Maxwell Power Consumption from Tom's Hardware

[3DVagabond]

Well I do see in some benches, there's a massive increase in SP at similar TDP to 680, so indeed it could be claimed its doubled efficiency too *for some tasks.

I think its definitely a mix of hardware & software, else we would get major efficiency improvements for Kepler via a driver update and NV isn't talking about that at all.

Even if possible, nVidia wouldn't update Kepler cards to make them more efficient. They want you to buy a new Maxwell card.

 

[Enigmoid]

Techreport.com confirms the 980 being below 970, in two separate measurements, with the 970 consuming more than all the other tested Nvidia cards, which is also curious, again I suspect 4 GB GDRAM being part of the problem. PCper.com also has a 2W difference favoring the 980. So even with their worthless, lazy method, they can pick up on the difference.

The 9.2 W and 15.8 W for the 980s vs. the 19.3 W for the 970, granted the first one appears to be an outlier. Still it's interesting because it's counterintuitive.

Techreport is comparing the reference 980 with a custom OC 970.

 

[Cerb]

The 9.2 W and 15.8 W for the 980s vs. the 19.3 W for the 970, granted the first one appears to be an outlier. Still it's interesting because it's counterintuitive.

Right. Tom's numbers are about double, while everyone else's are +0-20%. I think THG really needed to compare known cards on the new testing system, and work out some measuring and statistical kinks, before publishing a review with form conclusions like they did.

I suspect the difference in the cards is just binning, assuming Tom's speculated 970 is whack, and the stock OC GB card is a bit power hungry (since other reviews have different results with actual 970 cards). That, while some 970 (and 960?) GPUs are going to salvaged, many are likely to handle being fully enabled, but be a little on the hot side, so don't make the cut for the stated TDP fully enabled at full speeds.

 

[ocre]

Even if possible, nVidia wouldn't update Kepler cards to make them more efficient. They want you to buy a new Maxwell card.

Nvidia has already done all the possibly can with software, its great for controlling clocks and oversight. You can save power and boost up clocks pretty fast when the software allows it.

But the peaks toms is seeing in the oscilloscope is not software driven. Its much more instantaneous and exactly what you would expect in a system with ultra fine grained power gating.
You cannot achieve this with software updates, this is by design

 

[Abwx]

But the peaks toms is seeing in the oscilloscope is not software driven. Its much more instantaneous and exactly what you would expect in a system with ultra fine grained power gating.
You cannot achieve this with software updates, this is by design

For the time i think that there s inherently some hardware needed, the spikes can be explained by a lot of other things, there could be a part that is software driven but one more time the hardware must be designed accordingly, i m still doing some theories around the thing, since there s no official explanations we re left doing wild speculations about all the possibles implementations, some that are surely not even there...

 

[3DVagabond]

Nvidia has already done all the possibly can with software, its great for controlling clocks and oversight. You can save power and boost up clocks pretty fast when the software allows it.

But the peaks toms is seeing in the oscilloscope is not software driven. Its much more instantaneous and exactly what you would expect in a system with ultra fine grained power gating.
You cannot achieve this with software updates, this is by design

I wasn't commenting at all about Tom's measurements.

 

[Enigmoid]

Can someone with a 980/970 run a couple of compute benchmarks and give clockspeeds and power consumption to put this thread to rest?

 

[Abwx]

Can someone with a 980/970 run a couple of compute benchmarks and give clockspeeds and power consumption to put this thread to rest?

Look like power comsumption is not really a concern, after all....:biggrin:

 

[Enigmoid]

Look like power comsumption is not really a concern, after all....:biggrin:

What exactly is your point? Nothing like first hand tests.

 

[Abwx]

What exactly is your point? Nothing like first hand tests.

My point is that there s a thread full of 970/980 owners and that so far they do not seem to be concerned by power comsumptions numbers like you, otherwise they would had made some tests in this respect, but no, what matters are the graphic scores, i told you that absolute perfs is the first thing that interest gamers...

 

[Keysplayr]

So your entire campaign, judging by your mocking smiley face above, was to show that power consumption doesn't really matter? Just because they aren't talking about it non-stop in the 970/980 owners thread?
What on Earth for? They are just enjoying their GPUs. Figuring out options.
Enigmoid, as well as I, have asked what your point is. However I think we now finally know it.

 

[krumme]

Not true, AMD throttling is due to the chip trying to limit the temp but not power if temp is low enough, Nvidia has a power hard limitation that can be seen on Furmark score with a 970 getting the same score as a 280X.

This. Well done observations and points Abwx in this thread. Looking forward to more explanations from you regarding this matter as its interesting and prospective tech - btw i actually care about real efficiency in a desktop

 

[AtenRa]

Does anyone know what GPGPU application Tomshardware used in the power consumption measurements ?? Cant seam to find any info.

 

[Enigmoid]

Enigmoid, as well as I, have asked what your point is. However I think we now finally know it.

I'm still not sure what you are implying. I've already stated my point: Tom's isn't making correct measurements or is interpenetrating the data as their results differ greatly from the dozens of other review sites. Now their data could certainly be correct, however its likely that they are simply measuring an artifact of power delivery.

Fact 1: Toms is using different equipment and measuring equipment compared to previous results. Scientifically they have not demonstrated that the two methods are comparable and can be placed side by side. They have no instituted (or if they did then did not mention) any error check on system power in their results. They are not even giving the name of the stress application.

Fact 2: No other site has similar results (that I have seen).

