Brand new 290x memory clock stuck

[TookDk]

As you properly have realized you have a serious temperature problem.
I would recommend that you keep your average temperature under 80C (My target temp is 75-78C, and my cut-off temp is 85C).

You can properly get away with running the rig at 85C, but that will significantly reduce the lifetime of you graphics cards.
It is NOT recommended.

1-2 extra fans can normal do the job to bring down the temp.

[OllieB2000]

As you properly have realized you have a serious temperature problem.
I would recommend that you keep your average temperature under 80C (My target temp is 75-78C, and my cut-off temp is 85C).

You can properly get away with running the rig at 85C, but that will significantly reduce the lifetime of you graphics cards.
It is NOT recommended.

1-2 extra fans can normal do the job to bring down the temp.

Yeah that's a good point, I have adjusted some settings and get about 850 k/h @ 80c. It's not maxing the card out but then again I can at least mine 24/7 with reasonable temps.

[aigeezer]

A couple things:

You shouldn't need a dummy plug for 290x so don't bother with that.

The SDK software is of no use to you if you're using Catalyst 13.11 Beta or 13.12.

If you're eliminating all peripheral components(including other video cards) and risers, doing a power drain, wiping your drivers with either Display Driver Uninstaller or a sweeper and reinstalling and still having problems, you might have something corrupted in your OS or a MoBo issue.

Both of those issues would typically give BSODs.

If you're not getting any crashes, just weird hash rates, try a different cgminer version. Don't forget to eliminate cgwatcher for now and don't use GUIminer ever.  

Thanks for the tips, I shall give those things a try and see how I get on.

Update: I seem to be getting on ok now, not great but 800 k/h. I started with 1030mhz and intensity 13 getting me 500 k/h, I then lowered my gpu clock to 950 and upped the intensity to 20 getting me 800 k/h. Not sure what I can do next but I guess there's going to be a lot of messing around to get the kind of speeds that others are seeing with their 290Xs.

Thanks to everyone for their input, probably would have given up a while back otherwise.

I didn't notice this thread when it was fresh, but I was having similar problems both then and now with a 290X. As far as I can tell, the memory clock throttle-down to 150 occurs after a "driver has crashed and has recovered" message. From then on, cgminer and sgminer commands to set memory clock speeds are ignored. You can tune all the other arguments, intensity and so forth, and get more-or-less expected results, possibly not noticing that the memory clock has stayed locked at 150.

A reboot always seems to fix the problem until the next "recovery". Sometimes I may be tweaking settings and it is obvious that something has happened. Other times the mining program might become unstable while unattended, with no eyes to see the (faux) recovery message. As far as I can tell, the issue is present for all the recent AMD driver versions, including the latest 14.2 beta.

OllieB2000 - my interpretation of what is happening may not seem consistent with the breakthrough you reported back in January. I could just be wrong or perhaps your detuning prevented any further failures and their "recovery" messages, along with their memory clock locking. Do you have any new insights on the phenomenon? Anyone?

Thanks!

[m0nz]

try the same configuraiton without
“thread-concurrency” : “32765″, (just delete the whole line or set to "8192" instead)

If that do not work, then try to strip down the configuration file until it runs.
Should be able to work with only:


{
“pools” : [
{
"url" : "stratum+tcp://yourrpool1.com:3334",
"user" : "<USER>.<WORKER>",
"pass" : "<PASS>"
},
}
],
“scrypt” : true,
}

(If you don't specify a parameter then is the default value used)

This is completely on point. Thank you.  I have 290x's and when i switched the crashed gpu down to 8192 tc, it was immediately fixed.

Thanks.

[m0nz]

A couple things:

You shouldn't need a dummy plug for 290x so don't bother with that.

The SDK software is of no use to you if you're using Catalyst 13.11 Beta or 13.12.

If you're eliminating all peripheral components(including other video cards) and risers, doing a power drain, wiping your drivers with either Display Driver Uninstaller or a sweeper and reinstalling and still having problems, you might have something corrupted in your OS or a MoBo issue.

Both of those issues would typically give BSODs.

If you're not getting any crashes, just weird hash rates, try a different cgminer version. Don't forget to eliminate cgwatcher for now and don't use GUIminer ever.  

Thanks for the tips, I shall give those things a try and see how I get on.

Update: I seem to be getting on ok now, not great but 800 k/h. I started with 1030mhz and intensity 13 getting me 500 k/h, I then lowered my gpu clock to 950 and upped the intensity to 20 getting me 800 k/h. Not sure what I can do next but I guess there's going to be a lot of messing around to get the kind of speeds that others are seeing with their 290Xs.

Thanks to everyone for their input, probably would have given up a while back otherwise.

I didn't notice this thread when it was fresh, but I was having similar problems both then and now with a 290X. As far as I can tell, the memory clock throttle-down to 150 occurs after a "driver has crashed and has recovered" message. From then on, cgminer and sgminer commands to set memory clock speeds are ignored. You can tune all the other arguments, intensity and so forth, and get more-or-less expected results, possibly not noticing that the memory clock has stayed locked at 150.

A reboot always seems to fix the problem until the next "recovery". Sometimes I may be tweaking settings and it is obvious that something has happened. Other times the mining program might become unstable while unattended, with no eyes to see the (faux) recovery message. As far as I can tell, the issue is present for all the recent AMD driver versions, including the latest 14.2 beta.

OllieB2000 - my interpretation of what is happening may not seem consistent with the breakthrough you reported back in January. I could just be wrong or perhaps your detuning prevented any further failures and their "recovery" messages, along with their memory clock locking. Do you have any new insights on the phenomenon? Anyone?

Thanks!

In case I missed the point directed to you as well, the trick for me was to turn thread concurrency off or to 8192, change I=13 and run... you should get about half your normal - mine is ~480ish... then i put tc back to 32765 and I=20 and i'm cooking back at full speed. Weird why it works, but it does.  NOW, i haven't done it many times just today.  The other thing that worked for me once was to enable crossfire and then disable it... another thing worth the try.