon version), and right now that stands at 2x2.6, so with your math it's 5.2. But, that is not correct, and if you truly do not agree with the megahertz myth, then you won't use this nomeclature. It only promotes the myth that SMP systems are 2x faster (they aren't). SMP systems do promote much smoother multitasking.

[DharvaBinky]

Originally posted by kenohki


Intel has said Itanium is their next generation platform for the next decade or some catchy buzzphrase like that. Eventually (though not in the next year or two, but eventually) Intel is going to get consumers onto IA-64. They've implicitly stated x86 is done by their actions. It's a drain to have to maintain two desktop/server architectures as far as R&D goes. Besides, they just charge that much because they CAN. Put the economies of scale of the x86 line to the Itanium and they'll come down in price. It's just a matter of time to allow adoption. (HP sure is betting on it.)

I think they generally refer to it as their next generation *server* platform for now... <shrug> I can't see IA64 making a break for consumer space any time soon. There is a lot of resistance to the instruction set in the developer community since it not only requires different instructions, but a whole different "kind" of programming. The IA64 ISA is a brute force method to processing, to quote one analyst "Smart Compiler, Dumb Processor".

Additionally, I wouldn't look to see IA64 go anywhere outside of the high-end (for Intel) server market ever. There are confirmed rumors of a project at Intel called "Yamhill" which seeks to produce a 64-bit extensions to the IA32 ISA, similar to the approach that AMD is taking. Intel, ever conflicted about goals beyond pocketbook padding, has stated that AMD's approach (64-bit CPUs that are backward compatible to 32-bit apps) is undesirable since you don't gain 100% of the benefits of 64-bit addressing. However, they can see that software vendor resistance to *completely* rewrite their code (there is no simple "porting" in the Itanic world) may eventually cause them to change their minds.

I would take Intel's anti-crossover rhetoric (while maintaining a project to produce a crossover CPU) as a sign that they've made a bad business decision fomented in the throes of the DotBoom. Now that we're in the DotBomb, IT budgets are slashed, new enterprises are few, and there's little need for a new 64-bit platform when Sun, IBM, and HP all have strong platforms with development history.

Add to this that Sun, IBM, and HP platforms are all RISC based, a programming style familiar to all developers. Given that some computer scientists are estimating an entire *generation* of programmers will be lost when confronted with EPIC (IA64), I don't see IA64 as a threat to apple directly.

Binky

[kenohki]

Originally posted by DharvaBinky


I think they generally refer to it as their next generation *server* platform for now... <shrug> I can't see IA64 making a break for consumer space any time soon. There is a lot of resistance to the instruction set in the developer community since it not only requires different instructions, but a whole different "kind" of programming. The IA64 ISA is a brute force method to processing, to quote one analyst "Smart Compiler, Dumb Processor".

Additionally, I wouldn't look to see IA64 go anywhere outside of the high-end (for Intel) server market ever. There are confirmed rumors of a project at Intel called "Yamhill" which seeks to produce a 64-bit extensions to the IA32 ISA, similar to the approach that AMD is taking. Intel, ever conflicted about goals beyond pocketbook padding, has stated that AMD's approach (64-bit CPUs that are backward compatible to 32-bit apps) is undesirable since you don't gain 100% of the benefits of 64-bit addressing. However, they can see that software vendor resistance to *completely* rewrite their code (there is no simple "porting" in the Itanic world) may eventually cause them to change their minds.

I would take Intel's anti-crossover rhetoric (while maintaining a project to produce a crossover CPU) as a sign that they've made a bad business decision fomented in the throes of the DotBoom. Now that we're in the DotBomb, IT budgets are slashed, new enterprises are few, and there's little need for a new 64-bit platform when Sun, IBM, and HP all have strong platforms with development history.

Add to this that Sun, IBM, and HP platforms are all RISC based, a programming style familiar to all developers. Given that some computer scientists are estimating an entire *generation* of programmers will be lost when confronted with EPIC (IA64), I don't see IA64 as a threat to apple directly.

Binky

Why would you totally need to rewrite your code? If you're working in a high level language and using the Win32 APIs or writing portable *nix code, seems to me that you'd just need tweaking, optimization, rewriting any assembler (which you'd need to do with any architecture change) and a recompile, not *completely* rewriting your code. Add in the .Net framework and you're further abstracted. The Win32 API set was meant to be portable from the start. Thus the versions for Alpha, MIPS, and PowerPC. And Windows exists on IA64 as does Linux and Monterrey. So seems to me like the majority of the hard work has already been done.

