PDA

View Full Version : Top End G5 an overclock?


Converted2Truth
Jul 8, 2004, 12:17 PM
I am surprised that nobody else has mentioned this, but I think the Dual 2.5 is just an overclocked 2.0. Why else would they use liquid cooling? When they went down to a 90nm process, it also reduced heat. That's why the xserve G5 2.0 @ 90nm consummed fewer watts than the 130nm 2.0 in the last generation G5 desktop. So i think that they overclocked the G5 up to 2.5 using 90nm technology, and then were forced to liquid cool the steaming monster.

It's kinda funny that Steve Jobs said that they're improving in speed better than Intel at the developers conference, considering the fact that IBM has only changed their process to 90nm and apple has overclocked a 2.0. If you look at the picture from this angle, IBM hasn't made any speed improvements :eek: The new P4's run quite cool even amongst preformance increases.

Oh well, I still think i'd be awesome to liquid cool a G5. They should have added flourecent lights and stuff though. Too bad they don't sell an upgrade kit for those running 1.6/1.8/2.0 ...after all, it's just an added feature (non-nessesary) to the top-end powermac, right :confused:

Don't get me wrong though... I understand that the G5 rocks the P4 (in specific instances).

phiberoptik957
Jul 8, 2004, 12:50 PM
Your comments lack any trace of logical thought.


I am surprised that nobody else has mentioned this, but I think the Dual 2.5 is just an overclocked 2.0. Why else would they use liquid cooling? When they went down to a 90nm process, it also reduced heat. That's why the xserve G5 2.0 @ 90nm consummed fewer watts than the 130nm 2.0 in the last generation G5 desktop. So i think that they overclocked the G5 up to 2.5 using 90nm technology, and then were forced to liquid cool the steaming monster.

It's kinda funny that Steve Jobs said that they're improving in speed better than Intel at the developers conference, considering the fact that IBM has only changed their process to 90nm and apple has overclocked a 2.0. If you look at the picture from this angle, IBM hasn't made any speed improvements :eek: The new P4's run quite cool even amongst preformance increases.

Oh well, I still think i'd be awesome to liquid cool a G5. They should have added flourecent lights and stuff though. Too bad they don't sell an upgrade kit for those running 1.6/1.8/2.0 ...after all, it's just an added feature (non-nessesary) to the top-end powermac, right :confused:

Don't get me wrong though... I understand that the G5 rocks the P4 (in specific instances).

Laslo Panaflex
Jul 8, 2004, 12:54 PM
Your comments lack any trace of logical thought.

Wow, can you please elaborate on why his comments "lack any trace of logical thought"? I really don't like it when people accuse others of being trolls, and don't even bother to explain why.

Sun Baked
Jul 8, 2004, 12:55 PM
Your comments lack any trace of logical thought.Especially considering that the DP2.5PM may actually be running at a lower maximum power dissipation on the CPUs than the RevB DP2.0PM. ;)

There are quite a few hints and reliable reports that the RevB DP2.0PM is still on the 130nm process which places it's maximum power dissipation much higher than the 90nm DP2.5PM -- not quite the expected situation for an overclock. :rolleyes:

Converted2Truth
Jul 8, 2004, 01:18 PM
There are quite a few hints and reliable reports that the RevB DP2.0PM is still on the 130nm process which places it's maximum power dissipation much higher than the 90nm DP2.5PM -- not quite the expected situation for an overclock. :rolleyes:

They use the 90nm 2.0 G5 in the Xserves. The powermac line is all 90nm now (at least according to apples website...) So, considering that they have 90nm 2.0G5, and that they say they are using the 90nm 2.0G5 in the new Power Macs, then the only reason a chip which consumes less power(manufactured on a smaller process) would consume more power than it's predecessor is because it is OVERCLOCKED, and then liquid cooled.

If i lack any logical thought, let me clarify. The smaller the process, the less heat consumption. The fact that the chip generates less heat allows for the increase of cycles, thus increasing ghz etc. In this case, when they 'increased' the cycles (overclocked), the 90nm 2.0G5 got too hot, and they were forced to liquid cool the badboy. I think that air cooling would have worked, but probably caused the G5 to lose is 'quiet supercomputer' appeal. Liquid cooling is a successful way to stabalize an overclocked CPU. It seems to me that shrinking the process didn't allow for much of a speed increase (without drastic cooling measures).

