GPU load is different in game and running benchmarks

Valdis Volks

Honorable
Jul 21, 2013
4
0
10,510
Hello there,
I've just upgraded my system with another Gigabyte HD 7950 Windforce 3, 3gb and have them in crossfire, I've spent days to sort things out cos it gave me lost of trouble.



Now it seems like I have managed to get it up and running properly, when I run Heaven benchmark with both cards performance is about 95% better
than single card ( makes sense ) and both GPUs are running at 99% - 100% (no fluctuations nothing, running perfectly smoothly - without increasing power)

Both GPUs are overclocked to core clock: 1175 Memory clock: 1400, I haven't touched the voltage.

Heaven 4.0 Benchamerk results (both cards) at everything maxed out, resolution 1920-1080
FPS: 63.30
Score: 1593
Min FPS: 20.20
Max FPS: 142.60



GPU load used to fluctuate a lot ( when running metro Last light)
GPU 1 used to fluctuate between 100% - 0% and
GPU 2 normally stayed at 30%
so I increased power by 20% and that solved the GPU load fluctuaton problem, but now there is different issue

If I run Metro Last Light with increased power at max settings my GPUs wont go beyond 60% but there is no flactuation
If I run Metro last light without power increase GPU 1 fluctuates between 100% - 0% and GPU 2 normally stays at 30%

But if I run Heaen benchmark with OR without increasing the power both GPU load will be 99%-100% at all times.

why is that ? can it be fixed ?
If GPUs would run at max as they are with heaven I would have bout 40 FPS instead of 24 FPS

Im runing everything in full screen
I have not got multiple displays

PSU is Corsair GS600
CPU is i5-3570K 3.4 GHz overclocked to 4.1GHz (4cores)
RAM 8Gb

Mother Board is ASUS PBZ77-V LX2

Kind regards Valdis
 
Solution
PSU is a good point, but be aware of how the GPU load works

obviously in benchmark the card is capable, so PSU may or may not be an issue. Benchmark would likely bare this if an outstanding problem. Your first card will ALWAYS run heavier than second. The second card is and extension via the crossfire, the PCI bridge treats it not as a separate card, but as an extension of resources of the first. Therefore, unless the resources are needed, the second card will not fire up any more than it needs to. What your seeing is a completely normal occurrence. Benchmarking is designed to stress test all resources, so of course the second card goes to max. Gaming does not do this... unless high end gaming like Crysis, BF3 etc. with...
This is what you should be seeing

http://www.techpowerup.com/reviews/AMD/HD_7950_CrossFire/14.html

Your PSU is undesized

http://www.amd.com/us/products/desktop/graphics/7000/7950/Pages/radeon-7950.aspx#2

500W (or greater) power supply with two 75W 6-pin PCI Express power connectors recommended8
750W (or greater) power supply with four 75W 6-pin PCI Express power connectors recommended for AMD CrossFire™ technology

http://www.guru3d.com/articles_pages/his_radeon_hd_7950_review,8.html

Here is Guru3D's power supply recommendation:

Radeon HD 7950 - On your average system the card requires you to have a 500 Watt power supply unit.
Radeon HD 7950 Crossfire - On your average system the cards require you to have a 700 Watt power supply unit as minimum.

If you are going to overclock GPU or processor, then we do recommend you purchase something with some more stamina.
 

Valdis Volks

Honorable
Jul 21, 2013
4
0
10,510
I seee your point, but than again.
why is it ok with the benchmark? and not ok with the actual game ?
if i was low on juice would not the benchmarking tool do the same thing as Metro ?

I just want to make sure that isnt any software issue before i get into purchasing new hardware


If i have to get new PSU.
what do you reckon would this suffice ?
http://www.ebay.co.uk/itm/ARTIC-BLUE-850W-850-WATT-QUAD-RAIL-ATX-PSU-8-PIN-SATA-PCI-E-QUIET-POWER-SUPPLY-/281138247857?pt=UK_Computing_PowerSupplies_EH&hash=item41752538b1
 

Buzz247

Honorable
Mar 18, 2013
962
1
11,360
PSU is a good point, but be aware of how the GPU load works

obviously in benchmark the card is capable, so PSU may or may not be an issue. Benchmark would likely bare this if an outstanding problem. Your first card will ALWAYS run heavier than second. The second card is and extension via the crossfire, the PCI bridge treats it not as a separate card, but as an extension of resources of the first. Therefore, unless the resources are needed, the second card will not fire up any more than it needs to. What your seeing is a completely normal occurrence. Benchmarking is designed to stress test all resources, so of course the second card goes to max. Gaming does not do this... unless high end gaming like Crysis, BF3 etc. with settings maxed out. Metro is not near these, so no point of comparison.

Bottom line - you're fine - it's working like it is supposed to.

That all said, PSU may support fine... for now. But you are putting a load on it that it is not designed to handle, and will over time shorten it's life and cause a capacitor blow out. I would look into a larger supply. 2 cards of that caliber would do well on no less than 650 - advise 700-750w. This leaves headroom for case fans, lighting etc, and will not overstress your PSU, all the while not breaking the bank either. If you like the lighting option of the GS series, look at Corsair GS800 - rated well, uses Japan capacitors, and has low ripple. Even Johnnyguru reviews like it! lol do not skimp on PSU quality - for something of this power be prepared to spend 85-95USD. This is the heart pump of your system! dont get chinsey here
 
Solution

Valdis Volks

Honorable
Jul 21, 2013
4
0
10,510


Just checked with Crysis 3 and good news. GPUs are running at 85 - 100%

shame that it's still just 19fps with everything on max :D veery demanding game :)

But with AA off FPS is 50 - 100