Also, I think Intel has themselves backed into a corner. They've invested billions of dollars in IA64. They're not going to jeopardize that with Yamhill unless Itanium is a *complete* failure and AMD puts the hurt on with x86-64. For Intel to give enterprise an option that maintains their investment in their current architecture would seriously hinder adoption and pretty much mean the death of IA64. It would also be a huge financial blunder (investing all that money in IA64) on Intel's part.

[jettredmont]

Originally posted by DharvaBinky

I can't see IA64 making a break for consumer space any time soon.


Intel's roadmap includes Deerfield, a desktop "value" IA-64 processor based on Itanium-3. That doesn't mean that Deerfield will replace P4 (I doubt it will, especially as IA-64 and IA-32 programs can't mix and Banias, IA-32-based mobile processor, is planned to be alive and kicking for a few years). I don't see anything on Intel's roadmap for IA-32 beyond next year's Prescott/P5 and Banias. So, if Deerfield doesn't look to fulfill it's mission, Intel's going to have to revise their roadmaps.


There is a lot of resistance to the instruction set in the developer community since it not only requires different instructions, but a whole different "kind" of programming. The IA64 ISA is a brute force method to processing, to quote one analyst "Smart Compiler, Dumb Processor".


Yes, that is the problem with EPIC/VLIW and compiler-driver Out-Of-Order-Execution (OoOE): it places much more responsibility on the compiler than ever before. Some say compilers will never be smart enough to do the job. On the other hand, if the P4 can do it in hardware with a very short code visibility window, then there is no reason that a compiler shouldn't be able to do as good a job given a near-infinite theoretical visibility window. And that's just the first step. In theory at least, the compiler should be able to do a much better job at determining which instructions can be put out of order than the processor ever could.


Additionally, I wouldn't look to see IA64 go anywhere outside of the high-end (for Intel) server market ever. There are confirmed rumors of a project at Intel called "Yamhill" which seeks to produce a 64-bit extensions to the IA32 ISA, similar to the approach that AMD is taking. Intel, ever conflicted about goals beyond pocketbook padding, has stated that AMD's approach (64-bit CPUs that are backward compatible to 32-bit apps) is undesirable since you don't gain 100% of the benefits of 64-bit addressing.


Yes, the "halfway approach" is undesirable to Intel. The main reason is that IA-32 is an albatross around its neck and Intel would love to cast it off as cleanly as Apple did the 6800x0 instruction set. Remember that IA-32 is an extension of IA-16 which is an extension of IA-8. The chain has to end somewhere, and someone's going to be a bit pissed when it does. Designing IA-64 with a clean seamless IA-32 compatibility would have drastically reduced the potential life of IA-64. So Intel went with a "pure" 64-bit effort to develop what we know as IA-64/Itanium. AMD countered with the shorter-sighted Hammer extensions, which might be enough to get the job done today but likely won't be able to compete a decade from now.


However, they can see that software vendor resistance to *completely* rewrite their code (there is no simple "porting" in the Itanic world) may eventually cause them to change their minds.


Now, I haven't ported anything to IA-64, but my understanding is that the port is fairly straightforward, although today's compilers aren't smart enough to be able to get the OoOE efficiency P4s obtain. The programmer should not have to define "safe" simultaneous streams; that is the compiler's job.

That, of course, assumes that eventually (and before your product X needs to be ported) compilers come around to enough sophistication that they can produce semi-efficient IA-64 instructions.

Personally, I write cross-platform software. On the C-and-above levels of programming, there is no difference between coding for a RISC processor and a CISC instructon set, and there should not be any difference for an EPIC/VLIW instruction set with explicit OoOE. Granted, any instruction set is going to look different at the assembler/machine levels, but there are very few people who have to be proficient at those levels for a given platform (ie, the compiler people, and the game engine people; any C-level programmer should be able to learn enough about the underlying assembly/machine language to tune a bottleneck if absolutely necessary).

Once there is an EPIC/VLIW compiler that does a better job than the P4 does in hardware, "porting" software from IA-32 to IA-64 will be relatively easy for developers. Compared to recent efforts (Y2K, Internet, OS 9 to OS X, etc), this hardly registers as a blip on the screen of programming effort.

