Apple compares heat dissipation of G5 vs. x86: 90 nm G5 2.0 dissipates 55 Watts max

[Eug]

Apple marketing, so take it as you will... but I think the numbers are in the right ballpark.

Max heat dissipation --> :camera:

G4 1.33: 45 W
G5 2.0: 55 W
Athlon XP 3200+: 77 W
Itanium 2 1.5 6M: 89 W
Opteron 246 2.0: 107 W
Xeon 3.2: 110 W

 

[MadRat]

And what does their equivalent to Northbridge emit?

 

[Eug]

Originally posted by: MadRat
And what does their equivalent to Northbridge emit?

I have no idea. What does Intel's?

BTW, I found this: Chris Hare's Processor Electrical Specifications Page.

 

[Barnaby W. Füi]

Seems reasonable to me. Unlike all of the anti-apple zealots around here. Prepare for the onslaught...

 

[Eug]

Originally posted by: Eug
Apple marketing, so take it as you will... but I think the numbers are in the right ballpark.

Max heat dissipation --> :camera:

G4 1.33: 45 W
G5 2.0: 55 W
Athlon XP 3200+: 77 W
Itanium 2 1.5 6M: 89 W
Opteron 246 2.0: 107 W
Xeon 3.2: 110 W

Hmmm...

It seems Apple may have mixed up some numbers according to the Ars people... The Opteron numbers did seem high. It should be:

G4 1.33: 45 W
G5 2.0: 55 W
Athlon XP 3200+: 77 W
Opteron 246 2.0: 89 W
Itanium 2 1.5 6M: 107 W
Xeon 3.2: 110 W

Otherwise the numbers are accurate. That's pretty damn impressive. The G5 is HALF the wattage of the Xeon 3.2. That said, the proper comparison should probably be vs. the Xeon 2.8, since in benchmarks the G5 performs more like a 2.8, and the 2.8 has a max wattage of 86 Watts. The 3.06 is at 102 Watts.

 

[tart666]

Intel / AMD should start measuring heat dissipation in horsepower. 110W = 1/7 horsepower. That way, larger is better! I can just see the proud owners of a 3HP dell...

 

[Eug]

Originally posted by: tart666
Intel / AMD should start measuring heat dissipation in horsepower. 110W = 1/7 horsepower. That way, larger is better! I can just see the proud owners of a 3HP dell...

lol

BTW, it should be noted that the Xeon 3.2 has an extra 1 MB L3.

 

[aka1nas]

I thought the opterons currently consumed a lot less than that number? I.E. I thought that they are closer to 45 watts and the 107 watts is the maximum expected dissipation for the fastest model that they will produce on this core.

 

[Eug]

Originally posted by: aka1nas
I thought the opterons currently consumed a lot less than that number? I.E. I thought that they are closer to 45 watts and the 107 watts is the maximum expected dissipation for the fastest model that they will produce on this core.

The 107 Watts is a mistake. It should be 89 Watts. AMD lists 89 Watts for the 246, and Intel lists 107 Watts for the Itanium 2 1.5. It seems Apple just reversed the numbers by accident.

Either way, the 45 Watt number you quote for the Opteron 246 is way off - the value is twice that - according AMD itself. Actually, 89 Watts for the Opteron is just the TDP. The max power is higher than 89 Watts, but AMD doesn't document a max power rating.

 

[aka1nas]

Ok, thank you for the clarification.

 

[Pariah]

Is that Opteron number right? (the 89W value) Since its release all we have been hearing is how cool it runs compared to Intel's offerings. 89W for a 246 puts it above Intel's 3.2GHz P4 (82W), which doesn't run cool by anyone's standards. Is AMD's listing an actual or theoretical value for the entire line of chips? I think it's odd that on Chris Hare's page linked above that the entire line of Opterons from 1.4GHz through 2GHz has the exact same max heat dissipation while being built on the same process. Has AMD found a secret way to increase clock speed while not increasing heat dissipation? The AMD values are either theoretical or just not accurate.

