Low Power G5 Coming Soon?

[maya]

I thought the iMac G5 rev. A was quiet enough as it is. How could Apple make it produce even less noise while still upgrading it? Also, as far as I'm aware, there wasn't THAT big of a difference in noise between the two revisions of PowerMac G5.

The "noise issue" is shared by some iMac G5 users maybe not all however some. We all have different levels of sensitive when it come to hearing. And I fall in that group.

Since I have an iMac G5 and have used and owned other Mac's in the past, I feel that its false that Apple markets the iMac G5 as "more silent than a whisper."

Even on the "Automatic" setting the system can rev up and can get quite loud.

here are other smaller issues here and there also pertaining to noise issue(s). Changes to the rev B iMac G5 would be great if the G5 ran cooler (as mentioned in this article in regards to low wattage being), with the first will bring fewer "fans" as with the rev A PMG5 had more fans than the rev B PMG5 thus it is more quite to the end-user, better placed speakers (where it doesn't sound as if the noise was resonating from a hollow box), and an "audio headphone jack" no the keyboard would be GREAT.

Other than that the iMac G5 is quite well indeed.

The iMac G5 is not "silent" even under what Apple states as "normal" working load and noise. Silence is Bliss.

 

[ASP272]

A G5 powerbook would be like an UGM! An efficient Halo 2 machine perhaps!

 

[AidenShaw]

I'm not so sure. SPARC is a case in point: UltraSPARC is a 64 bit processor, but you still see a great deal of software compiled in 32 bit mode for Solaris. This isn't so much for compatibility -- a lot of this stuff is certified for later versions of Solaris that generally run on UltraSPARC-based systems -- as it is for performance, AIUI. If you don't need the 64 bit stuff -- ie, your memory needs are modest, and your integer arithmetic fits fine in 32 bits -- the odds are good that any appropriate Solaris stuff will be 32 bit code, not 64. This is basically where I was coming from. If the CPU works fine with both 32 and 64 bit code, I don't see a general need to recompile 32 bit code to 64 bit code just for the sake of it. x86-64 is a case where it wouldn't be for the sake of it -- you'd be recompiling to take advantage of the extra registers.

OK, I see your point. There are a couple of differences with the SPARC case, though.

• The 32-bit Solaris code is compatible with older 32-bit SPARC systems - so you could get by with one "thin binary" for apps that didn't need 64-bits. The "32-bit x64 with extra registers" would not run on x86 machines - so you'd have to build two.
• The first 64-bit SPARC machines were SLOW - memory was pre-SDRAM 60ns (that's 16 MHz). Reducing memory bandwidth was important - today's PCs have far more bandwidth available.
• Sun ships both 32-bit and 64-bit Solaris in the same kits - so many people are still running 32-bit Solaris on 64-bit capable hardware. (Just like most people run 32-bit O/S on x64 and only 32-bit O/S is released for OS X.)

On the other hand, your "32-bit x64 with extra registers" would require a recompilation, and would not be compatible with either 32-bit hardware or 64-bit operating systems (unless the OS makers included a third set of libraries and APIs).

In addition, the extra memory traffic due to 64-bit pointers is really noise for most applications. You'll need a very carefully run benchmark to separate the 64-bit burden from the noise.

Of course, there are some applications where it is noticeable - for example if your data is a huge doubly-linked list or doubly-linked tree with very small data payload per structure.

For those applications, most systems have compiler/linker options to allow 32-bit pointer usage on a 64-bit system. (If you can guarantee that the data structure resides in the low 4 GiB of virtual address space, you can safely truncate the pointers when you store in memory, and extend when you fetch.) In other words, you can do what you want today (at least with Windows - haven't looked at gcc).

So, your proposal has some merit - but IMO the disadvantages outweigh the advantages.


ps: There's a good PDF from AMD describing x64 at http://www.amd.com/us-en/assets/content_type/ white_papers_and_tech_docs/x86-64_wp.pdf. It goes into detail about the 16/32/64 bit modes and register usage. Your new mode is not provided in the hardware - the extra registers require the full 64-bit mode.

 

[cb911]

aawww.... i'm confused now.

i was seriously hoping for a dual-core G4. that would have been an awesome chip... but i guess everything is still up in the air? at least it's good to have options. but if it's a 3GHz G5, i won't be complaining...

but either way, it seems that it' will probably be about March, more likely April at least before we see a PowerBook revision.

also someone mentioned about the 'lack' of L3 cache, would that affect performance? i don't know if someone else has answered this... but no. the current PowerBooks (& iBooks i guess?) don't have any L3 cache, simply because there is no need for it, and it doesn't negatively impact performance.