On the consumer side, however, one will have to expect consumers to re-buy everything for the new platform. While Apple has a machine-specific packaging and install procedure well in place with OS X, on the Windows side the good folks at InstallShield will have to configure their compiler to install 32-bit code on IA-32 machines and 64-bit code on IA-64 machines so that developers can release dual-platform software. If developers don't release dual-platform software (ie, both install from a single disk, bought at the same price as the old IA-32 software), a consumer revolt is not just likely but almost assured.

[jettredmont]

Originally posted by kenohki


Why would you totally need to rewrite your code? If you're working in a high level language and using the Win32 APIs or writing portable *nix code, seems to me that you'd just need tweaking, optimization, rewriting any assembler (which you'd need to do with any architecture change) and a recompile, not *completely* rewriting your code. Add in the .Net framework and you're further abstracted. The Win32 API set was meant to be portable from the start. Thus the versions for Alpha, MIPS, and PowerPC. And Windows exists on IA64 as does Linux and Monterrey. So seems to me like the majority of the hard work has already been done.


The problem isn't in getting your C code to EPIC instructions; it is in getting them to efficient EPIC instructions. While the P4 does a lot of on-the-fly optimizations to the execution stream, EPIC/VLIW as implemented on the IA-64 does not. All out-of-order-execution decisions are left to the compiler. Which makes the processor much simpler and allows for complexity in other areas of the processor (or more L1 cache), but means that compiler programmers have to do a lot of extra thinking.

Yes, without doing any real hard thinking on the matter, one can write a straightforward C-to-Assembly compiler for IA-64. And, I strongly suspect, that is just a bit below what is out there right now. But taking advantage of the explicit execution ordering from the compiler will take time. Current comilers might take C-level programmer "hints" as well, which shouldn't be required but would make getting a compiler out the door a much quicker process. If this is the case, then, yes, the C programmer has to think about the fact that they are not in IA-32 world any more.

As for Win32 being portable ... have you looked at MS's scant non-IA-32 documentation? The shell of Win32 is available on other platforms, but there are enough "gotchas" and missing pieces that for a Windows developer to rely on Win32 to provide all hardware abstraction is downright foolish. Granted, those pieces might be filled in and the "gotchas" largely eliminated, but I don't think that MS is putting that much effort into Win32 on different platforms. Which, yes, is where .Net comes in.

[alex_ant]

Originally posted by Catfish_Man
I'm not so sure that this chip is good for the XServe. There's a reason why x86 1Us are still using P3s. The PPC970 is NOT being designed for low power consumption (according to anything I've heard), so for now it's only going to be useful in the PowerMacs. A little later when they transition it to .09 micron... we might start seeing TiBooks or XServes with it (they're pretty similar, actually, and have pretty similar requirements).
With a Celeron-size die, a 1.8GHz clock, and the manufacturing processes IBM will be using, I will be surprised if this chip has power requirements much beyond the G4's.

[kenohki]

Originally posted by jettredmont


The problem isn't in getting your C code to EPIC instructions; it is in getting them to efficient EPIC instructions. While the P4 does a lot of on-the-fly optimizations to the execution stream, EPIC/VLIW as implemented on the IA-64 does not. All out-of-order-execution decisions are left to the compiler. Which makes the processor much simpler and allows for complexity in other areas of the processor (or more L1 cache), but means that compiler programmers have to do a lot of extra thinking.

Yes, without doing any real hard thinking on the matter, one can write a straightforward C-to-Assembly compiler for IA-64. And, I strongly suspect, that is just a bit below what is out there right now. But taking advantage of the explicit execution ordering from the compiler will take time. Current comilers might take C-level programmer "hints" as well, which shouldn't be required but would make getting a compiler out the door a much quicker process. If this is the case, then, yes, the C programmer has to think about the fact that they are not in IA-32 world any more.


And not having directly worked with IA64, I'm ignorant on how efficient the compiler is. I had assumed that it was pretty whiz-bang from all the whitepapers I've seen on it. I mean, it does speculative loading and predication so I figured it was probably bleeding edge and had been enhanced by both Intel and HP during all the delays of the first Itanium. One would think that the compiler would be a major focal point of the project.

As for Win32 being portable ... have you looked at MS's scant non-IA-32 documentation? The shell of Win32 is available on other platforms, but there are enough "gotchas" and missing pieces that for a Windows developer to rely on Win32 to provide all hardware abstraction is downright foolish. Granted, those pieces might be filled in and the "gotchas" largely eliminated, but I don't think that MS is putting that much effort into Win32 on different platforms. Which, yes, is where .Net comes in.

