Folding on AMD Radeon 8610G- Driver Problems

[Yeroon]

As far as I know, your Radeon 8610G should not be listed as supported for folding. Is is from a generation of integrated GPUs as part of APUs sold by AMD that did not include support for doing double precision calculations (FP64). In addition AMD has limited full OpenCL support in newer drivers to more recent GPUs that the Terascale based ones like the 8610G.

If you can provide the PCI device ID number I can check the listing in GPUs.txt and have it corrected.

Joe_H,
I was going through the gpu.txt file today to see what falls under AMD species 5, and noticed what is probably the issue you are looking for. There are a few APU terascale gpu's listed as species 5, and a bunch as species 2, which are not able to do any DP compute.

For the 8610G mentioned in this thread, it is listed as 0x1002:0x990f:1:5:Tinity [Radeon HD 8610G]

Others that seem to be mis-labelled, including above:

Code: Select all

0x1002:0x990b:1:5:Tinity [Radeon HD 8650G]
0x1002:0x990d:1:5:Tinity [Radeon HD 8550G]
0x1002:0x990f:1:5:Tinity [Radeon HD 8610G]
0x1002:0x9996:1:5:Richland [Radeon HD 8470D]
0x1002:0x9998:1:5:Richland [Radeon HD 8730D]
0x1002:0x999c:1:5:Richland [Radeon HD 8650D]
0x1002:0x999d:1:5:Richland [Radeon HD 8550D]

If I were to guess why they were not assigned/re-assigned correctly, the spelling of trinity for those listings is wrong/listed as Richland. Remainder of Trinity is listed as species 2.

Also, a single species 4:
0x1002:0x983d:1:4:Temash [Radeon HD 8250]
While the rest of Temash gpus are species 5; though these are all so weak I'm not sure they could to a WU in time.

[bruce]

Confirmed. The Devastator Graphics HD 8610G does not support FP64 calculations and should be excluded from GPU folding. It reportedly does support OpenCL 1.2 so I can see where the confusion came from. I'm not sure that's really true.

The Kalindi graphics HD 8250 IGP does support FP64 and OpenCL 1.2 so technically it meets the minimum requirements for GPU folding. Whether it can meet the deadlines is a separate issue.

I've also corrected the spelling of Trinity.

I only see one Temash [Radeon HD 8250].

I'm not really confident that I've cleaned everything up. Apparently this all happened while ATI/AMD was going through some significant transitions. If anybody has iGPUs of this era, I'd appreciate specific reports, along with the GPU ID code.

[Yeroon]

Well, Richland was a respin of Trinity (minor clock/power fixes) with the same gpu, so my understanding of how this works is all the richland ids would be the same species as their trinity counterparts.
A few other id naming errors as far as my research goes:
0x1002:0x990c:1:2:Oland [Radeon HD 8670D] 648 - should be Richland [Radeon HD 8670D]
0x1002:0x990e:1:2:Oland [Radeon HD 8570D] 432 - should be Richland [Radeon HD 8570D]
0x1002:0x9999:0:0:R700 [Radeon HD 4600] -- should be Richland [Radeon HD 8510G]

All of which means little other than those are the only Richlands that are currently [accidentally] correctly sorted (1:2). None have DP compute.

EDIT:
If you are there correcting IDs to appropriate species:
0x1002:0x9905:0:0:[FirePro A300 Series]
0x1002:0x9906:0:0:[FirePro A300 Series]
Are both trinity based firepro apus as well. Same devastator gpu, same lack of DP
https://www.techpowerup.com/gpu-specs/f ... a300.c1889