All I am saying is that if you looked at the two chips (90nm 2.0, 90nm 2.5) they would be the EXACT SAME chips. Kinda like the Motorolla 7455 in my powerbook is clocked at 800mhz and 1000ghz and so on...

phiberoptik957
Jul 8, 2004, 01:24 PM
I am sorry that I wrote above without elaborating, I just get tired of these unsubstantiated claims that companies "over-clock" their processors just because they are hotter than the previous generation. I realize that this is about Apple but the same claims get made often when for example AMD increases the speed of their processors.

Also, I am sorry I jumped the gun and called you a troll, I em going now to edit my post and remove that.

Converted2Truth
Jul 8, 2004, 01:34 PM
I am sorry that I wrote above without elaborating, I just get tired of these unsubstantiated claims that companies "over-clock" their processors just because they are hotter than the previous generation.

I never said they 'overclocked their processors beccause they're hotter'... I am saying that they overclocked their cooler processor (90nm G5) and it got hotter than the prev generation(when they overclocked the 90nm 2.0G5 to 2.5), hence the liquid cooling. This really isn't a hard concept to understand, nor is it NOT logical.

Sun Baked
Jul 8, 2004, 01:35 PM
Actually the report on the RevB DP1.8PM and the RevB DP2.0PM was...
Processor version of new Power Mac G5 (http://episteme.arstechnica.com/eve/ubb.x?a=tpc&s=50009562&f=8300945231&m=9080959175&r=321009974631#321009974631)

Today, I saw two new Power Mac G5s (dual 1.8GHz and dual 2GHz) at local computer store. Both Power Macs were identified as PowerMac7,3. The PVR value of all G5s was 0x00390202. So, they are 970, not 970FX.
It seems that KeyLargo2 is updated. The revision ID of became 0x60, which was 0x20 in original Power Mac G5. The device revision of U3 is more complicated. One (dual 1.8GHz model) was 0xb3 and the other was 0x35. AFAIK U3 device-rev of original G5/dual 2GHz was 0xb3 and that of some initial model was 0x32.The dude is a reliable source, more so than your thoughts of what you think Apple said.

He provided quite a few Japanese translations for the rumor community, and you'll even find out he knows about overclocking -- just google (Michiro Isobe AND overclock)

Plus his findings seem to fit right in there with what Apple Release a couple weeks later...

http://forums.macrumors.com/showpost.php?p=911652&postcount=1

Frohickey
Jul 8, 2004, 01:46 PM
They use the 90nm 2.0 G5 in the Xserves. The powermac line is all 90nm now (at least according to apples website...)


Wanna bet money on that?

I'll bet my whole monthly paycheck that the whole PowerMac line is NOT all 90nm.

Converted2Truth
Jul 8, 2004, 01:53 PM
Wanna bet money on that?

I'll bet my whole monthly paycheck that the whole PowerMac line is NOT all 90nm.

Hey.. stevey said it at the dev conference, and it's on their website... if i was ordering a new G5, i'd want it to be what they advertised.

:rolleyes: but somehow, i fell you could be slightly correct...

osprey76
Jul 8, 2004, 02:00 PM
I am saying that they overclocked their cooler processor (90nm G5) and it got hotter than the prev generation(when they overclocked the 90nm 2.0G5 to 2.5), hence the liquid cooling.

That makes sense to me. I think you're using the term overclocking a bit loosely. I can't imagine Apple would use a processor at a higher clock speed than it was rated by IBM. The 2.0's and the 2.5's probably are from the exact same die (1.8's, too.) The 2.5's just came through the process closer to ideal and, therefore, can run at higher speeds.

Two thoughts on the higher heat load. First, I would guess that power consumption may not ramp up linearly with clock speed. So, there could be a large jump in heat generation between the 2.0 and the 2.5 chips. Second, with the smaller chip, you now have less area to transfer the heat across. This is very pertinent in heat transfer. So, if a new 2.5 uses similar power to the old 2.0, you'll need better cooling to make it happen. If the total chip size scales the same as the process size, then you have (130-90)/130 = 31% less area in the new 2.5 from the old 2.0. That is a large difference (if my assumptions are correct) and might possibly need another wind tunnel machine or liquid cooling.

smllpx
Jul 8, 2004, 02:02 PM
Just FYI
CPU Manufactures don't overclock, users do.

