I opted out of NVIDIA tasks because it threw too many errors, still got Intel jobs.

Could I throw out a suggestion for consideration. I'm not entirely sure how difficult it would be to do, but would it be possible to create a GPU validation WU that would actually test in turn each function of Open CL v1, v1.1, and v1.2 and then in the output report print the result of each function test?

Just because the card is reporting that it is capable of CL 1.2 after driver updates doesn't necessarily mean the underlying silicon chips are. My car has a speedo that goes up to about 160mph, but but nothing else on the car will.

Just wanted to say, that some WUs seemed to be stuck on my end as well. The timer continued counting up, but GPU util was at 0 percent. Suspending and unspuspending did the trick most of the time if a WU ever got stuck. Just curious and a bit annoying that I just found out after a couple hours of running beta tasks and some tasks already recorded ~ 2h runtime on a 1660 Super, just to apparently jump back to the last checkpoint and then finish within minutes.

Hu:
Thank you to all who contributed to the issue with my Geforce 660M V.C. in my Alienware laptop.
Here's the output from rebooting Bonic:
2021-03-10 4:12:46 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 660M (driver version 425.31, CUDA version 10.1, compute capability 3.0, 2048MB, 1691MB available, 730 GFLOPS peak)
2021-03-10 4:12:46 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 660M (driver version 425.31, device version OpenCL 1.2 CUDA, 2048MB, 1691MB available, 730 GFLOPS peak)
2021-03-10 4:12:46 PM | | OpenCL: Intel GPU 0: Intel(R) HD Graphics 4000 (driver version 10.18.10.5161, device version OpenCL 1.2, 1400MB, 1400MB available, 166 GFLOPS peak)
2021-03-10 4:12:46 PM | | OpenCL CPU: Intel(R) Core(TM) i7-3740QM CPU @ 2.70GHz (OpenCL driver vendor: Intel(R) Corporation, driver version 3.0.1.10891, device version OpenCL 1.2 (Build 76427))
2021-03-10 4:12:47 PM | | Host name: Clive-Alienware

When I checked TechPower Up GPU-Z and placed my cursor over OpenCL it indicates ver. 1.2 CUDA

Yet the GPU fails the Beta tests. So go figure.............
Clive

Good evening,

As for the question on our GPU cards to test, it is very limited but they mostly mid range cards. We don't have any of the cards that are causing issues on the grid. However, as mentioned before we do have a code fix to the opencl code that may fix the double precision issue. I'm guessing that some of these cards that are failing don't have double precision capabilities and aren't throwing an error to indicate it. The researchers found 2 locations in the code that might leave ambiguity in the definition of the float.

We have received the first production batches from the researchers a few hours ago and are working to get them downloaded, built, and loaded using all the automation steps on the servers to be ready for full launch. We will be testing these on alpha first and then moving them to beta. As a plus, one of the main developers of AutoDock has started to explore the points issue :)

Another issue that was diagnosed with the latest beta results is that the validator did have a bug in it that was causing some invalid results being marked as valid. This has been updated and will be tested in the next Beta test as well.

I'm sure I'm forgetting something, but again, thank you for all your help!

Thanks,
-Uplinger