IBM PowerPC 970 vs Intel Pentium 4

[MacCoaster]

Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by MrMacman
Any-how, inless you are operating a server or extreme graphics/EXTREME gaming you don't see many people using Dual Xeon's look at that Dell commerical (*ech*, sorry seen it too many times) they advertise it as a server, Which it was made for doing. Xeon's are nice, but only if you are willing to pay the price.

More assumptions! So I am more likely to see G4 towers around than iMac for a consumer who doesn't need all that power? Can you back up your claims with data?

Not only Intel is advertising the Xeon as a server chip, but as a workstation chip. It could be very well said that Xeons do compare with the G4 towers.

And same is true for a G4 tower. They are nice, but only if one is willing to pay the [disgustly overpriced] prices.

AGAIN, it doesn't matter if "less" percentage of people use Dual P4s than Dual G4s, what matters is: the option is available. And can you prove that there are less percentage of Dual P4 users than Dual G4 users? Out of what? Etc.

 

[will]

I think you'll find that study was comissioned by Microsoft and was misleading in many ways. For example it compared costs over five years so licensing costs were lower, over three years even the MS comissioned study showed Linux was cheaper. Read The Register article for more http://www.theregister.co.uk/content/4/28408.html I can't think of any area I would choose to use a Windows server, I would choose Sun or a BSD based system myself. The XServe is actually a very nice bit of hardware, but it will take a while for it to build up the support base and stability that exists for longs standing server systems like Sun.

 

[frogstomp]

Best of the Best!

Wow think of what that network system could do with some real power under their hoods ie 970's instead of the bottle necked shared bus rate of 1.3gb/s for both cpu's!

The P4's suffer from the inability to support seperate busses for each CPU also, but at least run at many times faster buss rates.

The Athlons can utilize a seperate bus for each cpu and therefore make for great data pushers...hmmmm Linux Rendering Networks....

So right now the fact that Xserves have DDR memory is moot. It is choked off at the system.
http://www.barefeats.com/xserve2.html

Ahh but bring on 64bit technology please! Not only is the 970 altivec compliant, but will also allow 32bit programs to run at the same speed advantage as 64 bit programming! Ohh and how about Bus Speed 900mhz sounds really nice ~ 5x the data through-put that apples ppc can handle right now!
http://www-3.ibm.com/chips/products/powerpc/newsletter/dec2002/newproductfocus2.html

Will the 970 @ 1.8ghz trounce a P4 at 3ghz? I haven't done the math yet...

cheers!

 

[will]

Re: Best of the Best!

Originally posted by frogstomp
Will the 970 @ 1.8ghz trounce a P4 at 3ghz? I haven't done the math yet...

Hard to tell as we have 3GHz P4s now, but the 970 hasn't shipped. So it's a bit of a moot point. Also performance very much depends on the benchmark you choose.

However I think it's important that Apple have processors that are at least in the same league as the P4/Hammer on SPEC INT and FP benchmarks. CPU performance does matter, particularly in areas important to Apple like graphics, 3D, and video editing.

I look forward to seeing a shipping Apple 970 system in Q3 2003.

 

[bluecell]

Re: Best of the Best!

Originally posted by frogstomp
Will the 970 @ 1.8ghz trounce a P4 at 3ghz? I haven't done the math yet...

No, the 970 @ 1.8GHz is a little behind the P4 @ 3.06GHz. That's why some see this as a potential problem for Apple. From the SPEC results that I've seen for the Athlon64, the PPC 970 seems to be the winner.

 

[MacCoaster]

Re: Best of the Best!

Originally posted by frogstomp
The Athlons can utilize a seperate bus for each cpu and therefore make for great data pushers...hmmmm Linux Rendering Networks....

Why do you think George Lucas' company used a cluster of Athlon MP-powered PCs running Linux to render their latest films.

 

[papolo]

...

hi guys
I just enjoyed reading the first tree pages, I learned a lot, but I have to leave now...
and I would like to continue my reading tomorrow expecting that somebody could answer this doubt I have:

In a posted message someone made the compairison between the 970 processor against the actual G4 processor, he says that the actual G4 use SDRAM memories structure against the DDR memory use of the 970... Could somebody tell me what does it mean? and why the Albooks17 uses DDR SDRAM memories??

thanks a lot

 

[maradong]

well they are using ddr right now, but the fsb between cpu and ram is still 133 at the max it s not doubled , resp quaded as with the pentiums, what gives a great bench of performance.
with the ppc 970 this should change.
i hope that helps you.

 

[daveg5]

Larger cache

A switch to a Larger 1MB or 2MB L2 cache will all by itself ad 20-30% to the performace if the 603-604-G3-G4 are any indication and could probably done sooner then higher clock speeds.

