C1 20 GPU acceleration uses wrong GPU due wrong assumptions
Hello everybody,
there were some post about this issue but not really an answer.
On my laptop (Windows 10, Version 2004) C1 20 uses the internal GPU Intel UHD Graphics 630 with up to 100 % and perform computing only under 10 % on the more powerfull discrete GPU Nvidia Geforce GTX 1060 with Max-Q Design.
When I have a look at the imagecore accelleration files (C:\Users\User\AppData\Local\CaptureOne\ImageCore\) you can see that C1 is based on wrong assumptions:
Intel UHD Graphics 630:
<GlobalMemoryMB>13024</GlobalMemoryMB>
<NumberOfKernels>1298</NumberOfKernels>
<NumberOfCompiledKernels>1298</NumberOfCompiledKernels>
<Benchmark>0.560687</Benchmark>
Nvidia Geforce GTX 1060 with Max-Q Design:
<GlobalMemoryMB>6144</GlobalMemoryMB>
<NumberOfKernels>1298</NumberOfKernels>
<NumberOfCompiledKernels>1298</NumberOfCompiledKernels>
<Benchmark>0.119552</Benchmark>
C1 uses a wrong Global Memory size for the Intel GPU (13GB instead of 1GB) and is calculating with a wrong Benchmark size (after C1 the internal GPU Intel is 5 times faster than much more powerful discrete GPU Nvidia)
Is it a bug?
It was not possible to change this behavior (change the driver Version of the GPUs, reset the immagecore, etc.)
Thanks for your suggestions.
Thomas
-
Thomas,
I cannot comment on the numbers used but can observe that the lines in the log files that I see are sometimes mixed together and perhaps a little difficult to be sure about which line applied to which device - i.e. Device 0 and Device 1.
Also the resulting Benchmark seems consistent. The lower the number the more powerful the assessment of the Benchmark. So the Nvidia is assessed as offering 5 times the power of the Intel.
Outright power of a GPU option is not necessarily the only factor involved in how to route data since the data transfer rates must also be considered. And also the GPU route is not always appropriate for all types of processing the image has applied. Sometimes CPU is more appropriate.
But if your really think there is a problem with the way the Intel facility has been assessed then you should contact the C1 Support Team using the "Submit a request" option found at the top of most Community pages and discuss things with them 1 to 1.
For what it is worth my older system discounts using the Intel inbuilt functionality and assesses my old Nvidia GPU Benchmark as somewhere between 2.5 and 3. It seems that under 3 means it it is sometimes worth using it.
0 -
The NVidia Control Panel has a way to force a program to use a specific GPU, if you have multiple ones. Look under 3D Settings -> Manage 3D settings. You will find a global setting and a Programs Setting tab. Use the Program Setting tab if you want to force a program to use a specific one. The field is called OpenGL rendering GPU.
0 -
Dear Kevin,
thanks for your answer.
After I had to upgrade to the Geforce Experience to get the latest Nvidia driver, the Nvidia control panel does not work anymore (the installed driver is not compatible). There is a discrepancy in usage, you can only use the control panel or the Experience App. In the Experience App you cannot set the specific Open GL settings (unfortunately it is only optimized for games).
But you can do it in the Windows settings under Display Setting, but maybe this is not really working.
Thomas
0 -
Thomas,
I have the Nvidia studio rivers installed. If you feel ambitious, you might try them instead of the game drivers. You can change between game ready Driver and Studio Driver in GEforce Experienc under Drivers and click on the three dots near the upper right hand area.
Kevin
0 -
Kevin,
I had also installed the studio driver.
I decide to make a clean installation of the Nvidia driver (change to the DDH driver) and finally I got rid of the Experience App and I have now the Nvidia Control Panel working again.
The C1 initialization of the ImageCore accelleration needs now 3 times longer as before and there is a little bit more use of the Nvidia GPU (it is now between 10-20 %)
Thomas
0
Post is closed for comments.
Comments
5 comments