Folding Forum

Hi!

I have a FirePro V7800, which is confirmed to be whitelisted, that after days of trying is still coming back with "No WUs available for this configuration" "Could not get an assignment". This issue is also present on another system running dual HD 6870's, again GPU's that are in the whitelist. Is FAH/Stanford just not making WU's for these models anymore? I've tried all the other troubleshooting steps like reloading the whitelist, pausing and restarting many times in a row, and letting it run for days trying to get a WU.

Thanks.

Welcome to the folding support forum.

Part of the issue has been a change in the requirements for many of the GPU projects. They found for many of them that they needed to use mixed precision, single precision calculations for most operations and using double precision for some critical calculations. AT the moment almost all active GPU projects require mixed precision.

A question has been sent to the project managers to see if some projects would be able to be run using just single precision, currently waiting on a response.

Related to this is the way the AMD GPU's were initially categorized for the GPUs.txt file and the use by AMD of multiple hardware architectures in each series of cards. FOr instance, your HD 6870's use a Terascale 2 chip that could in theory support double precision. But AMD only enabled double precision in the next generation Terascale 3 chip used in the HD 6900 cards.

Your Firepro V7800 using the same Terascale 2 chip does have double precision enabled. But it is lumped in with the same status as the other Terascale based GPU's that don't.

The GPUs.txt file is slowly being updated for individual cards that can support double in theory. It takes finding which can support mixed mode calculations. Not sure how long it will take to get the entry changed for the V7800. Once changed, let us know how well it works, we have had reports that the current GPU folding core does no perform well on Terascale based cards, that it does better on the newer GCN based ones.