SPCR

Is Furmark a good load generator? AFAIK both AMD and nV drivers detect Furmark and throttle the card.
And SPCR is using it...

Other than the standard throttling that would happen when reaching the TDP im not aware of application detection being built into graphics drivers, would you have a link to such info please ?

http://www.anandtech.com/show/4008/nvid ... gtx-580/17

As for FurMark, due to NVIDIA’s power throttling we’ve had to get a bit creative. FurMark is throttled to the point where the GTX 580 registers 360W, thanks to a roughly 40% reduction in performance under FurMark.

Yea looking into it some more it does seem they did throttle the GPU based on what is running and the power draw...

http://nvidia.custhelp.com/app/answers/detail/a_id/2857

Im still looking into if this is the same with newer GPU's, maybe someone can enlighten us

EDIT: it would seem even newer card do the same, but it seems EVGA makes something similar to Furmark http://www.evga.com/ocscanner/ also couldn't you just rename the Furmark .exe to something else to avoid the driver from detecting it ?

Not that i would want to test that But it does raise the question that you initially raised about why is SPCR still using Furmark ? or maybe it doesn't matter for testing the thermal load or acoustics of a GPU's cooling solution.

Maybe the performance of Furmark should be included to such that throttling is noticed and the performance / W ratio can be calculated. Power by itself is kinda meaningless.

Olaf van der Spek wrote:Maybe the performance of Furmark should be included to such that throttling is noticed and the performance / W ratio can be calculated. Power by itself is kinda meaningless.

Furmark already reads the temperature for cards so it would also sound sensible to read the frequencies, plot them on screen and then you could see when throttling was taking place. It you look at RivaTuner it has real time frequency display for both NVIDIA and ATI.

From what i have read it seems if the drivers detects Furmark or OCCT running it wont do anything until the power regulator comes under load, as soon as that happens it will underclock the card by %50 and regulate the power so you cant fry your card.
Also from what i gather you can disable the card from doing this but you would more than likely end up with a toasted card

I think that its a safe guard as in normal use the power regulator circuitry will sometimes over power the GPU, but im guessing this is for a short time where as these programs can overpower the GPU very quickly and for a long time.
So its almost like they are lowering the limit that the power circuitry regulates the power at IF the driver detects one of these programs running.

Furmark is a waste in my opinion. Its just asking you to fry your card. I've seen water cooled GPUs have trouble keeping temps down when running Furmark.

If you're looking to test your GPU for stability before or after an OC, then get 3DMark(http://www.futuremark.com) or Unigine (http://unigine.com/) and run those. Use GPU-Z while running either of the previously mentioned programs and you'll have all the information you need.

Spoon Boy wrote:From what i have read it seems if the drivers detects Furmark or OCCT running it wont do anything until the power regulator comes under load, as soon as that happens it will underclock the card by %50 and regulate the power so you cant fry your card.

Why would it have to detect Furmark before it'd do that? Sounds like a good idea to do anyway.

I undervolted my 7870. Got it down to 1.03 V with Furmark. Passed Furmark many times but then noticed driver crashes in game. Ran Windows index experience and the driver would crash as well (at the same point) . Upped it to 1.07V. No issues after that. Conclusion. Furmark is a very poor test for system stability. It might be ok to test max temperature readings but beyond that I'd use other tests.

Immortals wrote:Conclusion. Furmark is a very poor test for system stability. It might be ok to test max temperature readings but beyond that I'd use other tests.

This will vary from card to card. On your Radeon this happens to be the case but on other cards it may be different. Remember that Furmark is OpenGL whereas the Windows test and games you may have run would have been DirectX.

I'm not %100 sure but i think this started to be the case from Nvidia 500 series and AMD 6000 series onwards, its really just a safety feature AFAIK.

EVGA's PrecisionX stability test is popular on a lot of overclocking forums. Catches errors in compute that you don't see in Furmark.

See http://www.anandtech.com/show/6774/nvid ... nveiled/15 for example

http://images.anandtech.com/graphs/graph6774/53412.png
http://images.anandtech.com/graphs/graph6774/53413.png

In BF3, the 7970GE is much quieter than the Titan.
In Furmark, they're close.

The Unigine 'Heaven' bench seems more useful for simulating a gaming load than Furmark.

http://www.techpowerup.com/downloads/21 ... k_3.0.html

The Heaven benchmark has recently been updated to version 4.0 and can be downloaded direct from Unigine. The newer version shows additional information on screen including the GPU temperature.

gdr wrote:The Unigine 'Heaven' bench seems more useful for simulating a gaming load than Furmark.

http://www.techpowerup.com/downloads/21 ... k_3.0.html

Correct, though the best way to test a new OC is still through a gaming session.

Cod wrote:Correct, though the best way to test a new OC is still through a gaming session.

Exactly. Furmark is not even 'worst case' or such.

My GTX 670 with ghetto mod Accelero S1 (had to bend the rear part to fit it in the case) and a Scythe Gentle Typhoon does not pass Furmark. It always reaches >95°C, even with the fan at 1250 rpm, at which point I stop the benchmark. Playing Crysis 3 however, 2560x1440, high details, 30-odd FPS, I max out at 83°C with the fan at 800rpm.

I have no idea what Furmark is supposed to do. If you're a silent freak, it just makes you waste effort for a never-occuring scenario.