TITAN Downclocking

[Sarge104]

Little back ground in whats been going on. Was doing a bunker for an upcoming challenge when I noticed my TITAN card was down clocking even under load. I've uninstalled all drivers, reinstalled the updated drivers, fiddled with PrecisionX, uninstalled it, and now have been experimenting with Nvidia Inspector to see if I can get the card to run in something other then a low power state.

I've been unsuccessful.... (http://ic.pics.livejournal.com/sarge104/1100638/156497/156497_original.jpg)It's maddening.:mad:

Just to be clear, I've kept watch on temps and they've never broken 80C. I'm at a lost as to why I'm having such an issue keeping my card clocked under load.

[John P. Myers]

I recall something in precision X that allowed you to lock the freq. Only downside is you couldn't overclock it afterwards and in some cases the locked speed may have even been lower than your boost speed, but not by much. I don't know if this feature is still available but it's worth a check. I don't remember what it's called but i do remember the freq it would lock at was based on the voltage.

[MindCrime]

I can't remember the circumstances but I've seen nvidia cards not come out of a lower powerstate. I can't recall if gpuz shows the pstates but I know gpushark does. Most nvidia gpus come with 3 pstates that I've seen lately and sometimes boinc doesn't wake them up. I can't remember what would reset it besides a power cycle.


Edit: Yep I can see it in your pic P2 in pstate. I don't use any gpu management software anymore I do however use Rainmeter with system info on my main rig the one that would have problems. I can see the clock and usage of the main bits. Anyways I now just keep the driver up to date and don't fiddle with gpu clocks anymore since they're already so over clocked, have boost, use a ton of power and some projects recommend against it so I said it's just not worth the trouble of fiddling with them anymore.

[Sarge104]

Edit: Yep I can see it in your pic P2 in pstate.


Man, when you get frustrated you just forget to check basic things. Thanks for pointing that out. I'm still confused with how it goes to a balanced power state if its over 70% in use.

[Sarge104]

Currently running the card in the P2 state until I finish the last of the WU's to see if it down-clocks again.

[Sarge104]

Was able to check on the GPU during lunch...yep, down clocked again to P5 now. OMGWTFBBQ! The card is fully loaded, will open the case tonight and see if it is a possible hardware issue. Unfortunately I've only seen this issue after I've updated the software so I'm more likely to blame that then the card itself.

If it would be hardware I thought I would be seeing issues with voltage or the card itself not running. It is strange.

[Sarge104]

Brought the overclock down to 1137Mhz. I'm testing tonight to see if the P-state stays at P0 and the overclock stays. I'm still running into the issue where if the temp goes over 75C the card starts to down-clock. I may have to try switching back to PrecisionX to try and use the precision fan control to keep the fan speed down to allow more power to go to the card vice having the fan at 85% all the time.

[Sarge104]

Been a couple of hours since I re-installed the PrecisionX software, temps have stayed down with aggressive fan management and I've held my over clock with no down clocking that I can see. Going to hit the hay and see what the morning brings.

[Sarge104]

This morning I woke up to no down clocks, should equal less stress right? Of course I tried to increase the over clock by a couple Mhz with PrecisionX but the modifications would not take. I'm assuming its from both Invidia Inspector and PrecisionX fighting even after I shut down Inspector.

Aggressive fan curve has kept temps in the low 60C range, as stated before I've noticed the card start to down clock in the 75C range which in itself is strange to me when 80C should be a "safe" temp.

Probably going to take a few more restarts and dinking around to see what new issues need resolved. Been promising a tower cleaning for a while and it's always "one more WU...one more fiddle with overclock..." Heheheh.

[Al]

I used to count on my dust bunny's holding everything together. =))=))

[MindCrime]

Is this your only GPU in the machine? Do you have one on the processor as well?

[Sarge104]

Is this your only GPU in the machine? Do you have one on the processor as well?

Neg, no gpu on the processor AMD-8350. I probably just need to do a power down and while it's down give it some anti-dust bunny love. After going through so much hassle I'm of the opinion that a recent software update on the PrecisionX caused my voltages to be reset, and the fan curve to be reset. I'm just going to have to keep bumping both the overclock and voltages bit by bit to get it set back the way it was. Was of the opinion that I could just ramp up everything to where "I think it should be" and, surprise, things don't magically fix themselves.

Short story: Grogg sees error, Grogg beats error with club, Grogg gets mad when error laughs at him...

1998

[Sarge104]

After work I took the tower to the garage to shoot it, I mean clean it out with the air compressor. The amount of debris I expelled from the case gave me a little concern but afterwards I had a bit more fun with bumping the clocks on the TITAN now that it is not choking on dust.

My suspicions were validated as I tried to bring the clocks over 1135. The new PrecisionX software has over-voltage capped at 37mV so currently I'm only able to bring the card up to 1162 overclock. Compared to the Nvidia Inspector I had the card at 1257.

More fiddling to come...

[MindCrime]

Hows the Titan? Is it dropping to lower power states when it shouldn't be, still?