When chips come off the assembly line they are tested to see at what speed they can be run reliably then they are given a speed rating. As the process gets better and refined the chips are able to run at high clock speeds. I believe all the G4's come off the same line.

Overclocking occurs when a user purchases a lower rated chip and runs it at a higher speed or just running the chip past the speed rating.

edit: beaten. doh!

Sun Baked
Jul 8, 2004, 02:14 PM
>Converted2Truth

The 90nm chips are most likely still constrained, with XServe G5 users are still wondering if they'll get their machines before the end of the year, this wouldn't be the case if Apple had ample supply of 90nm CPUs for the PowerMacs.

Everybody thought the choice of 1.8/2.0 in the low/mid was an odd choice when the top end was 2.5 -- again could be lack of 90nm CPUs, and ample supply of 130nm parts.

Though we'll really see if the problems are cleared up when Apple ships the DP2.5PM, or delays it.

The iMac blurb can also point to 90nm problems if it is indeed using a G5, and that's at least September.

JFreak
Jul 8, 2004, 02:16 PM
liquid cooling doesn't mean overclocking but severe overclocking will mean liquid cooling. as the process shrinks, the heat also concentrates into smaller and smaller area, which is more and more difficult to cool with a fan. as the heat is in small area, so would the airflow have to be effective in that small area where the heat comes from. liquid cooling is more effective, as the heat will spread to larger area which is more trivial a task to cool down.

so in other words, the same amount of heat that comes from smaller area is more difficult to cool than the same amount of heat that comes from larger area. and if 2.5GHz@90nm generates the same amount of heat than 2.0GHz@130nm, that will not mean the new ones are overclocked - it just means that the heat becomes too hard to handle with air cooling.

just my reasoning. ask ibm if you want hard facts.

ddtlm
Jul 8, 2004, 02:22 PM
Converted2Truth:

I'd say its not overclocked, for a couple reasons.

First, I don't think its even possible for a machine from Apple, covered by warranty, at default specs, to be overclocked. Its the same silly arguement that we all had over 1.25ghz G4's and then the 1.42ghz G4's. Overclocking is running beyond the settings that the manuafacturer stands behind. I don't care if Apple pops a phase-change cooler on a G5, they are in a position to know if that is a reliable option, and we are not. Apple can push the limits as close as they want without being overclockers, cause they know the issues and they will fix it if something goes wrong.

Second, the thing about the 90nm G5 is that its got a very small die compared to the power it consumes. At 2.5ghz the chip might only use 65W peak, but its so concentrated that it cannot be cooled sufficiently by anything short of liquid. That doesn't mean that the processor can't go faster, it doesn't mean that it shouldn't run at 2.5ghz.

People that equate elaborate cooling with overclocking need to gain perspective on this. They say that the thermal power of new processors has on average increased by 22% each year for something like the last 20 years. Used to be that processors didn't even have heatsinks. Does that mean anything with a heatsink is overclocked?

maxvamp
Jul 8, 2004, 02:45 PM
I have seen this garbage over on the AMD boards too..

I seem to remember when the XSERVE G5 was announced that somewhere here on MacRumors was posted a presentation of some form that actually showed that the fx version of the chip was rated to run up to 2.5 GHz, which at that time lead to a lot of speculation to the next gen of PPC64 which was called the 980. Later, rumors came around about a 975, etc.

The comments about these being overclocked are garbage until Apple starts running them at 2.6 and above. No matter what the speed, these things will need cooled, and there is nothing wrong with water cooling.

On that note, remember, it is usually the heat given off by processors that kill off other components in a computer, and not usuually the temp that the processor itself runs. If you can keep your proc from belching heat, your entire computer will last longer, and Apples are known for longevity.