 

[Cubeboy]

Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by MrMacman
I like the part where you posted that Intel has them in Lab's. I'm sure apple has labs with a few 970's, but do I say that apple is ready to release them tommorow, NO.
Intel cannot get the P4 in its current condition to 4 GHZ. Intel has stated tha in 2003 the highest possible they will release is 3.6 GHZ. And look what it took for Tom's Hardware to get it to 4 GHZ, not monster fans, not water cooling. LIQUID NITROGEN. THEY BASICALLY FROZE THE CHIP.
What was it - 52 degrees Celcius?

Any-how, inless you are operating a server or extreme graphics/EXTREME gaming you don't see many people using Dual Xeon's look at that Dell commerical (*ech*, sorry seen it too many times) they advertise it as a server, Which it was made for doing. Xeon's are nice, but only if you are willing to pay the price.

Edit: UBB code not working...

Not quite how it works, we're not talking about the current Pentium 4 are we now? No, we're talking about Prescott, which in terms of the micron process itself is more than an entire generation ahead of the Pentium 4. Look at the difference between the .18 Pentium 4 and .13 Pentium 4, one was only able to scale to around 2 Ghz, the other can scale to well over 3 Ghz (3.6 is the estimated limit) before reaching the same temperature/limitations. With .09 micron process as well as strained silicon technology which allows a processor to scale to even higher clockspeeds, the Prescott will easily scale to 5 Ghz before reaching anywhere near the temperature/limitations of a current 3 Ghz Pentium 4. The reason they don't just come out with 4-5 Ghz parts from the outset is due to the fact that far more profits can be made by keeping higher yield parts (like a 3-4.5 Ghz Prescott) at premium prices while still offering superior performance to competitors (as was clearly seeing during the speed wars between AMD and Intel with Intel only releasing higher clocked Pentium 4s after AMD released a higher clocked/better performing Athlon that could outperform the Pentium 4s on the market at that time.)

Also, exactly what price is their to pay for a Xeon workstation, I've seeing dual 2.4/2.66 Ghz models go for equal or less than most Powermacs.

 

[jwdawso]

Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by Cubeboy
... The reason they don't just come out with 4-5 Ghz parts from the outset is due to the fact that far more profits can be made by keeping higher yield parts (like a 3-4.5 Ghz Prescott) at premium prices while still offering superior performance to competitors (as was clearly seeing during the speed wars between AMD and Intel with Intel only releasing higher clocked Pentium 4s after AMD released a higher clocked/better performing Athlon that could outperform the Pentium 4s on the market at that time.)

Also, exactly what price is their to pay for a Xeon workstation, I've seeing dual 2.4/2.66 Ghz models go for equal or less than most Powermacs.

Business doesn't work that way. If they have the parts, they sell them. There is no holding back. Of course they price them appropriately - high yields are cheaper than the rarer ones.

Give us a link to the Xeon pricing.

 

[Cubeboy]

Re: Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by jwdawso
Business doesn't work that way. If they have the parts, they sell them. There is no holding back. Of course they price them appropriately - high yields are cheaper than the rarer ones.

Give us a link to the Xeon pricing.

On the contrary, the longer you keep a part relevent on the market (and thus charge a premium price for it), the greater the profit margins. By constantly upgrading your chip line or starting your chip line with only the higher clocked models when the competition doesn't demand it, your losing profits for every chips you sell since your essentially asking for the same price for a chip with lower yield (since their is a limit to how high your can ask).

If you look at the speed wars between Intel and AMD during 2002, Intel only released a higher clocked Pentium 4 when a Athlon threatened to overtake the previous Pentium 4 model, even though the Pentium 4 at that time wasn't coming anywhere near it's architectural limit and could easily be clocked to higher speeds (i.e it's really not hard to come out with 2.4/2.6/2.8 Ghz parts when your processor can readily scale above 3 Ghz). The thing to remember here is, until a processor nears it's architectural limit, it's relatively easy to increase it's clock rate, only in cases where a processor reaches or exceeds it's limit does it take significant time or resources to clock it higher.

Also regarding Xeon workstations, you can customize a very well equipped Dual Xeon 2.4 / 2.66 Ghz workstations from Dell for around $2000.

 

[jwdawso]

Re: Re: Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by Cubeboy
On the contrary, the longer you keep a part relevent on the market (and thus charge a premium price for it), the greater the profit margins. By constantly upgrading your chip line or starting your chip line with only the higher clocked models when the competition doesn't demand, your actually losing profits for every chips you sell since your essentially asking for the same price for a chip with lower yield (since their is a limit to how high your can ask).