Edit:
Upon closer inspection, the Opteron is listed at 84.7W, the A64 is 89W. And that value includes the thermal power of the memory controller, which the other CPU's in the list don't include. Not really and apples to apples comparison (no pun intended).

Also, it appears AMD is going to release rack mountable mid and low power consumption Opterons that consume 55W and 30W.

 

[Mingon]

IIRC the AMD value is 89watts max under heavy load (as quoted in their CPU document) whereas intels is usually quoted under normal load. As for the G5 thats 55watts per CPU so in effect you still have 110 watts to match intels performance (p4 3.2). So its seems to me that the net result is (drum roll) they are all quite similar in their heat production when comparing similar performance. what a surprise

edit>

its 89watts for both 2.0 and 2.2 and assumes a case temp of 42degrees! but interestingly the mobile 2.0ghz is 62watts so it seems that these are from better yields

 

[Eug]

Originally posted by: Mingon
IIRC the AMD value is 89watts max under heavy load (as quoted in their CPU document) whereas intels is usually quoted under normal load. As for the G5 thats 55watts per CPU so in effect you still have 110 watts to match intels performance (p4 3.2). So its seems to me that the net result is (drum roll) they are all quite similar in their heat production when comparing similar performance. what a surprise

Yeah, the AMD value is 89 Watts under heavy load. As I understand it the Intel value (TDP) is also under heavy load, but of course doesn't include heat for the memory controller (for obvious reasons).

The G5 is 55 W per CPU, but a dual 2.0 is much faster than a single P4 3.2 for floating point, if you use the fastest compilers on both platforms. For double precision fp a single G5 2.0 performs on average about the level of a P4 2.8. The guys who do computational fluid dynamics stuff say the with the IBM compilers, a dual G5 2.0 performs on par faster than a projected P4 5 GHz (using Intel's compilers).

Thus, one would expect a dual 2.0 G5 to compare in the ballpark of a dual Xeon 2.8, at least for scientific code (ie. *nix apps, etc.). That would be 110 Watts for the G5 compared to 176 Watts for the dual Xeon.

Now if you're talking games, that's a whole different kettle of fish. Games often suck on Macs, including fast Macs, presumably from optimization, compiling, porting issues, etc. Indeed, A P4 3.2 single CPU system would destroy a G5 2.0 dual for most games.

 

[arsbanned]

lol, Apple reversed the numbers by accident? I wonder what else they might do "by accident"

 

[Accord99]

Originally posted by: Mingon
IIRC the AMD value is 89watts max under heavy load (as quoted in their CPU document) whereas intels is usually quoted under normal load. As for the G5 thats 55watts per CPU so in effect you still have 110 watts to match intels performance (p4 3.2). So its seems to me that the net result is (drum roll) they are all quite similar in their heat production when comparing similar performance. what a surprise

Intel's definition of TDP is set at a value that is slightly higher than the highest observed power dissipation using a large suite of CPU intensive applications, so it is highly unlikely than any useful realworld application will ever exceed Intel's TDP. AMD doesn't fully define their definition of TDP in their A64 docs, though in their XP docs, it is defined in a similar way to Intels.

 

[Pariah]

Originally posted by: arsbanned
lol, Apple reversed the numbers by accident? I wonder what else they might do "by accident"

Not only did they reverse it, they got the wrong number anyway. The Opteron 246 is listed by AMD at 84.7 W. The AMD64 line is listed at 89 W.

 

[mosco]

Originally posted by: Pariah

Originally posted by: arsbanned
lol, Apple reversed the numbers by accident? I wonder what else they might do "by accident"

Not only did they reverse it, they got the wrong number anyway. The Opteron 246 is listed by AMD at 84.7 W. The AMD64 line is listed at 89 W.

Click Here

shows 89W for the 246 model on the last page.

 

[Pariah]

Thanx for the link, that certainly answers my question about why the Opteron's thermal power doesn't increase with clock speed as Chris Hare's page shows. The answer, his page is wrong and it does increase.