Yes, I know how bad it is. But they also don't support any architectures other than IA32 anymore so the documentation has stopped. Alpha was the last one to be dropped. However, one of the original design goals of NT was to provide a portable environment. Back in the early days of NT and Win32 (when other microarchitectures were supported), there was a big push to make things portable between processors. I'm sure this was because at that time, RISC was shiney and new and MS didn't want to be left out of the game had Alpha maintained it's lead or PowerPC lived up to the promise of the PPC 620.

In today's environment though, I would assume MS is making sure that code is portable from Win32 to Win64 or whatever they're calling that version of Advanced Server.

[nixd2001]

Originally posted by kenohki


And not having directly worked with IA64, I'm ignorant on how efficient the compiler is.

I likewise have no direct knowledge. However, getting the sort of optimisations that Intel require has long been very difficult. To this end, though, it's worth noting that Intel purchased a company called Kuck & Associates a couple of years ago. These dudes are some of the best at performing optimisations - mainly as source->source transformations for simple compilers to handle. Strip mining, now a common technique to avoid data caching beating, is something they bought to the market after hiring the dude (Wolf?) who did his PhD in this area, following on with Superoptimising Supercompilers as a book.

In short, I don't think Intel are there yet, but they've got a lot of the right bits to stand a chance of getting there.

(alex_ant - anything you can chip (:eek:) in here re K&A., etc?)

[alex_ant]

Originally posted by nixd2001
I likewise have no direct knowledge. However, getting the sort of optimisations that Intel require has long been very difficult. To this end, though, it's worth noting that Intel purchased a company called Kuck & Associates a couple of years ago. These dudes are some of the best at performing optimisations - mainly as source->source transformations for simple compilers to handle. Strip mining, now a common technique to avoid data caching beating, is something they bought to the market after hiring the dude (Wolf?) who did his PhD in this area, following on with Superoptimising Supercompilers as a book.

In short, I don't think Intel are there yet, but they've got a lot of the right bits to stand a chance of getting there.

(alex_ant - anything you can chip (:eek:) in here re K&A., etc?)
I think you would know much more about the Itanium than me... I used to call it the Sh-itanium, but with these latest McKinley SPEC scores, I feel humbled. Whatever Intel/HP is up to, their compiler must be a work of art.

(Isn't the Microprocessor Forum happening today? Where's all the juicy news already?!?! :D)

[nixd2001]

Originally posted by alex_ant
I think you would know much more about the Itanium than me...
I was thinking of Kuck & Associates. But don't worry - minor part of the thread.

[bacon]

Originally posted by arn


Please read the rest of this thread.

Quick question: Would you rather have a higher MHz number or Faster Performace?

Remember, the IBM PowerPC is a different architecture... you can't compare the new chip to current Motorola G4's. You certainly can't extrapolate Mhz.

The 1.3 GHz IBM Power4 benchs close to the 2.8GHz Pentium according to numbers cited in this thread.

arn

So, the 1.3 GHz Power4, which is really 2 processors each running at 1.3 GHz, running effectively at 2.6 GHz benches close to the Pentium 2.8GHz chip? One would hope.

[nixd2001]

Originally posted by bacon


So, the 1.3 GHz Power4, which is really 2 processors each running at 1.3 GHz, running effectively at 2.6 GHz benches close to the Pentium 2.8GHz chip? One would hope.

If you're thinking of the SPEC benchmarks, I thought they specifically only used one processor core, so no, not 1.3*2=2.6....

[alex_ant]

(edit: redundant)

[Telomar]

IBM POWER4 based processor. - PPC 970

Target process - 0.13um SOI 8 layer Cu interconnect
Target frequency 1.4 ~ 1.8 GHz
Target sample date 2Q03
Target ship date 2H03
Power 42W @ 1.8 GHz 1.3V (low power mode @ 1.1 V)

Est. SPEC INT 937 @ 1.8 GHz
Est. SPEC FP 1051 @ 1.8 GHz

Chip features: Elastic unidirectional point-to-point interconnects between CPU and "companion chip" (i.e. memory controller/northbridge) Elastic link may run "UP TO" 900 MHz, offering 6.4 GB/s of memory bandwidth.

POWER4 internals. Max of 8 inst fetch per cycle, 8 inst issue per cycle, and 5 (4 + branch) inst dispatch per cycle.

32 KB L1 Dcache, 64 KB L1 Icache.
512K L2.

32 64 bit GPR (general purpose register)
32 64 bit FPR (float)
32 128 bit VRF (vector)

Note that that vector unit uses separate registers and is altivec compatible.