 

[sjl]

On the other hand, your "32-bit x64 with extra registers" would require a recompilation, and would not be compatible with either 32-bit hardware or 64-bit operating systems (unless the OS makers included a third set of libraries and APIs).

In addition, the extra memory traffic due to 64-bit pointers is really noise for most applications. You'll need a very carefully run benchmark to separate the 64-bit burden from the noise.

[snippage by sjl]

So, your proposal has some merit - but IMO the disadvantages outweigh the advantages.

Ah, ok, I see where you're coming from. (See, kids? People with differing opinions can come to an agreement. ) Your comments do have merit. Certainly, I'd now be looking at benchmarking to compare the performance of an appropriate load before making a firm and fast decision on such a matter. And besides -- playing around with benchmarking of this sort for the sake of playing around with the benchmarking is fun!

 

[MacinDoc]

aawww.... i'm confused now.

i was seriously hoping for a dual-core G4. that would have been an awesome chip... but i guess everything is still up in the air? at least it's good to have options. but if it's a 3GHz G5, i won't be complaining...

but either way, it seems that it' will probably be about March, more likely April at least before we see a PowerBook revision.

also someone mentioned about the 'lack' of L3 cache, would that affect performance? i don't know if someone else has answered this... but no. the current PowerBooks (& iBooks i guess?) don't have any L3 cache, simply because there is no need for it, and it doesn't negatively impact performance.

I hope you're not expecting a 3 GHz PowerBook. The source suggests that the PowerBook will be clocked at a much more modest 1-6-1.8 GHz. And, to reduce heat, it will likely have a 3:1 bus ratio, similar to the iMac. So, it probably would benefit from L3 cache, but I don't expect any L3 in the next rev.

Having said that, if Apple did come out with a 3 GHz PowerBook this spring, it would shock the computer world, and it would likely be the best-selling Apple product of all time. Well, we can always dream...

 

[Rincewind42]

I hope you're not expecting a 3 GHz PowerBook. The source suggests that the PowerBook will be clocked at a much more modest 1-6-1.8 GHz. And, to reduce heat, it will likely have a 3:1 bus ratio, similar to the iMac. So, it probably would benefit from L3 cache, but I don't expect any L3 in the next rev.

Truthfully, I'm expecting more like a 4:1 bus ratio. Not the fastest, but faster than any Powerbook to date at least. I do hope they can get up to 2Ghz though, as I don't think too many people will be impressed by a 1.33/1.5 to 1.6/1.8 but the same people would probably buy a 1.8/2.0 split.

Having said that, if Apple did come out with a 3 GHz PowerBook this spring, it would shock the computer world, and it would likely be the best-selling Apple product of all time. Well, we can always dream...

Heh, Dual 970MP 3Ghz PowerMac and 970GX (or better yet 970MP!) 3Ghz PowerBook - we certainly can dream .

 

[MacinDoc]

Heh, Dual 970MP 3Ghz PowerMac and 970GX (or better yet 970MP!) 3Ghz PowerBook - we certainly can dream .

Now you're talking!

 

[maya]

Apple/SJ is going to introduce the new PowerBook G6 that is faster than any PPC or x86 to date, they have a time machine and brought the lowest spec laptop they can find and will market it to all they users and call it a Blade Server with a screen and will keep your coffee, tea. legs and palms warm in the winter.


or you can buy the



new "SpacePod" that will hold 10,000 passengers and reach Mars in 15 battery hours.

 

[thatwendigo]

Truthfully, I'm expecting more like a 4:1 bus ratio. Not the fastest, but faster than any Powerbook to date at least. I do hope they can get up to 2Ghz though, as I don't think too many people will be impressed by a 1.33/1.5 to 1.6/1.8 but the same people would probably buy a 1.8/2.0 split.

Or, you know, it could just not have a front side bus at all by using on-die memory control and eliminating the northbridge... Gee, two cores, DDR2 667mhz, two 128-bit dual precision AltiVec units, on-die network control, on-die memory control, and roughly 15-15 watts for the whole package? Nah, let's just try to shoehorn a processor that was cooked down from big iron into a laptop instead, since that makes so much more sense than using a newer and more power efficient core design!

 

[~Shard~]

Or, you know, it could just not have a front side bus at all by using on-die memory control and eliminating the northbridge... Gee, two cores, DDR2 667mhz, two 128-bit dual precision AltiVec units, on-die network control, on-die memory control, and roughly 15-15 watts for the whole package? Nah, let's just try to shoehorn a processor that was cooked down from big iron into a laptop instead, since that makes so much more sense than using a newer and more power efficient core design!

Thanks thatwendigo, you just broke my sarcasm detector - now I have to go and fix it again..

And good thinking by the way - I like it...