I'm not being biased here. I certainly will say that Nvidia is overplaying efficiency as even gaming benchmarks indicate only 60-80% gain over the 680 at similar power usage not 2x as marketed. I don't think AMD is doomed or whatever, from the other thread.

Having a 970/980 owner run a luxmark/litecoin/furmark bench at stock clocks will quickly show whether the card is drawing 285W of power and allow a conclusion to be drawn with reasonable clarity. I don't see what anyone has against it.

Abwx seems to think somehow I'm being sneaky here by mentioning efficiency and power consumption. How about reading the thread title? What else should I talk about? And yes while gamers are concerned about performance efficiency is the primary factor in how well that performance can scale.

 

[96Firebird]

Does anyone know what GPGPU application Tomshardware used in the power consumption measurements ?? Cant seam to find any info.

I've looked, multiple times... nothing. Horrible way to do a review.

 

[Cerb]

My point is that there s a thread full of 970/980 owners and that so far they do not seem to be concerned by power comsumptions numbers like you, otherwise they would had made some tests in this respect, but no, what matters are the graphic scores, i told you that absolute perfs is the first thing that interest gamers...

Most of us have, at best, a Kill-a-Watt. But, we can all do performance tests that can be readily compared with others (and are useful to see what driver/CPU bottlenecks may exist that reviews, often using OCed i7s, might not show).

Also, that very thread is quite full of high-end SLI users, who, aside from, "can I cool it," of course are minimally interested in heat/power/noise. I'm less worried about minor power differences, myself, than I am the fact that actual power use directly translates to higher noise, with any given active cooling system. The light usage cases, like web browsing keeping it above idle, and old games that can't bring GPU usage to 100%, matter, to me, but I can't easily test those differences, myself.

As soon as I can figure out who's is really the quietest stock cooler, my money's getting spent (MSI looks like it, so far).

 

[wand3r3r]

So your entire campaign, judging by your mocking smiley face above, was to show that power consumption doesn't really matter? Just because they aren't talking about it non-stop in the 970/980 owners thread?
What on Earth for? They are just enjoying their GPUs. Figuring out options.
Enigmoid, as well as I, have asked what your point is. However I think we now finally know it.

Why not stick to the topic. It appears ad hominem without even discussing the data toms brought out or Abwx brings up. D:

He's discussing power consumption and what's going on with Toms measurements? To some of us we are actually interested in what it means, whether it's unusual or not. :biggrin:

The voltage fluctuation can probably be proven to be normal or not, and what it means can probably be proven as well. The facts should sort themselves out.

 

[Attic]

Someone do GPGPU and check clocks and power draw on the 970/980 vs Unigine Valley Maxxed.

Too bad crypto on gpu died, this would be more definitive already.

 

[sontin]

Tom's Hardware updated their article:
Gigabyte's custom cards have a much higher power limit - 250W for the GTX970 and 300W for the GTX980.
Their reference GTX980 had a problem. The new one doesnt really exceed the TDP limit of 165W - 177W in the torture test: http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-13.html

 

[geoxile]

Tom's Hardware updated their article:
Gigabyte's custom cards have a much higher power limit - 250W for the GTX970 and 300W for the GTX980.
Their reference GTX980 had a problem. The new one doesnt really exceed the TDP limit of 165W - 177W in the torture test: http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-13.html

It would be nice if we had some performance metrics to accompany those numbers. It would be strange if the gigabyte model used up to 2x the power just to achieve the same performance

 

[wand3r3r]

Interesting, I wonder if NVidia sent them a new card.

I also wonder if the gigabyte cards can actually be overclocked better if it's not being as throttled?

 

[mindbomb]

I feel this is how things should be. You have ~200 watts for a reference card, and you can have premium overclocked models for 250+ watts. Looking at previous cards, with hawaii, I think a lot of ppl felt that the premium trixx model was the only viable one, cause it was 300 watts to start with.

 

[Abwx]

Abwx seems to think somehow I'm being sneaky here by mentioning efficiency and power consumption. How about reading the thread title? What else should I talk about? And yes while gamers are concerned about performance efficiency is the primary factor in how well that performance can scale.

Perhaps i used the wrong smiley, it was not directed at you actualy, it was some irony about the fact that 970/980 owners seems interested firstly in their GPUs scores and not really about power usage since orherwise we would have numbers popping here and there.

 

[Enigmoid]

Perhaps i used the wrong smiley, it was not directed at you actualy, it was some irony about the fact that 970/980 owners seems interested firstly in their GPUs scores and not really about power usage since orherwise we would have numbers popping here and there.

For most people absolute performance rules the roost. I agree completely. However, focusing on efficiency allows for dominance in mobile (which Maxwell is doing) and allows the chips to be scalable. Its also terribly important for industry.

There does appear to be an upper limit. Release chart leading performance at >480 levels of power, noise, and heat and sales will be noticeably lower than if the card is cool and quite. I agree that it comes down to perf and price but given two options with similar prices and performance and I'll take the quieter and cooler one any day.

 

[selni]

It would be nice if we had some performance metrics to accompany those numbers. It would be strange if the gigabyte model used up to 2x the power just to achieve the same performance

This is exactly why power measurements without performance measurements are meaningless with respect to efficiency. The Tom's article doesn't even try to account for this this - and thus statements like "it's not more efficient at GPGPU" are essentially unsupported, but that doesn't appear to stop them. You can say "it uses a lot of power under said GPGPU test" or "absolute maximum possible power draw is similar to previous cards" but that's hardly the same thing.