If you look at the speed wars between Intel and AMD during 2002, Intel only released a higher clocked Pentium 4 when a Athlon threatened to overtake the previous Pentium 4 model, even though the Pentium 4 at that time wasn't coming anywhere near it's architectural limit and could easily be clocked to higher speeds. It's really not hard to come out with 2.4 or 2.6 Ghz parts when your processor only begins to have issues at over 3 Ghz. The thing to remember here is, until a processor nears it's architectural limit, it's relatively easy to increase it's clock rate, only in cases where a processor reaches or exceeds it's limit does it take significant time or resources to clock it higher.

Also regarding Xeon workstations, you can customize a very well equipped Dual Xeon 2.4 / 2.66 Ghz workstations from Dell for around $2000.

Business doesn't work that way - that's an outsiders view of big, bad corporations. It's simple - you've got the parts, you sell them. Every company has pricing formulas that they follow that relate to cost and consider uniqueness that allows a premium. Higher clock speeds roll out from constant engineering analysis. When a new chip comes out, it's projected high end comes from improving processes, improved components, and small board changes.

I don't disagree about the price - just provide a link to a configured system. I refuse to go to a PC site.

 

[Cubeboy]

Re: Re: Re: Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by jwdawso
Business doesn't work that way - that's an outsiders view of big, bad corporations. It's simple - you've got the parts, you sell them. Every company has pricing formulas that they follow that relate to cost and consider uniqueness that allows a premium. Higher clock speeds roll out from constant engineering analysis. When a new chip comes out, it's projected high end comes from improving processes, improved components, and small board changes.

I don't disagree about the price - just provide to a configured system. I refuse to go to a PC site.

It's not big or bad, it's just how chip makers do things, why do you think Intel clocked their first northwood Pentium 4s at only 2.0 Ghz when that exact same core can be easily pushed to over 3 Ghz? It's simply how things are done and have always been done.

Also, here's a link to a Xeon workstation I just configured. Just to name a few features:

Dual Xeon 2.4 Ghz
512 MB Dual Channel DDR226
36 GB Ultra 320 SCSI drive
QuadroFX 500
Total price: 2235 dollars

http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=04&kc=6W463&l=en&oc=ws450min

 

[DakotaGuy]

Originally posted by pgwalsh
I don't know anyone that has a computer with the 3 Ghz Intel inside.

I have a friend with a 3.06Ghz P4 Gateway. The thing has been locking up a lot. When it runs, it's fast. He has had it back to Gateway 3 times now because of overheating problems. The problem is more Gateways then Intel though. He bought one of the first ones they put out and Gateway did not beef up to cooling enough for the chip. It is a piece of junk for all the money he paid for that thing. I don't care what a lot of people say about speed. If they had that Gateway or a new Dual 1.42 sitting there and said, take your pick it's your birthday, the decision for me would be a no brainer.

 

[jwdawso]

Re: Re: Re: Re: Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by Cubeboy
It's not big or bad, it's just how chip makers do things, why do you think Intel clocked their first northwood Pentium 4s at only 2.0 Ghz when that exact same core can be easily pushed to over 3 Ghz? It's simply how things are done and have always been done.

Also, here's a link to a Xeon workstation I just configured. Just to name a few features:

Dual Xeon 2.4 Ghz
512 MB Dual Channel DDR226
36 GB Ultra 320 SCSI drive
QuadroFX 500
Total price: 2235 dollars

http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=04&kc=6W463&l=en&oc=ws450min

If you mean by "easily pushed" that it takes not a lot more engineering and not much of a change & improvement to components, I guess we agree. But I would suggest that "easily pushed" does cost money, and that the price reflects it. In other words, Intel can produce a 2Ghz chip for $x, but by "easily push[ing]", can produce a 3Ghz chip for $x+"a small amount". I would also suggest that the time to market for the 3Ghz chip - even if Intel wanted to produce it ASAP - is some delta after the 2Ghz could be produced. But I do not believe that Intel - for the same price and at the same exact time - could produce the 3Ghz as the 2Ghz. The reason? Because Intel would make more money and garner greater market share by releasing the "giant leap forward" chip. For Intel to only release the 2Ghz when at the same time and for the same money it could release 3Ghz, then AMD and Intel must have an agreement to release chips incrementally in unison. If they don't, then why would they ever increase the speed?

When you say the "exact same core", do you mean there are no component changes at all? Even if that is true, there are process changes that enable producing faster & faster chips at higher yields. Small component and process changes are the inputs to faster chips and higher yields. This takes time.

Thanks for the info/link on Dual Xeon.

 

[trebblekicked]

i saw a couple of references the "the best apple can offer" being a dual 1.25. unless i'm missing something i believe this is the top of the line under $3000.00

anyway, i second this emotion:

I think some of us should reconsider our priorities. Which do we want most, an awesome new processor in our Macs or an 'Intel killer'? Personally if I had to choose between the two the former would win every time.

how the 970 compares to the P4 (prescott or not) in non-specific benchmarks is mainly irrelevant for the market the 970 is aimed at. apple pro users run programs like FCP, AE, Photoshop etc, all of which are multi-processor aware and i believe altivec enabled. this puts real world pro useage on a comperable stage as the pc platform. that's all we need. not to bury the P4 (although we'd love to see that). i bought my desktop relatively recently, and if apple can provide me significanlty decreased render times in after effects, photoshop and FCP, i will upgrade.