 

[Mac-Xpert]

Heh, Dual 970MP 3Ghz PowerMac and 970GX (or better yet 970MP!) 3Ghz PowerBook - we certainly can dream .

Yeah, that would be great, for all sort of things..

 

[AidenShaw]

Sounds like a great chip !

Or, you know, it could just not have a front side bus at all by using on-die memory control and eliminating the northbridge... Gee, two cores, DDR2 667mhz, two 128-bit dual precision AltiVec units, on-die network control, on-die memory control, and roughly 15-15 watts for the whole package? Nah, let's just try to shoehorn a processor that was cooked down from big iron into a laptop instead, since that makes so much more sense than using a newer and more power efficient core design!

Wow - that would be a great design for a laptop. Power efficient *and* dual core - it'd blow any PPC970-based system right out of the water.

It wouldn't be 64-bit, but until you can put 8 GiB of RAM in your laptop you won't need 64-bit!

 

[Cindy O.]

Will top drawer G4 PBs cut it with Tiger?

....The source suggests that the PowerBook will be clocked at a much more modest 1-6-1.8 GHz....

If Apple does attempt to save some face quickly in January with a 1.6 / 1.8 PB to hold all waiters over till the big PB update next summer, I'd be interested to know on a technical note how the last of the great G4 powerbooks will fair running Tiger? Will Tiger put a time stamp on how long the strongest G4 PBs (even current 1.33 & 1.5's) will be able to keep up with the next version of MAC OSX?

 

[dacloo]

I guess Tiger will just run as well as 10.3, perhaps even better, but certainly not worse.

Try 10.0 running on a G4 PB. Now THAT is unoptimised.

 

[oingoboingo]

Or, you know, it could just not have a front side bus at all by using on-die memory control and eliminating the northbridge... Gee, two cores, DDR2 667mhz, two 128-bit dual precision AltiVec units, on-die network control, on-die memory control, and roughly 15-15 watts for the whole package? Nah, let's just try to shoehorn a processor that was cooked down from big iron into a laptop instead, since that makes so much more sense than using a newer and more power efficient core design!

That is a really crap idea. Your first suggestion was way better, you grubby G5 fanboi you.

 

[thatwendigo]

Wow - that would be a great design for a laptop. Power efficient *and* dual core - it'd blow any PPC970-based system right out of the water.

It wouldn't be 64-bit, but until you can put 8 GiB of RAM in your laptop you won't need 64-bit!

Features of the MPC8461D:

• Dual-cores that begin at 1.8ghz and ramp up from there, using the modified e600 design that will maintain compatibility with POWER and PowerPC instruction sets. The pin-out is wholy incompatible with previous generations, and the processor itself is manufactured on CuSSOI2 at 90nm with a heat output of 15-25 watts under full load.
• The on-die memory controller eliminates latency by removing the northbridge, giving direct access to RAM up to DDR533 or DDR2-667. The advantage to using DDR2 in a laptop is its far lower voltage, and therefore a greatly reduced heat consumption. Also, with time, the new format for DDR2 will yield denser memory chips to fill the PCB on a ram stick, helping to pave the way for the e600's 64-bit sucessor, the e700.
• The processors are fully SMP and SMT enabled, with each one hosting an independent, low-latency 1MB of L2 cache at core frequency. There are inbuilt sharing routines to allow the processors to fetch from each other's cache in order to keep the pipelines full, rather than going to memory or the drives. The single chip presents itself as at least 2 separate processors, with the interesting addition that each one can be booted into a fully independent system, running a separate OS at time of power-on. Also, the L2 cache can be enabled in ECC mode.
• Included in the System-On-Chip design are four hardware MACs at gigabit speed, providing for low-latency control over networking tasks. In addition, hardware support for major encryption standards will speed security tasks involving hashing and verification of MD5 checksums.
• The superior (to IBM, at least) implementation of AltiVec in Freescale's design is further enhanced with dual-precision (important for many scientific tasks) math units that can handle 128-bit vector operations, one attached to each core.
• The support fabric for the processor is an on-die controller for both FreeScale's RapidIO and Intel's PCI-Express channels, supporting throughput in excess of the PCI-X and AGP currently in use in Apple's systems. It's not necessary that both be included in a motherboard layout, as they serve a very similar function.

That's just off the top of my head, too. So, if you ignore the hype and FUD about the bitness of a processor and just look at features, the dual-core chip from FreeScale offers far more bang than a single processor 970 would. This is assuming that IBM hasn't managed some kind of incredible feat in their labs and held out on us in the meantime, which is a distant possibility. Without major architectural improvements that go far beyond just turning the clock up a bit, the current G5 will most likely be left crying in the dust, especially in a mobile platform where power draw is important.

Really, you'd think mac users would be wise to the megahertz myth by now.

