Hello. I see many messages regarding wrong results that AMD RX 5700 cards producing.
Those results are consistent (being WRONG) enough to pass statistical validation.
Why AMD being long-term and knowable corporation allows itself to put such broken product on market?
And when to expect fix for OpenCL part of drivers for those cards??
Also, who responsible in AMD for OpenCL drivers quality control and is he still not dismissed?
It would be helpful if you can provide more details about the OpenCL related issue(s) that you've mentioned above. I'll check with the concerned team and try to get an update on this.
Thanks.
I have no time for debugging right now but at first glance it seems wrong computation on very initial pipeline stages of SETI@home app. Perhaps, first FFT already wrong.
So, check available in sources oclFFT work on those cards.
Thank you. If you suspect that some FFT calculation is producing wrong result on RX 5700 with the latest driver, we can try to verify/reproduce it at our end. To reproduce the issue, we need below inputs:
Thanks.