Hello,
It looks like, the computing works, but, there is an error with gpu info.
I have these 3. I think Arc computes, but it shows that AMD done this.
How to show proper hardware?
https://pasteboard.co/I8HLbWbWS0rV.png
https://pasteboard.co/481D0zSWNZgB.png
GPUs Info
Moderators: Site Moderators, FAHC Science Team
-
- Site Admin
- Posts: 7936
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2 - Location: W. MA
Re: GPUs Info
The GPU info is probably correct, the Arc 770 is not enabled to do folding currently as it does not support double precision in hardware. Intel provides emulated DP calculations through its driver, but tests have shown that does not work properly with the current GPU folding core.
It looks like you are using Task Manager to see GPU utilization. By default that displays video rendering, not the calculation processing. You would need to change to one of the optional Compute modes to show F@H utilization of the AMD GPU. Or use something like GPU-Z which will directly show that.
It looks like you are using Task Manager to see GPU utilization. By default that displays video rendering, not the calculation processing. You would need to change to one of the optional Compute modes to show F@H utilization of the AMD GPU. Or use something like GPU-Z which will directly show that.
iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
-
- Site Moderator
- Posts: 6359
- Joined: Sun Dec 02, 2007 10:38 am
- Location: Bordeaux, France
- Contact:
Re: GPUs Info
There's nothing wrong here ... the client seen the the Vega M GPU and the Intel HD 630 integrated in your CPU since they are both capable to fold. Although the Intel iGPUs are too slow for most work, which mean why it doesn't get work.
The Intel dedicated GPUs (Arc) are no capable of folding, so no one ever bother entering them in the GPUs.txt file, which explain why it doesn't appear in the client.
Please post soe logs if we didn't answer your question : viewtopic.php?f=24&t=26036
The Intel dedicated GPUs (Arc) are no capable of folding, so no one ever bother entering them in the GPUs.txt file, which explain why it doesn't appear in the client.
Please post soe logs if we didn't answer your question : viewtopic.php?f=24&t=26036
-
- Site Admin
- Posts: 7936
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2 - Location: W. MA
Re: GPUs Info
It doesn't appear in the web control, but there actually is an entry in GPUs.txt to blacklist the Arc A770:
Code: Select all
0x8086:0x56a0:::DG2 [Arc A770]
Re: Información de las GPU
Isn't it enough for a GPU to have OpenCL support to work with FahClient? What version of OpenCL is required? In addition to support for that version of OpenCL, does the GPU need to have FP64 support?
Last edited by jlopz on Sat Jul 13, 2024 4:40 pm, edited 3 times in total.
-
- Site Admin
- Posts: 7936
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2 - Location: W. MA
Re: Información de las GPU
Minimum OpenCL version is 1.2. Yes, FP64 support is required. The current GPU folding cores use some double precision calculations where needed to maintain sufficient numerical accuracy. They investigated use of emulated FP64 calculations, they did not always work out and would slow all GPU folders down.
Current GPU cores also use NVidia CUDA instead of OpenCL on GPUs supported by the CUDA libraries included in the code. The oldest Nvidia GPUs supported by F@h with Kepler chips fall back to OpenCL, Maxwell and newer chips use CUDA if the drivers and support libraries are installed correctly.
iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3
Re: Información de las GPU
Thanks for your answer. It's a shame not to be able to use the Intel ARC, they would perform great.Joe_H wrote: ↑Sat Jul 13, 2024 7:03 pmMinimum OpenCL version is 1.2. Yes, FP64 support is required. The current GPU folding cores use some double precision calculations where needed to maintain sufficient numerical accuracy. They investigated use of emulated FP64 calculations, they did not always work out and would slow all GPU folders down.
Current GPU cores also use NVidia CUDA instead of OpenCL on GPUs supported by the CUDA libraries included in the code. The oldest Nvidia GPUs supported by F@h with Kepler chips fall back to OpenCL, Maxwell and newer chips use CUDA if the drivers and support libraries are installed correctly.
-
- Site Admin
- Posts: 7936
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Pro 2.8 quad 12 GB smp4
MacBook Pro 2.9 i7 8 GB smp2 - Location: W. MA
Re: GPUs Info
Perhaps it will be possible with a newer generation ARC line of chips. It is possible Intel will reconsider taking out the hardware FP64 support and include it again. It was present in a few generations of their iGPU designs before ARC which worked for some smaller F@h projects..
iMac 2.8 i7 12 GB smp8, Mac Pro 2.8 quad 12 GB smp6
MacBook Pro 2.9 i7 8 GB smp3