Would you return this?

[pandem0nium]

Hi Chaps,

I think my second MSI 390x might be a dud and can't sustain it's stock clocks.

I'm using OCCT gpu to test it which I believe uses furmark.

The settings I'm using are:

Shader level 4, full screen, 4k resolution and error checking enabled. FPS limit is disabled.

I get errors within the first 10 minutes. on stock settings except power limit + 20. If it's set at zero, then powerplay kicks in and it clocks itself down.

Would you return it or do you think I'm being too picky?

Has anyone got another method of testing that they'd recommend?

Cheers

 

[pandem0nium]

Typically if it can't sustain clocks then it's because the card is running to hot. How about testing it on it's own instead of in crossfire? I would remove the other card just in case to eliminate power/temps/airflow issues.

You might also want to read this - http://forums.overclockers.co.uk/showthread.php?t=18710139

It's installed on its own at the moment. It's in my second machine since my main machine motherboard needs to be rma'd. Which unfortunately doesn't have the slot spacing to run two 2.5 slot cards together.

Yeah, I've seen that and posted in there. I've repasted with thermal grizzly kryonaught. What I'll do tonight is check the application and maybe do it again using the line method.

 

[pandem0nium]

What are your frtc settings in catalyst, and are you using vysnc/freesync?.
What clock rates are you seeing in games and what settings are you using for cf in afterburner.

I've turned off frtc for the test, it's usually on a 60fps limit.

Freesync is enabled, but will give it another go tonight with it disabled.

Clocks are sustained during gaming, I'll have a go with a few games tonight and do some more testing.

 

[pandem0nium]

I'd also test in other benchmarks/situations. Furmark is known to be particularly stressful and damaging so GPU makers have put in failsafes to prevent this.

Yeah, furmark is pretty harsh, but it's the only test I know which is similar to cpu stress testing and reports errors.

 

[pandem0nium]

I've used valley and heaven for testing and still do. What I have found though is it could be perfectly stable in those only to fall over in games.

 

[pandem0nium]

I've repasted the gpu using the line method and it's helped a bit with temps.

I've retested, comparing against my known 'good' card, lets call it card A.

10 mins of OCCT with the same settings as the first post. stock clocks but power limit set to +50.

Card A - 0 errors in 10 mins, max temp was 81*C

Card B - 2 errors in 10 mins, max temp of 85.

So I'm fairly sure the card is a bit dodgy. The temps are quite high, but well below the 95*C maximum. I'll run Valley now for a while.

Thanks for the help guys.

 

[pandem0nium]

Shadows of Mordor has stopped working...

Temp was only 73*C when it failed. I think it's RMA time.

 

[pandem0nium]

Yeah, it could be. Try testing with occt with each card and see if you get errors.

 

[pandem0nium]

OCUK advised me to try running some benchmarks. I have a feeling that proving my 95% stable card isn't right is going to be a pain

Oh wells, if I can find a bench which finds a fault, at least I can use that as a good test of stability going forwards.