[qwerty2]

Check it out and use Sherlock or Watson to translate if you can't read French. It's under this title: 2004-08-24 - G5 2.5 overclock�s

http://croquer.free.fr/

What do you think? IBM has been having such problems that I find it is easy to believe this. What if Apple only had a 2.2 GHz to show from last year all the way up until now? That would have been a total disaster and would be motivation for overclocking a 2.2.

[djohnson]

Originally posted by qwerty2:


http://croquer.free.fr/

[Catfish_Man]

Of course, IBM says it's speced to run at 2.5GHz, but what do they know?

[qwerty2]

Originally posted by djohnson:

Well, I agree -- there is no proof and it is likely just a ploy to get easy hits. (If that's what you mean by the roll eyes.) I wonder if this can be easily proven false?

[qwerty2]

Originally posted by Catfish_Man:
Of course, IBM says it's speced to run at 2.5GHz, but what do they know?

Since it is IBM who is having all the trouble that is delaying Power Macs excessively, isn't that a little like the proverbial "fox guarding the hen house" to say we are to trust that it's a "true" 2.5 because IBM said it is? Then again, I know of no other way to confirm or deny that it is what it is except to go along with what IBM says.

[Link]

They're just trying to score hits. There's always been rumors flying around because some moron posted the same thing about dual 1.42s but that's just not true.

Apple isn't that stupid. IBM PROBABBLY wouldn't let them sell overclocked processors marketed as such to begin with. Sure one can point at the cooling system but that's just to keep it running *SILENT*. Then again apple forgot one of the biggest heat producers in the g5 with that water cooling monster --- the system controller

[Catfish_Man]

Originally posted by Link:
They're just trying to score hits. There's always been rumors flying around because some moron posted the same thing about dual 1.42s but that's just not true.

Apple isn't that stupid. IBM PROBABBLY wouldn't let them sell overclocked processors marketed as such to begin with. Sure one can point at the cooling system but that's just to keep it running *SILENT*. Then again apple forgot one of the biggest heat producers in the g5 with that water cooling monster --- the system controller

Actually iirc the water cooling system helps cool the system controller as well. I can't remember where I read this though. The main reason for the water cooling seems to be that a 2.5GHz 970fx dissipates 50 watts over 66 mm^2, while a 2GHz 970 dissipates 66 watts over 118 mm^2 (iirc). Less heat total, but much more per mm^2 (I'm pretty sure we had this discussion on this forum recently though...)

My main issue with people saying chips are "overclocked" is that overclocking means running it faster than its rated speed. The manufacturer can rate it at whatever speed it'll run at. The only way it could be overclocked is if Apple was running it faster than IBM rated it, but we know that IBM rates it at 2.5GHz. There do seem to be some system life span issues at high voltages for the 970fx though, so I suppose running it at those could be considered something like overclocking.

[Maflynn]

I always thought the 2.5 was an overclocked 2.0.

IBM was unable to produce a 3GHz and they are having yield issues with their fabrication plant, and apple installed a water cooling rig inorder to make it silent. What about all of those claims apple made that they put in 9 fans that run at low speed so that the G5 would be quiet.

It would seem to me that if someone wanted to overclock a 2.0GHz they would want a water cooling system installed since that beast would probably melt into a puddle of slag w/o it.

Mike

[Eug Wanker]

What is it with this infatuation with claiming Apple uses overclocked chips? This has been claimed with a bunch of different chips Apple has used, and every time these claims have been proven wrong.

Note also that IBM freely publishes info about their 2.5 GHz parts:



BTW, croquer is a complete waste of time, not just with their current overclocking article, but with all of their articles.

[EMC]

It's not like IBM built a 2.2Ghz processor and overclocked it. You don't build a clock speed into a design. IBM builds lots and lots of G5s, about which only 1/4th work. Of the working chips, each one is tested at a variety of clock speeds. The chips which function over 2.5 Ghz, get put in the 2.5Ghz bin. The ones that fail above 2.0 Ghz get put in the 2.0Ghz bin. Once they've determined this, Apple hardwires the motherboard to run a clock rate compatible with these chips.