Max.

varmit
Jul 8, 2004, 02:54 PM
ITS JUST TO MAKE IT QUIETER!!!! Didn't you get that when they started the G5. This sets them up if the cooling unit works well. Then all G5s after these will be liquid cooled because it works a lot better, and they can make the fans not have to work so hard, making it quieter even though the G5s get faster.

Flynnstone
Jul 8, 2004, 02:57 PM
Two thoughts on the higher heat load. First, I would guess that power consumption may not ramp up linearly with clock speed. So, there could be a large jump in heat generation between the 2.0 and the 2.5 chips. Second, with the smaller chip, you now have less area to transfer the heat across. This is very pertinent in heat transfer. So, if a new 2.5 uses similar power to the old 2.0, you'll need better cooling to make it happen.

power consumption used to be (nearly) linear with clock rate. There are two parts to the power equation; 1) leakage current and 2) switching current. There are likely more, but this is the simple explaination.
Switching current is linear with frequency. Double the frequency, double the current , double the power. Leakage current has always been part of the equation, and used to be a small part. It used to be negligible. But as you shrink feature size this has become a bigger part of the equation. I believe leakage power is in the same area as switching power. The 90nm devices "leak" like a sieve! This is a problem Intel & IBM & others are trying to deal with. Using tech like SOI and strained silicon ...

We are seeing a lot more dual (or more) systems to deal with this.
Apple is nicely working in this area :) . I'm glad a PowerMacs are duals now.

King Cobra
Jul 8, 2004, 03:04 PM
Heh, they're all using the 90nm line currently: http://www.macrumors.com/pages/2004/06/20040609083915.shtml

And I don't think there has been a processor overclocking in a PowerMac that has then been released since Motorola's latest G4 revisions.

maxvamp
Jul 8, 2004, 03:16 PM
specs (http://www.macrumors.com/pages/2004/02/20040211143241.shtml)

Sorry, just caught the speculation part. Still, I would suspect that this guy knew something.

Max.

MacNeXT
Jul 8, 2004, 04:27 PM
I understand the OP's thoughts on this, but as someone else said too, a computer out of the box is never overclocked by definition.

A computer is supposed to work ok, if it doesn't, the vendor should solve the problem. Even if the vendor configured a system to run at a higher clock speed as was prescribed by the CPU supplier. Whether it's a bad product or not has nothing to do with that.

Same goes for Apple. Even if the liquid cooled processor runs at a higher speed than IBM's specifications, you can assume it's still Apple quality. Whether that's good or not is another discussion ofcourse :rolleyes:

However, if that's the case, it could turn out to be a bad decision to do it like this because for example the liquid cooling breaks down after a couple of years after warranty (just like, for example, the badly designed powerbook wallstreet hignes did).

In my opinion a liquid cooling system is just something that breaks easily and is very expensive. Therefore I would never buy it. Even if I had the money :rolleyes: It's like having a car that's so fast that it needs a parachute to slow down because there are no breaks that can handle the speed.

ddtlm
Jul 8, 2004, 04:56 PM
MacNeXT:

However, if that's the case, it could turn out to be a bad decision to do it like this because for example the liquid cooling breaks down after a couple of years after warranty (just like, for example, the badly designed powerbook wallstreet hignes did).
Heh, this is the reason I still prefer my CPUs and video cards to be passively cooled. ;) Not that such a thing is always possible.

MacNeXT
Jul 8, 2004, 05:04 PM
MacNeXT:


Heh, this is the reason I still prefer my CPUs and video cards to be passively cooled. ;) Not that such a thing is always possible.

True, even a fan breaks relatively quickly or starts to make a lot of noise after a while.

dudewheresmymac
Jul 8, 2004, 05:40 PM
guys when you shrink the manufacturing process down it decreases the size of the dye and because the dye is smaller then heat is consentrated in a smaller area therefore a more effective cooling method is needed. Also Apple isnt legally allowed to say that the G5 is something that it is not.

