Thanks for reporting the problem.
Actually the clock frequency reported by the runtime (hence, clinfo output or clGetDeviceInfo call in any OpenCL application) should match with the "core clock" number shown in the AMD Radeon settings panel. Could you please check it and share your observation? It would be helpful to trace the issue accordingly.
WattMan shows correct value (1200Mhz) while OpenCL runtime returns 2000Mhz. Moreover, I've changed memory frequency within WattMan to 2035Mhz but OpenCL runtime still returns 2000Mhz, so it doesn't looks like they are really related.
Also, RadeonSettings.exe crashed a couple of times, part of the log:
Sig.Name=Fault Module Name
Sig.Name=Fault Module Version
Sig.Name=Fault Module Timestamp
UI=Radeon Settings: Host Application has stopped working
Thanks for sharing the information. I've already forwarded the issue to the runtime team. As I get any feedback, I'll come back to you.
The OpenCL team does not think it's a runtime issue, rather suspecting a base driver issue. As they confirmed, the OpenCL runtime reports core or memory frequency as provided by the base driver. I'm trying to contact with the concerned team for more information.
Meanwhile it would be helpful if you could share the clinfo output and screenshots of the Radeon settings including current settings in Wattman.
OK, I've uploaded .zip archive with 2x .jpgs & 2x clinfo's .txts here:
Basically, I've only changed state 6 & 7 within WattMan to 1200Mhz/1050mV.
Thanks for providing those information. Once I get any reply from the concerned team, I'll get back to you.
Please can you attach a DXDIAG report to your next response so we can check your system spec and driver details?
Just to inform you that a bug report has been filed against the issue.
Give me a link to your previous topic.
It seems quite unbelievable, that such small frequency decrease can resolve overheat problems