Overclocking is the process of raising the clock rate above spec. Any one of the 2.5 Ghz chips might be capable of being run at a higher speed. IBM just knows that raising the clock speed will on the average cause to many chips to fail. So, if you overclock your proc, you might find you were lucky enough to get one of these better manufactured chips. Unfortunately, while these higher speeds may function initially, they can also heat of the chip to higher temperatures, and heat exponentially increases the electron transport of the copper wiring on the chip itself, causing it to literaly dissolve away much faster and break the chip.

Thus, if the 'water-cooled' heat pipe can disperse more heat than a traditional fan, it is possible that the chips can be run at a higher rate than with conventional cooling. Still, they are not 'overclocked' in the traditional sense. IBM just approves their use at 2.5Ghz so long as they are appropriately cooled

[Maflynn]

Originally posted by Eug Wanker:
What is it with this infatuation with claiming Apple uses overclocked chips?

Well for me, I wouldn't catorgize it as a infatuation.
Based upon the what I see from IBM and Apple. I think that either IBM really has heat issues with the G5 or Apple OC'd the hell out of the 2.0.
I cannot understand why apple would install such a complex and expensive cooling system just in the name of quiet there's more to the story.

Mike

[Shaddim]

Correct me if I'm wrong here, but it's only considered "overclocking" when the end user is manipulating the processor frequency beyond it's "bin" rating. By the term's very nature, it's not possible for a manufacturer to overclock anything.

[BrunoBruin]

Croquer also claimed, in May, that Apple would unveil PPC975-based Power Macs at 2.2, 2.4 and 3.0GHz at WWDC.

[Eug Wanker]

Originally posted by Maflynn:
Well for me, I wouldn't catorgize it as a infatuation.
Based upon the what I see from IBM and Apple. I think that either IBM really has heat issues with the G5 or Apple OC'd the hell out of the 2.0.
I cannot understand why apple would install such a complex and expensive cooling system just in the name of quiet there's more to the story.

Mike

If you look at the IBM specs for the chips, the 2.5 puts out no more heat than the 1.8 GHz G5 970. In fact, the G5 2.0 970 is hotter than the G5 2.5 970FX. If you look at the above graph, IBM pegs the max power of the 2.5 at 100 Watts, which is toasty, but not unreasonably so. Intel chips are currently in the 130 Watt range.

However, the 90 nm 970FX chips are much more dense, because their die size is very small. Conventional cooling has trouble dealing with high heat density, even if the overall heat isn't increased. I posted more on this here.

[fiesta cat]

Originally posted by qwerty2:
Since it is IBM who is having all the trouble that is delaying Power Macs excessively, isn't that a little like the proverbial "fox guarding the hen house" to say we are to trust that it's a "true" 2.5 because IBM said it is? Then again, I know of no other way to confirm or deny that it is what it is except to go along with what IBM says.

We are talking about machines around $3000. Some of the 64-bit/PPC based servers that IBM sells start at $30,000, and some setups can easily top $300,000.

I don't think IBM would play fast and loose with numbers on the G5s. If they did, and were caught, they wouldn't be concerned about a bunch of pissed of Mac owners, they'd have larger problems, namely those IT managers who spend $100,000s and more every year who would be questiong the rest of their 64-bit/PPC lineup.

[booboo]

Originally posted by BrunoBruin:
Croquer also claimed, in May, that Apple would unveil PPC975-based Power Macs at 2.2, 2.4 and 3.0GHz at WWDC.

Well maybe Croquer were right - just premature ;-)

[deboerjo]

I don't think many of you understand what "overclocking" means, particularly based on querty2's statement. Read EMC's post. If a chip meets certain specifications, and is operated within those manufacturer's specifications, it's not overclocked. Overclocking means running a chip outside of it's specifications. IBM has set the specifications for the G5 2.5GHz. Apple is meeting those specifications. By definition, no overclocking is involved.

Now, maybe it's more accurate to say IBM is pushing the limits of the design to get it up to those speeds, much like what Intel did with the 600MHz Katmai, and the 1GHz/1.13GHz Coppermine Pentium IIIs, and is doing now with the Pentium 4s. What I mean by this is they raise the voltage and lower the maximum safe operating temperature specification. For example; early coppermine Pentium IIIs have a Tmax of 82 degrees C, and a specified voltage of 1.6V. That means the computer manufacturer is required to ensure that the processor temperature never rises above 82 degrees, and the motherboard has to supply 1.6V (+/- 2.5%) power. As the Coppermine pushed past 1GHz, Intel lowered Tmax to 62 degrees, and the specified voltage went up to 1.75V. Many people at the time called this an "Intel-certified overclocking". Again, by definition, this is not the case. However, it did demonstrate that Intel was under intense pressure from AMD at the time and was shipping products that were not entirely polished. Ultimately the first batch of 1.133GHz Pentium IIIs were recalled. I don't think IBM is going to quite such extremes as what Intel did in that example, but I'm fairly sure the 2.5GHz G5 has a lower Tmax and higher Vcore specification than the 2.0GHz parts.

