Dedicated rig High end GPU Low end CPU?

[JonasTheMovie]

I am a bit curious on getting a Athlon x4 870k (socket FM2+) and see if I get any improvement, considering the sanity checks that seems take time, but never utilize full cpu %.

[bruce]

I monitored the sanity check recently and it only took a second or two -- although I'm sure it varies depending on the atom count and I don't remember which protein it was. I suppose it would have taken longer if all CPUs were in use folding another protein but one CPU per GPU is probably sufficient.

[JonasTheMovie]

Maybe it is not the sanity check I see then. I see a paus on GPU and spike in CPU load every 20 minutes, like every 3-4 frames. What could it be?

[toTOW]

It is the sanity check (and checkpoint write). Interval (1%, 2.5%, 5%, ...) varies from project to project, since it's a choice from the project owner.

[bruce]

I see a paus on GPU and spike in CPU load every 20 minutes, like every 3-4 frames. What could it be?

That's what you should see, but what's the duration of the CPU spike?

[JonasTheMovie]

It varies depending on WU, but 30-60s load on CPU. Load on GPU usually stays up at 100% but sometimes it goes down during ~30s. Will need to measure it more precise.

[JonasTheMovie]

My rig has locked up a three times times the last days now. Not sure what it could be really, when starting the display its all black. The beta drivers have worked flawless for a week before that.
Started logging of sensor data to see if there is something wrong with the cooling. My best guess right now is gupdate for chrome, which always seems to be the last event found before a crash.

[muziqaz]

I would advise against using chrome or watching movies while GPU is folding. AMD has this little power state bug which came back from dead with Fury X. It was present with 290x, but it was fixed with omega drivers I think. AMD is aware of that issue. I am guessing yours is same problem. I would suggest reporting it to AMD feedback page for good measure

[JonasTheMovie]

I uninstalled Chrome to see if the problem goes away, since the rig is dedicated I have no use of it. Nice to see the next WU show predicted ppd of 500k, by chance.
Wonder why the problem started this saturday, no updates have been done lately.

[bruce]

If you've tweaked ANY of the driver settings, be sure that you have a temperature limit set with a higher priority than any of the other settings.

[foldy]

Maybe the ClockBlocker Tool may help your Fury X.
http://forums.guru3d.com/showthread.php?t=404465
https://www.reddit.com/r/Folding/commen ... ixed_with/

[JonasTheMovie]

The use of Clockblocker seems to be removed with the latest beta drivers. Clock stays at 1050Mhz.

[bruce]

Not-overclocked and stable is always better than a little faster and unstable. As I said, you need to make sure that the temperature limit has priority over all other settings. At times, FAH is efficient enough to push the temperature a little higher than what the rest of the settings call for.

[JonasTheMovie]

All seems well after removing Chrome, so gupdate seems to have been the culprit. Kind of sucky.

I keep stock clocks, the temp limit at 70C, fan limit at 50% and power limit at 10%+. I wish I could change priorities, but I can´t see how in the standard GUI.
Temp stays between 58-64C while the fan keeps steady at 15%/1000rpm except for some small rev up after sanity check, during some WUs.

[Nathan_P]

Jonas

Thanks for all the info - I'm now pondering something similar in terms of cpu power, i'll post a thread when I get started.