Question specifically for my GPU crunching Seti Guys

[TomWoolford]

Ok, so I know this could probably go in the GPU section of OC Forums, but I wanted to throw this out here since it's directly related to seti.
I've been told that raising the Shader clock on my GTX 260 helps with crunch times, but don't overdo it.
Well I've noticed even 40mhz helps in regards to about 45 seconds per 10 minute w/u which is worth it to me, but my problem is occasionally I start getting computational errors on cuda w/u's with the overclock. Is this something I'm going to have to live with, or should I just leave it at stock speeds?

 

[Sir-Epix]

Ok, so I know this could probably go in the GPU section of OC Forums, but I wanted to throw this out here since it's directly related to seti.
I've been told that raising the Shader clock on my GTX 260 helps with crunch times, but don't overdo it.
Well I've noticed even 40mhz helps in regards to about 45 seconds per 10 minute w/u which is worth it to me, but my problem is occasionally I start getting computational errors on cuda w/u's with the overclock. Is this something I'm going to have to live with, or should I just leave it at stock speeds?

Computational errors can come from anything and not just the overclock. If you are getting a healthy dose of them than I would suggust lowering your overclock. I also believe the VLAR killer will show up as a computational error, and sometimes you just get a bad batch of cuda work. What is your GPU at? I run mine at 702/1512/1053 and I sometimes get computational errors but most of the time it crunchs right along.

 

[TomWoolford]

After another batch of alot of errors, I lowered it back to stock. It only happens on overclock, and it was a mild one at best. 675/1050/1455

 

[Sir-Epix]

After another batch of alot of errors, I lowered it back to stock. It only happens on overclock, and it was a mild one at best. 675/1050/1455

When you overclock what are your temps at? Also do you just overclock the shader or to you overclock the core and the memory also?

 

[TomWoolford]

Temps are fine since it's aftermarket cooling, but no I don't play with core or memory clocks, but after running for a few hours I'm still getting the errors so I'm just going to go with the bad WU's and crank the clocks back up.

 

[Careface]

Hmm. I cannot be certain as the SETI site is down for maintenance at the moment, so I can't check your WUs, but you say you're getting a lot of computation errors - after how long do these WU error out? after 1-2 seconds, or a few minutes into the WU?

If its a huge bunch (i.e ~10-20 WU in a row) of 1-2 second computation errors, then I can almost guarantee you its the VLAR killer doing its job. If however, you're seeing errors several minutes into a WU, then there's most likely a problem with something else (drivers/hardware).

So, first thing to check - if the WU are erroring after 1-2 seconds, check on the SETI site (from memory) under account -> computers -> click the computer in question -> tasks -> click the link up the top which sorts WU into "errors" -> click TASK ID (not Workunit ID) -> check down the bottom to the stdout and in there will be a line saying something like "WARNING: VLAR Workunit detected! Aborting!" or something similar.

If the WU are taking several minutes to error out, then again, check the process above, but look in the stdout for a line saying something like "unhandled C++ exception in XXX" - this is a sign of an unstable oc, or driver issue. If this is the case, then.. well, lets cross that bridge if we come to it

On a side note, I see that if your cards clocks are indeed 675/1455/1050 (C/S/R), then your core and shader clocks are linked - are you 100% that you're only upping the shader clocks, and not the core clocks? Whats your card's stock clocks?

Hopefully we can get this sorted! crunch on!