-Jon

[Scotttheking]

Originally posted by qwerty2:
Since it is IBM who is having all the trouble that is delaying Power Macs excessively, isn't that a little like the proverbial "fox guarding the hen house" to say we are to trust that it's a "true" 2.5 because IBM said it is? Then again, I know of no other way to confirm or deny that it is what it is except to go along with what IBM says.

Erm, IBM ships it rated to run 2.5GHz. It runs at 2.5GHz. Thus, it's a 2.5GHz chip. If a chip it rated to run at a speed, marked accordingly, and it runs at that speed, it isn't over-clocked.

[Maflynn]

But thats not what we are saying. I believe that apple and/or ibm are using 2.0's in the 2.5 and that's the reason why they had to put a liquid cooled unit into the box.

Mike

[Catfish_Man]

Originally posted by Maflynn:
But thats not what we are saying. I believe that apple and/or ibm are using 2.0's in the 2.5 and that's the reason why they had to put a liquid cooled unit into the box.

Mike

All evidence so far (Apple and IBM's public statements, as well as logic) is against that, but there's no real harm in you believing it.

[lngtones]

Originally posted by Maflynn:
Well for me, I wouldn't catorgize it as a infatuation.
Based upon the what I see from IBM and Apple. I think that either IBM really has heat issues with the G5 or Apple OC'd the hell out of the 2.0.
I cannot understand why apple would install such a complex and expensive cooling system just in the name of quiet there's more to the story.

How do you know the water cooling wasn't planned from the beginning? It could have easily been planned to always be in the G5 but it wasn't ready when the rest of the system was so they released the G5 without it. Probably not as exciting as conspiracy stories...

[Maflynn]

Originally posted by lngtones:
How do you know the water cooling wasn't planned from the beginning?

I don't know and its quite plausible that Apple planned this radical design for a long time.
They've been known to push the envelope on design from time to time

As for the G5 2.5 not being OC'd that too is possible and in all likelyhood probable. The actions and circumstances surronding the G5 2.5 really do lead me to believe that the 2.5 could be overclocked.


Mike

[FlatLyna]

Perhaps Apple is using the 2.5s to mass test the liquid cooling unit. If the thermal output is in the same bracket as the old 970 1.8s then this solution may be overkill and not yet really necessary. That then paves the way for when 3Ghz+ becomes available the cooling solution will be proven.

Then again it might just be about noise. The existing heatsync design may have needed the fans to be spinning higher whereas the liquid cooling abates that need.

[Catfish_Man]

Originally posted by FlatLyna:
Perhaps Apple is using the 2.5s to mass test the liquid cooling unit. If the thermal output is in the same bracket as the old 970 1.8s then this solution may be overkill and not yet really necessary. That then paves the way for when 3Ghz+ becomes available the cooling solution will be proven.

Then again it might just be about noise. The existing heatsync design may have needed the fans to be spinning higher whereas the liquid cooling abates that need.

Apple has stated that the liquid cooling is because the 970fx dissipates more heat per mm^2, although less total heat.

[Groovy]

Originally posted by Eug Wanker:
If you look at the IBM specs for the chips, the 2.5 puts out no more heat than the 1.8 GHz G5 970. In fact, the G5 2.0 970 is hotter than the G5 2.5 970FX. If you look at the above graph, IBM pegs the max power of the 2.5 at 100 Watts, which is toasty, but not unreasonably so. Intel chips are currently in the 130 Watt range.

However, the 90 nm 970FX chips are much more dense, because their die size is very small. Conventional cooling has trouble dealing with high heat density, even if the overall heat isn't increased. I posted more on this here.

true the die is smaller but the package is the same size and the heat is pretty
even across the surface (as tested by us in our server room) so I do not totally
buy into the density argument. Furthermore, the 970FX 2.0 has no liquid cooling
system. If it did then yeah I would be more inclined to believe there is a major
density issue. note I'm not saying there isn't, I'm saying it is not a big an issue
as apple is making it out to seem.