 

[cb911]

each version of OS X has brought speed improvements no matter what the hardware it's been running on... so there's definitely no need to worry about Tiger having more system requirements. but that's going slightly OT...

 

[Mac-Xpert]

Features of the MPC8461D:
..........

I think you mean the MPC8641D based on the e600 core
This could certainly be a attractive alternative to the current G5 offerings. But I do think they will not be available till second quarter next year, although I couldn't find anything about that on the Freescale site, so I could be wrong.

EDIT: Update:
Reading more about it online, I'm under the impression that this chip is or available now or soon, since no one seems to indicate any release dates.

However the chip seems to be running on 1.5 GHz and expected to exceed that. The 15-25 watts power consumption seems to be for the 1.5 GHz speed.

 

[bobelmore]

Waiting for the right time...

Like many of you, I am waiting for that right time to buy a new Powerbook, or more correctly, avoid the wrong time (like a month before new models are announced). My G4-667 will be three years old in March. I was hoping to wait long enough for a G5, but now I am not sure if that makes sense. All we can do now is wait and see what is announced at Macworld in January.

 

[~Shard~]

Like many of you, I am waiting for that right time to buy a new Powerbook, or more correctly, avoid the wrong time (like a month before new models are announced). My G4-667 will be three years old in March. I was hoping to wait long enough for a G5, but now I am not sure if that makes sense. All we can do now is wait and see what is announced at Macworld in January.

As I've said a million times before, if you always wait for the next best thing, you'll always be waiting.

That being said, I know it's frustrating when a new model is released just after you make a significant cash committment to the older model - these things aren't cheap. Nonetheless, if you have bought a machine that you like and one that does what you need it to, you can't let yourself fall into that trap of feeling ripped off. Buy the machine you need, when you need it, and you should never feel disappointed.

 

[wrldwzrd89]

As I've said a million times before, if you always wait for the next best thing, you'll always be waiting.

That being said, I know it's frustrating when a new model is released just after you make a significant cash committment to the older model - these things aren't cheap. Nonetheless, if you have bought a machine that you like and one that does what you need it to, you can't let yourself fall into that trap of feeling ripped off. Buy the machine you need, when you need it, and you should never feel disappointed.

I couldn't agree more. If you need a machine now, buy one now - don't worry about how soon Apple will release another update. If you need to, just stay away from MacRumors for a while

For those that want this kind of info, there's always the Buyer's Guide right here.

 

[mai]

the best what we can expect in jannuary for a pb

a dual core g4 is not available at the moment,
if its true, then what would u prefer?
a g5, at 1.6/1.8 ghz, 90nm, 512kb l2 cache, at 400/450 mhz (1:4) fsb or
a g4, at 1.6/1.8 ghz, 90nm, 1mb l2 cache, at 200 mhz fsb

not a big differece, hm?
maybe just the power consumption/heat generation
and i think g4's would be much cheaper than a new generation g5 with new casing, anything else would be a surprise

 

[thatwendigo]

a dual core g4 is not available at the moment,
if its true, then what would u prefer?
a g5, at 1.6/1.8 ghz, 90nm, 512kb l2 cache, at 400/450 mhz (1:4) fsb or
a g4, at 1.6/1.8 ghz, 90nm, 1mb l2 cache, at 200 mhz fsb

There won't be a dual-core "G4" any time in the near future. The MPC8641D is a derivitive in the same way that ther Pentium-M is a derivitive of the Pentium 3, in that the core logics are more advanced versions of a very efficient design. Also, if you'd been paying any attention to what people have been saying, the 8641 cores won't have a frontside bus, thus eliminating that potential slowdown. Just to be fair to both sides, I'd also like to point out that the AntaresMP and AntaresSP 970 cores are supposedly going to have 1MB of L2 cache per processor, as are the 8641 and 8641D.

not a big differece, hm?
maybe just the power consumption/heat generation

In a laptop, "just the power consumptuion" is an enormous consideration that defines the entire system. Dismissing something like that completely misses the entire point of portable computing.

 

[wdlove]

There won't be a dual-core "G4" any time in the near future. The MPC8641D is a derivative in the same way that there Pentium-M is a derivative of the Pentium 3, in that the core logics are more advanced versions of a very efficient design. Also, if you'd been paying any attention to what people have been saying, the 8641 cores won't have a frontside bus, thus eliminating that potential slowdown. Just to be fair to both sides, I'd also like to point out that the AntaresMP and AntaresSP 970 cores are supposedly going to have 1MB of L2 cache per processor, as are the 8641 and 8641D.

Are you also saying that we shouldn't expect a significant upgrade to the PowerBook line any time in the near future? Because it also seems that there are many things to overcome with the use of a G5 also.