Mord
Jul 8, 2004, 05:41 PM
we'll just have to wait untill a dual 2.5GH owner pops the lid and looks on his g5 dies to see what speed they are marked at i did this with my mates dual 1.4GHz g4 back when it was debated weather it was a 1.4GHz cpu or an overclocked 1.25GHz one, the die was marked ppc 7455 1400 in little gold writing over the purple die. proof that they were 1.4GHZ moto rated chips

MikeLaRiviere
Jul 8, 2004, 06:03 PM
A fact I believe no one has yet mentioned is that Intel experienced greater heat when it moved to the 90 nm technology. For example, the P4 3.0 is based on the old technology; the P4 3.0 Prescott is based on the 90 nm process. The Prescott, testers have found, leaks more heat than the Northwood (the older technology), even at the same clock speed. Because chips are now manufactured at 90 nm, heat leaks far more readily.

My guess is that the G5 2.5 processors are not overclocked, but that the 90 nm process in these IBM chips simply requires liquid cooling. I can't rule out the possibility that the 2.5s are overclocked; however, I would have to think that Apple would favor stability over higher advertised clock speeds. As we all know, and as someone in this thread has previously pointed out, clock speed is not the sole determinant of how "fast" the processor will run. Those of us who use Macs know this fact very well.

If there are any objections to/problems with my post, I welcome the replies.

Mike LaRiviere

ddtlm
Jul 8, 2004, 09:00 PM
MikeLaRiviere:

However since the 90nm P-M actually uses less power than the 130nm version, I think we can conclude Prescott is just a very poor design. Everyone is still left wondering what on earth all the transistors in there are doing.

MikeLaRiviere
Jul 8, 2004, 09:46 PM
ddtlm, were you talking about the Pentium-M? I didn't know the M was on a 90 nm process, but then again I haven't been following this line closely. I am, however, impressed with the M line's performance.

The Prescott, in fact, uses more power than the Northwood, if I understand what I read correctly. I think you concur with this fact.

But this is beside the point; I don't imagine the 2.5 GHz G5 is overclocked.

Mike LaRiviere

Sun Baked
Jul 8, 2004, 09:55 PM
we'll just have to wait untill a dual 2.5GH owner pops the lid and looks on his g5 dies to see what speed they are marked at i did this with my mates dual 1.4GHz g4 back when it was debated weather it was a 1.4GHz cpu or an overclocked 1.25GHz one, the die was marked ppc 7455 1400 in little gold writing over the purple die. proof that they were 1.4GHZ moto rated chipsYes, but that was enough proof for everybody to say they were overclock -- even if only by 20MHz. ;)

Apple improved the heatsinks, which allowed them to run a hotter chip.

Apple was also the only customer for a CPU in the high W power dissipation range from Motorola, and since they no longer used, them Motorola isn't supplying them nor testing for them.

The PowerMac G4 upgrade market is just too small a niche market for Freescale to make the effort.

ddtlm
Jul 9, 2004, 03:34 AM
MikeLaRiviere:

Yep yep yep. The 2MB P-M's are the 90nm ones. (Why did IBM only put 512k of L2 on the 970fx? Makes me sad.)

Zaty
Jul 9, 2004, 04:58 AM
Apple can't sell machines using overclocked CPUs. Why? Overclocking may decrease the lifespan of a CPU. Apple would get into lot of troubles if many of those $3000+ machines died after say two years or so because of their CPUs running at clockspeeds above what they're were rated for by IBM.

johnnyjibbs
Jul 9, 2004, 05:40 AM
I don't think they are overclocked. Going from 2GHz to 2.5GHz is a hefty leap and, even on a smaller processes, would run hotter than the previous generation 130nm 2.0GHz chip. The gains for going to 90nm in terms of heat and power consumption are obviously less than IBM and Apple (and Intel for that matter) originally anticipated.

As there is no difference in performance between a 90nm and 130nm chip rated at the same speed then it makes sense to put the old 130nm 1.8 and 2.0 chips in the lower and mid-end G5s, plus maybe 90nm chips that are not stable at 2.5GHz (although these would fit nicely into Xserves and upcoming iMacs).

As someone else pointed out, this must be the reason why they stuck with 1.8 and 2.0 G5 machines and not introduce a 2.2 or similar spec. It also ties in nicely with the iMac situation, which is obviously going to go G5 in September and was obviously originally intended to be released at WWDC or around this time.

Oh, and as someone else also pointed out, quietness is also key. The liquid cooling replaces the alternative - a massive and very noisey fan.