Also what bin a part goes into depends on what IBM wants to set the test rules at.
Are the 2.5 GHz parts rated at standard room temp of 72F with standard heat sink
fan in mind or did apple tell IBM they will have liquid cooling and here are the
specs and then IBM rates the parts based on the liquid cooling that they KNOW
will be used. A standard rating may only yield a 2.2 GHz part but with the LQ
rating it would rate a 2.5GHz part.

Is that over clocking? I would say yes but the only way to know for sure is
to find out how IBM is rating the parts.

Which brings up another interesting thing. The Army uses liquid nitrogen cooling
in their thermal sights (the M1-Abrams and Bradley's etc..) A very compact system indeed
that can cool to -300 degrees. The same cooling is used on many battle field servers.
It is a very compact system and could easily be used in desktop PC's. If Apple
ever decided to use them then the 2.5 GHz parts I bet COULD reach 3.5+ GHz. I'm not
kidding. The Army has gotten MAJOR speed boost from the liquid nitrogen cooling
system they use. So now apple says to IBM "we are going to use liquid nitrogen cooling"
IMHO IBM would need to change the rating system since it is that far off from normal
cooling. So did they do it here? I wish I knew

[qwerty2]

Originally posted by Groovy:
Also what bin a part goes into depends on what IBM wants to set the test rules at.
Are the 2.5 GHz parts rated at standard room temp of 72F with standard heat sink
fan in mind or did apple tell IBM they will have liquid cooling and here are the
specs and then IBM rates the parts based on the liquid cooling that they KNOW
will be used. A standard rating may only yield a 2.2 GHz part but with the LQ
rating it would rate a 2.5GHz part.

Is that over clocking? I would say yes but the only way to know for sure is
to find out how IBM is rating the parts.
....
So now apple says to IBM "we are going to use liquid nitrogen cooling"
IMHO IBM would need to change the rating system since it is that far off from normal
cooling. So did they do it here? I wish I knew

Well, thanks for that, Groovy. I couldn't articulate it but I think you put well what my concern is -- what measurement IBM used to set the test rules at for this batch of problematic CPUs. That's not just simplistically asking "Did IBM overclock 2.0 or 2.2 GHz CPUs?"

[Maflynn]

Yes well put.

Mike

[power142]

I don't see a good reason why we should be concerned about how or why Apple/IBM rates and cools the chips so long as they are prepared to stand by warranty commitments. It's a fact of modern highly rated computer design that alternative methods of cooling be explored - keeping a system component cool is not challenge unique to IBM/Apple.

It will be interesting to gauge however, 1 year from now, if the liquid cooled machines prove to be as reliable as their brethren - I see no reason to suspect they won't be so. After all, IBM isn't in the business of selling second-rate computer parts any more than Apple, and I can't imagine IBM giving Apple such parts just to keep them quiet - it's just more than their name is supposed to be worth.

[qwerty2]

"It will be interesting to gauge however, 1 year from now, if the liquid cooled machines prove to be as reliable as their brethren - I see no reason to suspect they won't be so." - power142

Well, I agree that it will be interesting. I wouldn't have purchased one if I thought it was very likely to be problematic, however the liquid cooling and the trouble getting to 2.5 GHz bothered me enough that this was the first Mac that I've purchased Apple Care for. I've had 5 other Mac that I never purchased Apple Care for and never really needed it. So, I guess that says that although I don't think my 2.5 will have CPU or liquid cooling problems, that I feel a bit more risky buying this machine than the others. If I get 3 solid years of use out of it then I'll be happy. I'll be wanting a new machine by then anyway.

[Eriamjh]

There is no such thing as overclocking. The processor simply runs or it doesn't. The processor either requires a little cooling or a lot to run properly. Companies have made CPUs run at 4,5, or 6 GHz while submerged in liquid nitrogen.

Reliability is what really matters. Will a 2.5GHz G5 still be working after 4 years on constant use compared to a 2GHz G5 without liquid cooling? Only time will tell.

Weren't the 1.4GHz G4s supposedly overclocked? Whatever happened to them?

[power142]

I really excited about my future dual 2.5 and I'm hoping it'll be quieter than the rev A dual 2.0 I have at work.... that thing really howls when it gets given some work to do.

