PDA

View Full Version : Real world power consumption measurements from triple 7970's



Stu @ MSD
03-23-2012, 10:23 AM
Hi,
Now my water cooled rig is up running and stable, I thought I would offer some info I have recorded thats kind of interesting.

During benching and gaming, I have been creating a spreadsheet of everything I do and that includes power draw at the mains outlet. I was very surprised to find that most games dont stress the PSU anything like benchmarking. I thought the figures would be much closer.

The Test Rig: (For power consumption reference)

ASUS Maximus IV Extreme-Z
i7 2600K @ Std settings
3x AMD 7970HD in Tri-Fire @ std settings
2x 240GB OCZ SSD's
3x 1TB Velociraptors
Corsair Vengeance 8GB DDR3 @ 1866
XFX 1250W PSU
1260 & 360 Radiators fed with D5 pump
A few lovely lights... lol



Results as follows:
(All settings maxed for each game, using eyefinity at 5760x1200)


Benchmarking:

3D Mark 11 (Performance) – 880w
Heaven 2.5 – 898w
3D Mark Vantage (Performance) – 906w
Furmark on 3 screens – 908w
Aida 64 stability test - 1092w


Gaming:

Mass Effect 3 - 565w
Shift 2 unleashed – 570w
Syndicate – 580w
Skyrim – 620w
Hawx 2 – 641w
Alan wake – 720w
Crysis 2 – 780w


There is little point to this topic, other than to show that during gaming, your PSU may not be quite as stressed as you think. Its during benchmarking that you need the extra watts for synthetic loads.

Please Note:
These figures were taken with everything at standard clocks so its a valid benchmark. Overclocking will increase the current required.

HiVizMan
03-23-2012, 10:29 AM
Stu I think this is a very interesting thread mate, and one I have now subscribed too. Do you have your figures for when you did over clock.

I would be keen to see the numbers of just having the CPU over-clocked, and then the VGA over-clocked in say 50MHz increments. You up for the task mate?

Stu @ MSD
03-23-2012, 10:55 AM
I would be keen to see the numbers of just having the CPU over-clocked, and then the VGA over-clocked in say 50MHz increments. You up for the task mate?

I dont see why not mate, just need to decide on a definate test program to start with.
I would suggest Aida64 stability as it loads them up real fast, can have stable readings within 30 seconds. God knows what it does!

HiVizMan
03-23-2012, 11:13 AM
Yeah that sounds good. I would suggest the two graphic tests from Vantage myself as they are the single most difficult benchmark graphic test we have to run.

Stu @ MSD
03-23-2012, 11:46 AM
Yeah that sounds good. I would suggest the two graphic tests from Vantage myself as they are the single most difficult benchmark graphic test we have to run.

Yeah, I know what your saying, but the consumption is all over the place with those and no two tests seem quite the same, consumption wise, which is why I was thinking Aida.

HiVizMan
03-23-2012, 12:07 PM
Fair point - that does beg the question as to why? On my scope I get pretty similar peaks and troughs throughout the benchmark with the 580GTX cards. Have not hooked up my HD7970,s yet.

Stu @ MSD
03-23-2012, 01:24 PM
Fair point - that does beg the question as to why? On my scope I get pretty similar peaks and troughs throughout the benchmark with the 580GTX cards. Have not hooked up my HD7970,s yet.

Well, it was a long nights testing so will look at it again this weekend with fresh eyes and open mind, see what I can figure out.
Got some sexy samsung 30nm sticks to go in just landed on my desk, reading goog things about overcliocking them up to 2400mhz. :)