My 6950 Power Draw Tests

[doveman]

So I borrowed this power meter thing, which doesn't appear to be particularly reliable or accurate, but it's probably close enough. Not sure these results will show anything that you all don't already know (except that I've got a funky meter!) but I thought I might as well share in case someone found them useful/interesting.

With just the HD3300 IGP and my X4 idling at [email protected], idle draw was around 84W. The OCCT tests resulted in the following:
CPU: 140W, increase 56W
GPU: 106W (fluctuating 73, 85, 91W), increase 22W
PSU: 145W, increase 61W

With the HD6950 (shaders unlocked, idle clocks 250/150 (single screen) - VDDC 5A) and the X4 idling at [email protected], idle draw was 102W.
CPU: 162W, increase 60W
GPU: 272W, increase 170W. This was wierd as it mostly hovered around 260-270W, but quite often dropped to 40-60W, which is clearly nonsense! GPU-Z showed a VDDC of about 87A, but I'm not sure how accurate that is. The Core clock never really hit the full 800Mhz either, and was constantly changing between about 513-675Mhz, with a few spikes into 700+ territory, so this might not be a great test of gaming power use, where I imagine it would constantly be at full clock. This test ramped the X4 into full power mode ([email protected]) and loaded it about 19%.

With the unmodded HD6950 idling at 450/1250 (running dual screens, VDDC 9A) and the X4 idling at [email protected], idle draw was 135W.
CPU: 200W, increase 65W (I'm sure this is wrong as the X4 idling at full power should mean the increase was less surely?)
GPU: 257W, increase 122W

So at idle, with the 6950 my system draws about 18W more than with the HD3300 IGP, and with a fully loaded CPU and mostly loaded GPU, the system draw would be about 330W.

This page http://www.techpowerup.com/articles/ove ... idcard/159 puts the unmodded 6950 at 169W with Furmark, which is 47W higher than I measured with OCCT. It shows that jumps to 252W when the 6950 is shader modded, clocked to 6970 speeds and PowerTune upped to 20%, but the FPS gains are substantial.

I'm only running a 400W PSU at the moment though, so I've probably got enough power to up the clocks to 6970 speeds, but not to put PowerTune at 20% as well, although without doing so the (Furmark) FPS gains are apparently precisely nil (for a 33W power increase!) so I might see more of an improvement by leaving the clocks alone and just upping PowerTune, but I don't know if this will translate from Furmark to real games.

[zyrobs]

The card speed hovering is caused by the Powerplay thing, that scales back the card in extreme applications, so the power usage doesn't get extreme. You can give it a plus 10% at the overclock panel of CCC, to make the clockspeeds stable.

I'd quite like to get readings with that options as well, if possible.

[doveman]

Thanks for the advice. I'll be happy to re-do the tests with Powertune at +10%, although it won't be for a few days as my TV just went on the blink so I've had to disconnect everything and re-arrange my room into a repair workshop!

[doveman]

OK, I've had a chance to do the test at Powertune +10% as requested.

As you quite rightly said, it does stabilize the clocks. Conditions were as before with the shaders unlocked, idle clocks 250/150 (single screen) - VDDC 5A) and the X4 idling at [email protected], idle draw was 102W.

OCCT GPU test: 290W, increase 188W. VDDC Current - 96A

I hope that's what you wanted to know. I'll probably leave Powertune at +10% as it should increase my gaming FPS a bit by preventing the core clock scaling down.

[zyrobs]

It is indeed. Thank you.

[doveman]

You're welcome. Glad to be able to help.

[Zargon]

very cool thanks for the info!