Or you can just go here. (http://www.realworldtech.com/forums/index.cfm?action=forum&roomID=11) There is more infor out there but I am too lazy to grab it. There is a good thread over at Ars Technica (http://arstechnica.infopop.net/OpenTopic/page?q=Y&a=tpc&s=50009562&f=8300945231&m=3470943335&p=9).

*waits for the complaints to start*

[alex_ant]

Not bad at all. 90%+ the number-crunching ability of the Power4 (not including AltiVec!) at 1/3 the power draw. I'm sure this is in the realm of what people were expecting. It will fit in an Xserve in pairs. Cut the energy requirements of the 1.4GHz version by 1/3 and it will easily work in a TiBook.

Maybe it's just that, compared to the G4, anything seems fantastic. :) At the very least, it's a fresh start of sorts. I'm not gonna complain.

[nixd2001]

Originally posted by Telomar
IBM POWER4 based processor. - PPC 970
.....
*waits for the complaints to start*

What speed gain is likely to be had from moving from .13 to .09 (Fishkill being built eventually to do 0.09 ISTR)?

Either way, two of those in a quality hardware design with a quality operating system and quality applications will probably keep me content. Especially if each processor can maintain it's own memory bandwidth (now that needs a lot more details to know whether its possible - coherency etc aside).

It may be that a PC users can get more useful operations done per second (to be seen), but I'll probably get my task done first.

[topicolo]

Wow, I just saw the scores for the 2Ghz AMD Opteron.

Running on a mobo with dual channel PC 2700 DDR memory (that's 333Mhz DDR boys and girls), the AMD obtained a SPEC CPU 2000 INT score of 1202 and a SPEC CPU 2000 FP score of 1170. Apparently, when running on a 64bit clean compiler, the scores are about 20% higher.

Crazy. They just HAD to beat us by one step. ;)

[solvs]

Originally posted by jadam
and who reported this story to you MR ARN!$@?????

Wow, you are an idiot.

This is the second time today you've said something stupid that I've seen (the first being that the smarter a person is, the more money they make, and how teachers are stupid for making only $30,000 a year. My Mother's teacher. And just because you go to a good school, that doesn't make you smart.). This information is everywhere. Has been for days.

And you do know ARN is one of the administrators here, don't you? What was that about you being so smart?

Dumbass. :mad:

[ddtlm]

Est. SPEC INT 937 @ 1.8 GHz
Est. SPEC FP 1051 @ 1.8 GHz

These scores mean the PPC970 is expected to get about 520 SPEC int per ghz compared to 620 per ghz for a full Power4. The PPC-970 is apparently expected to get 580 SPEC float per ghz, compared to 920 for a full Power4.

If these figures are true it is unlikely that the PPC-970 will have higher SPEC scores than the top P4s or Athlons when it is first available, since the integer score is already behind and the float score only leads the current top P4 by perhaps 12%. Well, it's a lot better than a G4, anyway.

[sturm375]

Originally posted by rice_web
The one thing that I am hopeful for is the bus speed. IBM's G5 has been rumored to support 6.4GBs of throughput, while their G3 supports 3.2GBs. Using some math....

If 3.2GBs = 200MHz x 2 system bus (200MHz with DDR) = 400MHz

Then 6.4GBs = 200 x 2 x 2 (200Mhz with DDR and double-pumped) = 800MHz

AMD Hammer: 19.2 GB/s!


Q:How many HyperTransport™ links does the “Hammer” architecture support?
A:The “Hammer” architecture are designed to support up to three HyperTransport links. The combined peak bandwidth of the HyperTransport links is 19.2GB/sec.

from: Here (http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_4699_875%5E1024,00.html#1540)

[ddtlm]

sturm375:

Yes, but that is only for multiprocessor Hammers (the Opterons) which cost more than normal ones, will have big L2 caches, and are generally designed for workstations and servers rather than consumer desktops.

Not to mention that all those HT links go different places, whereas the IBM bus goes one place. When transferring any single thing, there is no advantage to the Hammer as far as raw bandwidth scores.

Also note that the Hammer's HT links do not go to memory, they go to other processors and to things like AGP bridges. The Operon has dual channel DDR, presumably DDR 333, so would have something like 5GB/s memory bandwith, less than the PPC-970 when getting things from memory. (This is somewhat more complicated in multiprocessor Operton systems where each chip has its own memory and a single chip can in theory get more than 6.4GB/sec of bandwidth.)