Forum > GPU crunching
Latest drivers (NVidia and ATI)
Urs Echternacht:
--- Quote from: Josef W. Segur on 29 Dec 2011, 05:46:44 pm ---I expect Urs was looking for control of the 2 seconds a GPU kernel can run before Windows thinks it has hung and forces a reinitialization. He's probably already found it, but the MSDN Timeout Detection and Recovery of GPUs through WDDM article defines the registry keys and how they interact. A web search for TdrDelay or TdrLevel turns up many pertinent discussions with advice.
Joe
--- End quote ---
Actually with Win7 i feel much like a newby, so didn't even know the correct terms to look for. In Linux and XP this is or was called "VPUrecover" and it does or did restart possibly hanging drivers. Actually i wanted the "old" behaviour back to extract the place of failure in the driver and derive a dump containing that info. This dump would help AMDs developers to look up where they need to improve their drivers.
With my current setup, which uses the default settings after a fresh install of Win7, the dumps do only show the failure occuring inside windows components and mention the actual failing AMD driver as a secondary point of concern. That would be of NO help.
So, Joe and k6xt, if WDDM (which i guess means something like "windows device driver model", have to look explanation up!) is what makes a GPU driver recover and restart on Win7 then that is what i'm looking for but did not find.
Thanks to both of you for trying to help. :)
k6xt:
--- Quote from: Mike on 29 Dec 2011, 10:26:28 am ---
And 560 TI if you have the skills to control it.
regards
Mike
--- End quote ---
Mike can you elaborate on "...skills to control it." ? What is it that needs controlling? Maybe there's a link to a discussion here? Today I just about decided on 560Ti or, if I can't "control it" a 460.
Jason G:
--- Quote from: k6xt on 29 Dec 2011, 07:00:05 pm ---Mike can you elaborate on "...skills to control it." ? What is it that needs controlling? Maybe there's a link to a discussion here? Today I just about decided on 560Ti or, if I can't "control it" a 460.
--- End quote ---
I'll inject what info I've got here. The 560ti issues have narrowed down to a small set of caveats, after extensive investigation.
- Some card manufacturers put just enough core voltage to get 'game stable' , If the factory voltage is lower than 1.05 Volts, it'll probably need boosting & stability checking with something like OCCT at max complexity for 1 hour, with no artefacts.
- Some past drivers leave rubbish in a hidden Cuda kernel Compute Cache. A clean install is mandatory when installing the drivers.
- Latest Cuda Windows release forces use of a specific Cuda binary for each kernel on compute capability 2.1 GPUs, and has improved fault tolerance, so x41g should be used, as opposed to any prior stock or optimised release.
- For some (unknown) reason some of these cards appear to seat badly on the first try if care isn't taken. Reseating the card can sometimes clear issues.
- These cards use significant power, so sufficient PSU +12V rail current, and decent case airfolw is advised.
Jason
Mike:
I only want to add there is always the possibility to increase voltage or reduce core clock spped with MSI Afterburner.
Or similar tools.
Of course i agree with Jasons comment as well.
Mike
Urs Echternacht:
Finally managed to get a useful bluescreen on Win7sp1 x64. Preparing to post the info over there at AMDs forum.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version