Meanwhile, my nearly 18-month old dual 1.42 G4 continues to rule the roost at home - I'm ready to upgrade to the 2.5 now for the extra speed gains, but the G4 will still get plenty of useful work to do in the coming years.

[blakespot]

Originally posted by power142:
I really excited about my future dual 2.5 and I'm hoping it'll be quieter than the rev A dual 2.0 I have at work.... that thing really howls when it gets given some work to do.

Meanwhile, my nearly 18-month old dual 1.42 G4 continues to rule the roost at home - I'm ready to upgrade to the 2.5 now for the extra speed gains, but the G4 will still get plenty of useful work to do in the coming years.

My dual 2.5 will ship shortly. Glad to see it soon. But I must say my 3 year old dual G4 800 does not feel slow. Desktop apps fly and HALO is pretty close to glass.



blakespot

[Maflynn]

Originally posted by Eriamjh:
There is no such thing as overclocking.

I have to disagree with you, overclocking is raising the speed of a cpu beyond its rated megahertz. pushing it further may require radical cooling solutions, i.e., liquid.

Some of the questions in this thread are, mre related to how IBM rated the chips, they may have been "2.0" but they changed their rating system and introduced the need to cool them via liquid, is that overclocking technically no.

How about this: Apple took 2.0's and pushed them to 2.5, is that overclocking, yes because ibm only rated them at 2.0. (I really don't know if apple did this or not).

The point is that manufacturers specifiy how fast the chip can go. If another company, or a user increases the cycle rate beyond its rated speed then its overclocking, which then means the cpu may be prone to crashes and early failure.

Its a fine line when your talking about the two vendors that supply the cpu and computers to us. Do I personally believe that IBM changed their rating, or used an incredible liberal rating system? Not really they've been making chips for a long time. Do I believe that Apple really did overclock the 2.0. Well no, but it wouldn't surprise me if they did.

Mike

[Groovy]

Originally posted by Maflynn:
I have to disagree with you, overclocking is raising the speed of a cpu beyond its rated megahertz. pushing it further may require radical cooling solutions, i.e., liquid.

Some of the questions in this thread are, mre related to how IBM rated the chips, they may have been "2.0" but they changed their rating system and introduced the need to cool them via liquid, is that overclocking technically no.

IMHO it is still overclocking, just an IBM sanctioned overclock since they did it
and are willing to back its quality and life expectancy with their warranty.
One funny thing which may point to it being an IBM sanctioned overclock is that
you can not buy a 2.5GHz parts directly from IBM. You have to go through Apple
via towers. We tried for our servers and we couldn't. When we asked why,
we were told there is not adequate cooling.

Things that make you go hmmm...

[qwerty2]

Originally posted by Groovy:
IMHO it is still overclocking, just an IBM sanctioned overclock since they did it
and are willing to back its quality and life expectancy with their warranty.
One funny thing which may point to it being an IBM sanctioned overclock is that
you can not buy a 2.5GHz parts directly from IBM. You have to go through Apple
via towers. We tried for our servers and we couldn't. When we asked why,
we were told there is not adequate cooling.

Things that make you go hmmm...

Man, Groovy, you have my attention with that. That would be interesting to verify. What you are saying is that IBM can't sell the 2.5s because they don't have the cooling system that Apple has. It was clear that you said that (with different wording), but I just had to repeat it because it is pretty huge.

[*Mhz]

Originally posted by Groovy:
IMHO it is still overclocking, just an IBM sanctioned overclock since they did it
and are willing to back its quality and life expectancy with their warranty.
One funny thing which may point to it being an IBM sanctioned overclock is that
you can not buy a 2.5GHz parts directly from IBM. You have to go through Apple
via towers. We tried for our servers and we couldn't. When we asked why,
we were told there is not adequate cooling.

Things that make you go hmmm...

You have it right. There is no such thing as absolute chip speed. If people took a while end checked out some data sheets of different electronic components they would find out that the oparation frequencies depend on thermal conditions.

Regarding capacitors their lifetime is dependant on their operating temperatures. And if the question is about processors their rated frequencies are ALWAYS dependant of the junction temperatures. E.g. even a 2.2GHz chip might not be overclocked if it runs at 2.5GHz if it meets the required thermal conditions. If chips are runned at higher freqencies than IBM has in its specs than the chips are overclocked.