 

[Cubeboy]

Re: Re: Re: Re: Re: Re: Re: There are less % of Dual P4 Users that Dual G4 users.

Originally posted by jwdawso
If you mean by "easily pushed" that it takes not a lot more engineering and not much of a change & improvement to components, I guess we agree. But I would suggest that "easily pushed" does cost money, and that the price reflects it. In other words, Intel can produce a 2Ghz chip for $x, but by "easily push[ing]", can produce a 3Ghz chip for $x+"a small amount". I would also suggest that the time to market for the 3Ghz chip - even if Intel wanted to produce it ASAP - is some delta after the 2Ghz could be produced. But I do not believe that Intel - for the same price and at the same exact time - could produce the 3Ghz as the 2Ghz. The reason? Because Intel would make more money and garner greater market share by releasing the "giant leap forward" chip. For Intel to only release the 2Ghz when at the same time and for the same money it could release 3Ghz, then AMD and Intel must have an agreement to release chips incrementally in unison. If they don't, then why would they ever increase the speed?

When you say the "exact same core", do you mean there are no component changes at all? Even if that is true, there are process changes that enable producing faster & faster chips at higher yields. Small component and process changes are the inputs to faster chips and higher yields. This takes time.

Thanks for the info/link on Dual Xeon.

Yep, same Northwood Pentium4b core, all you have to do is increase the multiplier or fsb or both, how do you think hardware enthusiasts overclocked their 2.4 Ghz Pentium 4s to 3+ Ghz models (this is a particularly popular setup in the overclocking community)?

As I've said before, the industry revolves around providing the fastest processor while making the most profits. Back in the day it was introduced, a Pentium 4 2.0 Ghz costed around $500 dollars, right now, that same Pentium 4 costs around $150 dollars. A 2.80 Ghz Pentium 4 was initially priced at $508 dollars, now it costs around $200. It's counterproductive and frankly incredibly stupid to introduce a part at it's highest frequency if lower clocked higher yield parts well fetch you the same sales and go for the same price. If a 2.2 Ghz Pentium 4 is the highest performing chip on the market (it was), and sells for $650 dollars (it did), why on earth would you lower it's value to well over half of what it was before with a 3.06 Ghz Pentium 4 that will probably cost much more to make (it does), fetch around the same number of sales (it did), and go for the same price (intro'd at 650 dollars as well). Not only are you significantly cutting profits by a rather large amount, your also preventing every single Pentium 4 before the 3.06 Ghz model from ever fetching a decent profit. Thats why Intel only released a 2.8 Ghz model right after (and I mean right after) AMD released it's 2600+ and 3.06 Ghz model right after AMD released it's 2800+.

 

[scem0]

They are both better for different things and to argue about which one is better is useless and won't get anyone anywhere.

 

[Vlad]

The 970 is a 64 bit cpu. This will make a difference for new software and the OS update, Panther. Consumer PCs will have to rely on a new windows version and an update to their cpus such as the new Athon Operton. IBM will continue to push the 970 for their server lineup and also because it will be the basis for the Playstation 3 (which will be using 4 x modified 970s). The key going forward is definately performance like it always has been; however, Apple has also greatly benefited from power usage which will only get better and now they will also get better pricing. Apple has a road forward without another completley new processor architecture switch while remaining competitive--this is good . . . all good.

 

[fred_garvin]

For 95% of the Mac users out there, performance worries disappeared a couple hundered mHz ago. A 1.8 gHz PPC 970 won't be discernably different from an 800 mHz G4 (or even G3 for that matter). Most of the apps that people run are relatively simple and the biggest change may be that the Aqua overhead is no longer noticed, unless Apple makes it even more burdensom to force us into hardware upgrades (does that sound jaded? )

Apple will always find a use for the speed, and consumers will want it. Sometime around MWSF 2001, Steve mentioned "look what we can do now that we have a gigaflop to play with!" I think he was showing off Aqua, the genie effect, etc. Some of this is fluffy eye candy, other parts are very useful. I'm sure Apple has a slew of advanced UI features they'd like to release, but are just too heavy for current hardware. People have speculated about piles, but I think piles will be more useful when it can be rendered in 3d.

Try to think into the future, and what OS features we'll take for granted in 10 years. At some point we will have good speech recognition built in, which takes cpu, extremely useful artificial intelligence, more cpu, and 3d rendered ui.