• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

forcing an R9 290 to a clock speed...

Associate
Joined
6 May 2011
Posts
1,467
Location
Milton Keynes
Collective GPU minds of OCUK:

I am having some more black screen issues with my R9 290.

Basically I bought my 290 just before Xmas to replace my 7950, now when I first got it I had some major black screen issues when playing low intensity games, mainly DOTA2 (although anything which maxed the card never crashed).

Luckily the 14.1 beta drivers seemed to fix whatever was causing this issue and I have had trouble-free gaming from Jan-April. Now roll on a few days ago I have been having massive black screen problems, only in DOTA 2, mostly on the main menu, its the usual black screen hard lockup, although sometimes it will recover with the usual display driver stopped responding rubbish. If it does recover however, the memory will be locked at 150MHz and needs a re-boot to allow it to clock back to 1250MHz.

Now when I run MSI afterburner in the background whilst playing DOTA, the memory will sit at 1250MHz happily, but the GPU clock speed will fluctuate continuously between 400 and 600 MHz, when it should be locked at 975MHz in 3d mode. The GPU usage is also all over the place depending on how high the card is clocking itself.

I believe that this continuous change in GPU clock causes the display driver to stop responding, probably just because it gets confused eventually.

I have seen a similar problem in X-com, where having v-sync on didn't put enough load on the GPU to warrant the full 3d clocks, making it sit between 400 and 600MHz meaning it blackscreened every 10 minutes. I have a 144Hz monitor so I just turned V-sync off and the GPU usage went up to 100% at 975MHz and I managed to have a 3 hour session with no crashes.

It must be this changing core speed that causes the crash because I can run BF4 and tomb raider for hours, which will happily sit at 975MHz core speed 100%GPU usage and never black screen. (I have a custom afterburner profile for fan speed which keeps the temps below 82c)

I have tried every driver between 14.1 and 14.4 beta, none of which seem to fix this. Although I have no idea why it has suddenly gone bad again after 4 months of no-issues, although I suspect it's a random windows update or similar. Even turning the V-sync off in DOTA just dosn't generate enough GPU load to force the card to full clocks.

Now my question: Is there way like a BIOS flash or something that I can do which will force my 290 to sit at full clocks rather than this odd intermediate state between 2d and 3d mode? I'm 100% sure this is the cause of my crashes but I can't find a way to fix it in this particular case.

It really is a great card 99% of the time, but it just seems so daft I can run BF4/NS2/tomb raider for hours on end with no crashing, and yet the low requirement DOTA2 crashes on the menu after 3 minutes making me curse it and wish I had gone with the geforce 780!
 
Not sure how much help this is. Anyway when I first bought a 7950 couple years back I would also get black screen. It turned out to be the PSU power cable, with the Corsair PSU being Modular I changed the cable for another and didn't have the black screen since.

Might be worth a try if you can to change the cable?

I may try swapping the PCI-E cables for different ones when I get home, my HX850 has about 4 unused tucked in the back chamber of my case.

I'm not sure it's a card fault though, as it really does seem to be as a direct result of the core speed just going all over the place, which I presume is due to the drivers not being able to make their mind up. I'll have to take an afterburner screenshot when I get home just to show how odd the behavior is in DOTA 2 compared to literally anything else.

Still strikes me as odd that I can run really intensive games for hours with no crashes ever but as soon as I sit on the DOTA menu for more than 3 minutes it bombs out.
 
Ah brilliant, thanks LtMatt, this seems like it's what I need. Would you suggest a particular driver version or should I stick with 14.4 for this?

Also DOTA2 is like the least demanding game ever, I think the minimum requirements are Radeon 2600, which is why the GPU usage is so low and the core speeds get confused.

Either way I'll give it a try when I get home and report back.
 
I didn't get chance to have a proper test last night but it certainly seems to have locked my clocks to the right speed and I was able to leave DOTA on the main menu for a good 15 minutes without a crash, which is probably long enough.

Will report back once I have had a proper try over the weekend.
 
well that seems to have fixed my issues*, managed to play dota 2 for a full 90 minutes last night with no crashes, even after a long stint on the menu waiting for the dam matchmaking.

*touchwood
 
Well after I thought this was all fixed and my card was working fine I have had to restart my PC twice tonight in the space of a 40 minutes trying to play natural selection 2, just a black screen with no audio needing a hard reset.

I'm all out of ideas, I would love to send my card back for RMA, but what is the chance of OCUK actually finding anything wrong with it when it is so intermittent for me. Even then I would rather exchange it for a geforce 780 of some kind just to avoid this kind of rubbish happening again :/

Last AMD card I buy I think, anybody thinking of bothering with a 290 just save yourself the hassle and buy a 780.
 
I have now swapped my PCI-e cables for a different pair, will see how this gets on.

My afterburner settings are all just stock clocks with +50 power limit. I have also set it to "force constant voltage" in the compatibility properties.

Probably won't get chance to do any gaming till Monday so I'll have to report back then.
 
Right, after much experimentation I have managed to have several longer gaming sessions with no crashes. It basically came down to just setting profiles for different games in radeon pro (I have also changed my PCI-E cables around as well).

In low intensity games where I was having issues, DOTA2, X-com etc I am actually running an underclocked GPU profil, like 900MHz core and 1200 Memory. This is perfect as the games still run well in excess of 60FPS, and the temperatures are reduced significantly along with noise.

In higher intensity FPS games such as BF4/NS2 I am running my cards default clocks at 975MHz 1250Mem.

As long as I launch through radeon pro the clockspeeds get locked an the powerplay seems to work properly and I haven't had any issues.
 
In less demanding games, i.e. DOTA/Xcom If I don't use Radeon Pro and just leave say afterburner or GPU-z running, then it looks as if the clock speeds fluctuate all over the place from 400 - 600MHz (ish) this usually results in a crash of some sort pretty quickly. Either a black screen or a display driver stopped responding error.

Higher intensity games which can keep the GPU at 100% usage (tomb raider/BF4) I don't need to launch through radeon pro as the clock just goes to 975 and stays there according to afterburner. Never had a crash in either of these games

Natural selection 2 is an odd game, it's pretty shoddily optimized as it is and I have seen it crash a few times for no apparent reason, although launching it through Radeon Pro seems to fix this, which i guess comes down to the power limit actually working.
 
Just to provide a comparison for anybody who is curious, this picture shows what my GPU is up to on the DOTA menu without running through radeon Pro and when launched through radeon pro. The same pattern of clock fluctuations occur in game, I guess it just can't make it's mind up what speed it needs to run at.

3TMSSdR.png

Notice how the clock is all over the place, as well as the memory, when left to it's own devices without radeon pro keeping it in check...

No wonder the dam thing was crashing the display drivers all the time.
 
Last edited:
Back
Top Bottom