Now the real question is who knows the specs? I think only Apple and IBM. And the specs might be different for a good customer whos solutions are well known unlike John Doe. That is why all this is only apeculation without facts.

[Groovy]

Originally posted by qwerty2:
Man, Groovy, you have my attention with that. That would be interesting to verify. What you are saying is that IBM can't sell the 2.5s because they don't have the cooling system that Apple has. It was clear that you said that (with different wording), but I just had to repeat it because it is pretty huge.

Actually it is easy to verify. Contact IBM and ask about 2.5 GHz parts and their
availability to you.

Anyway, I EXPECT IBM to improve the 970FX so that 2.5GHz for use in Xserves will
be sooner rather than later. I bet at that time we shall see 3GHz in the towers. Same
chips but in different thermal environments. nothing new.

if IBM actually has true 2.5GHz bin parts for use in standard cooling boxes then
where are they? Even VT was only given 2.3GHz parts and they have a down right
chilly server room.

[Groovy]

Originally posted by *Mhz:
That is why all this is only speculation without facts.

yes speculation and some facts. "without facts" is incorrect as we do have some
facts. One fact being you can not get the part any way other than through Apple via
towers. Another fact is the Xserves do not have 2.5GHz parts and even VT was only
given 2.3Ghz parts and they have a down right cold server room. In fact I bet that
is the main reason why they were able to get 2.3GHz rated parts. Their server room
temperature is so low and thus the air intake so low it is almost as good as the liquid
cooling in towers. The VT setup is another case of IBM/Apple knowing the thermal
environment ahead of time.

[Eug Wanker]

Originally posted by Groovy:
yes speculation and some facts. "without facts" is incorrect as we do have some
facts. One fact being you can not get the part any way other than through Apple via
towers. Another fact is the Xserves do not have 2.5GHz parts and even VT was only
given 2.3Ghz parts and they have a down right cold server room. In fact I bet that
is the main reason why they were able to get 2.3GHz rated parts. Their server room
temperature is so low and thus the air intake so low it is almost as good as the liquid
cooling in towers. The VT setup is another case of IBM/Apple knowing the thermal
environment ahead of time.

So what? If IBM says a 2.5 GHz 970FX requires a certain level of cooling then what's the difference? The 3.6 GHz P4s are also very hot, but I wouldn't call them overclocked either.

These "2.5 is overclocked" statements really make no sense.

[Al G]

Groovy, to be thorough and put things in context here, you should probably make it clear on which, if any, 970 and 970FX parts IBM is willing to sell to anyone other than Apple. IOW, can anyone buy 1.8 or 2.0 parts?

[qwerty2]

Originally posted by Eug Wanker:
So what? If IBM says a 2.5 GHz 970FX requires a certain level of cooling then what's the difference? The 3.6 GHz P4s are also very hot, but I wouldn't call them overclocked either.

These "2.5 is overclocked" statements really make no sense.

So let's get away from the term "overclocked" and just say that the 2.5GHz chip is one that requires extraordinary cooling, one that IBM is having a great deal of trouble producing, and one in which IBM has yet to use in any of it's own systems. Will IBM work out a similar liquid cooled system in order to use them in their own machines that use PPC once supplies are better and Apple gets all they need, or are is IBM just going to pass on this one and pick with with the next PPC that has lower heat? Jobs was talking about other companies using the flat panels that Apple rejects, but is Apple using a chip that IBM rejects? It's a lovely rejected chip, if so, because I have a 2.5 and I'm enjoying the machine so far. It's kind of noisy and the fans come on more than I would like, and I've started using the ceiling fan in my office. No, just kidding about that last part...sort of.

[Eug Wanker]

IBM doesn't use the 2.5 because it doesn't really have big plans for desktop PPC machines at this point. There will be some, but IBM's big iron stuff is POWER4 and POWER5, and the cluster stuff is BlueGene.

It does use the G5 2.2 for blades, and that's faster than even Apple's fastest Xserves, and blades are often more cramped than Xserves.

[Groovy]

Originally posted by Eug Wanker:
So what? If IBM says a 2.5 GHz 970FX requires a certain level of cooling then what's the difference? The 3.6 GHz P4s are also very hot, but I wouldn't call them overclocked either.

These "2.5 is overclocked" statements really make no sense.

3.6 GHz P4s do NOT need liquid cooling and you buy them without having
to go through a box maker. that is just 2 of many HUGE differences you
are turning a blind eye to.

