A64 vs. P-M desktop PC: 50% greater power draw

P

Pete

Moderate Nuisance
Moderator
Legend
Now that's impressive. Probably not enough to make me pony up the extra cash, but if I had money to burn, I'd slap a P-M and 6800 or X800XT into an Antec Aria in a heartbeat.

But then, I buy fluorescent light bulbs even though I really doubt I'll make up for the higher up-front costs through lower energy bills. Those suckers blow almost as often as regular incandescents, and their warranties have dropped from a whopping 7 years to a less reassuring 1-2 years.

While it's impressive to see an A64 3500+ with a 6800GT, 1GB RAM, and a Raptor draw only 200W at max load (P95 + RTHDRIBL), it's mind-blowing to see a P-M system do so at only 135W.

I'd love to see them throw Dothan Cel-Ms and 1MB L2 A64s into the mix (for greater price and performance parity, respectively).
 
What's even better is that it is possible to undervolt both the Pentium M and Athlon64 considerably whilst still running at stock speeds.

This article shows you can undervolt the Pentium M to just 1.1V at stock speeds. Athlon64 can be undervolted to a similar level, also at stock speeds. :)
 
Pete said:
While it's impressive to see an A64 3500+ with a 6800GT, 1GB RAM, and a Raptor draw only 200W at max load (P95 + RTHDRIBL), it's mind-blowing to see a P-M system do so at only 135W.
Click to expand...
The test is quite flawed though. I don't doubt the power consumption numbers are correct, but the text specifically mentions "Pentium-M consumes 1 third less than A64" whereas the measurements of course include everything, not just the processor!
It is certain that quite a bit of the difference is due to platforms, not the cpus itself - you know that when in ide, throttled the difference is still more than 40W whereas the A64 according to the datasheet is guaranteed to draw less than 10W...
So you have probably a 35W difference due to platforms alone.
 
Possibly, tho the mobile is 130nm, and they used a 90nm A64 b/c of its supposed power savings. The extra L2 probably would've helped with the benchmarks, tho.

mczak, I agree that they got a little sloppy with their terminology, but I took care to use the term "desktop PC." Yes, you have to take into account the fact that the A64 has the memory controller built-in, while the P-M (and every other desktop PC) farms that task (and the associated power draw) out to the north bridge. I also thought they could've used more price parity, tho their argument was that the A64's extra power draw is offset by its lower price.

But you think the nF4 could draw a significant amount more than the P-M MB? Maybe the SLI version is less frugal than the non-SLI version.

Mariner, thanks for the link. Those are some crazy numbers that Ralf squeezed out of his Dothan. Man, Shuttle needs to get to work on a P-M SFF made for laptop HDs and that can accomodate a double-wide GPU.
 
unless the systems used to test the CPUs are 100% identical then the test results are meaningless. One CPU contains 2MB cache and the other contains a dual challen memory controller.
 
YeuEmMaiMai said:
unless the systems used to test the CPUs are 100% identical then the test results are meaningless. One CPU contains 2MB cache and the other contains a dual challen memory controller.
Click to expand...

I agree that the components like video and sound should be the same, but these are different CPU architectures with different strengths and weaknesses. Just cuz one has more cache doesn't mean it's not comparable in a benchmark situation. And, Dual Channel is an advantage of A64 and as such should be included when judging it against a competitor. Though it would be interesting to see how much of an advantage that dual channel gives, that can be seen elsewhere in other reviews of A64 itself.
 
You can only get so identical when we're talking two competing OEMs like Intel and AMD. I suppose you could eliminate the dual channel variable by using a single DIMM, but that still leaves the gaping price difference. So you can't be perfect with this sort of competition.

It's a nice start, tho, and I'd love for GamePC to expand upon it with more processors and perhaps different A64 MBs.
 
Pete said:
But you think the nF4 could draw a significant amount more than the P-M MB? Maybe the SLI version is less frugal than the non-SLI version.
Click to expand...
Well, I don't know WHAT draws more power - I simply know that at least 35W of the difference can't be from the cpu (when you take psu efficiency into account, it's still a 25W or so difference which is somewhere burnt on the board).
maybe the PCI-Express graphic card uses slightly higher voltage? The northbridge certainly will use some more power (not 25W more though, more like 5 probably). Maybe the rest of the 25W are spread out to several chips, since the i855 boards, while being desktop boards, might have a lot in common with notebook designs optimised for low power.
Whatever it is, the comparison might be valid for comparing power usage of the whole systems as you would buy them, but to say all the power consumption difference is because of the cpu is just silly. I don't doubt that some of the difference is due to the cpu, but how much you can just as well roll a dice, would be at least as accurate.
 
Actually, I wonder if the GamePC tests had Cool 'n' Quiet activated when testing the idle loads? They insinuate they do but perhaps it wasn't activated and this is the reason for the disparity? :?
 
Mariner said:
Actually, I wonder if the GamePC tests had Cool 'n' Quiet activated when testing the idle loads? They insinuate they do but perhaps it wasn't activated and this is the reason for the disparity? :?
Click to expand...
It wouldn't affect the full load power consumption. And the DFI Pentium-M motherboard doesn't support Enhanced SpeedStep either.
 
Can't the 90nm athlon 64s be greatly undervolted at stock speeds, so much so that they can reach pentium m levels?
 
Sxotty said:
Actually they can be undervolted, but not enough to be as low as the pentium Ms
Click to expand...

There was some article on tom's hardware that had them at a power draw of 35w(at 2.2ghz) under load, but I don't know if it was undervolted or not. I believe the 1.8ghz was at 30w, which is about the same as the pentium m's with the 533mhz fsb's peak isn't it?
 
The 2GHz Dothan uses about 16W at full load. The TDP of the 533MHz FSB is 27W, but it probably won't be reached till 2.4GHz at least.
 
Current A64 chips can be undervolted pretty well (reports of 2.0GHz at just 1.1V in some cases). Still not quite to the power consumption level of Dothan, however.

I expect 'Turion' to be little more than the new revision 'E' of Mobile Athlon64 (which is codenamed 'Lancaster', I believe) which incorporates strained silicon. This should reduce heat dissipation and enable lower voltages to bring it within reach of Dothan. AMD are quoting 25W for the slower Lancaster chips (up to 3200+, I think) and 35W for the faster ones. Lancaster also has twice the cache of Winchester which ought to make per-clock performance improve.

I notice that the new Opteron 252 is a Revision 'E' chip but unfortunately, Anand's review (the only one I've seen thus far) doesn't look at power consumption. :(

Personally, I'm waiting for 'Venice', the desktop Revision 'E' chips before I upgrade to a new system but I doubt they will be around for a few months yet.
 
You must log in or register to reply here.
Back
Top