In the Case of 970FX at 2.5GHz, it needs LQ, else you can not run at that speed.
The 3.6 GHz P4s do not need LQ. The only reason 2.5GHz was attainable was
because of the LQ system in towers. Apple even admitted this

Please stop going on about stuff making "no sense" and start using some
common sense

[Groovy]

Originally posted by Al G:
Groovy, to be thorough and put things in context here, you should probably make it clear on which, if any, 970 and 970FX parts IBM is willing to sell to anyone other than Apple. IOW, can anyone buy 1.8 or 2.0 parts?

We tried to get 970FX 2.5GHz parts. Anyone can buy parts from IBM but your
volume will determine exactly how/who/where.

Could YOU call up and buy just 1 directly from IBM, no of course not. They would
point you to someone who can though. In our case there was no one to get that
part from. It was implied that part it is a made for Apple only part to be used only
with their cooling system.

[Groovy]

Originally posted by Eug Wanker:


It does use the G5 2.2 for blades, and that's faster than even Apple's fastest Xserves, and blades are often more cramped than Xserves.

You are NOT helping your case by saying even IBM is not running 2.5GHz in their
OWN blades.

Blades are mainly used in server rooms which almost always have dedicated AC with
sub 70 temps and still IBM is not running 2.5GHz in their own blades.

VT is running 2.3GHz Xserves so your 2.2GHz theory has even more holes in it.
VT is the bleeding edge on Xserves and even they couldn't get the thumbs up to
to use 2.5GHz parts.

[BurningBright]

whither does IBM use 970s in blades?

They use Intel Xeon and POWER series processors to power their blades. There is one 970 option, clocked at 1.6, yes 1.6ghz. Not 2.2, certainly not 2.5 ghz.

That is just not feasible in their designs.

[Link]

There actually are 970 powered blades.. can't remember what model or whatever

[macgfx]

Don't the G5's get "Marked" when IBM bins them out?

IF so then every one with a G5 2.5 pull of that LQ cooling and report back with the die markings, Quick I'll hold my breath............................................ .................................

If Apple is using a 2.0 bin marked chip @2.5 it is over-clocked.

If IBM marked them 2.5 it is not.

If IBM marked 2.0's 2.5 only for APPLE with LQ cooling it's still not over-clocked. It's like a 675HP 351 C.I.D with a Blower giving that Last 275HP.

I don't think anyone has ever proven that OCing a Chip will make it Fail early as long as it's stable and does not overheat.

[Maflynn]

What an interesting thread which definitely has some insights to people think.

We've gotten that there is no such thing as overclocking, apple overclocks, ibm overclocks, and every permutation of these three.

fwiw, I've been on the side that it looks like apple OC'd the 2.5s but I really don't know, i.e., no empirical evidence to support my thesis.

Anywho, I don�t want to let ibm off the hook either, if they rated the chip at 2.0 but because of pressure from apple, said you can run them at 2.5 as long as you cool the hell out of them, then that�s as bad a consumer doing it, i.e., it is overclocking.

As to the perceved life span, who knows, it seems that a lot of people here (myself included) go through computers pretty fast, even though I keep saying this one will last me a while 3 years (if I'm lucky) roll around and bam a new mac on my desk. So any shorten life spans won't be noticed in those circumstances.

Mike

[macgfx]

Ok new Term..............

Over-Marked ,OMing...............a process by witch a chip maker marks a chip to a speed higher than it has been rated in the past in order to make the chip able to run @ a higher clock speed with one or more of the flowing:

1. Vboost (see voltage boost)
2. Higher Bus speed
3. Added cooling

From what has been said here my money is also on the 2.5 being OMed, and as has also been pointed out IBM is not the only Chip maker to do so(see Intel).

Seems the trend these days is OCing before sale(see Giga G4 upgrades,see also Geforce OC, see also Radeon XT.

What it all seems to come down to is the high demand for speed and the premium we are willing to pay for it has pushed Chip makers and 2nd party manufactures to HAND sort chips Rather than BIN sorting.

That said I don't like it. In the past you could buy a CPU GPU RAM what have you and feel you hand an amount of head room. Now all head room is sorted and marked UP. It has good points for those who do not par-take of the OCing, but those of use who do seem Doomed to Die in a Zipper-less Camel suite.