I just upgraded from an ASUS ZenBook with Intel Core i5-2557M (2 Core/4 Threads, HD3000 graphics) using DxO Optics Pro 10 to a Lenovo Yoga with AMD Ryzen 5 4500U (6 Core, Radeon RX Vega 6 graphics) using DxO PhotoLab 3.
For both systems, enabling OpenCL doesn’t have any effect. It looks as if the option is ignored. Are those GPUs too weak for OpenCL to make a difference? I just checked the sensors of the new Ryzen and there is no increase in GPU clock and usage after enabling OpenCL.
I also profiled the processing of 2 images in parallel in the hope that processing can be parallelized and distributed to both, CPU and GPU. Still OpenCL does not make a difference.
So how is DxO’s OpenCL implementation actually working? It looks like it is disabled internally even though the option is actually enabled (application has been restarted after toggling the option). What kind of graphics card is actually needed for OpenCL to have an effect?
Hello @daniel.b and welcome to the forum,
Let me ask @Lucas to reply you.
If you really mean “Radeon Vega 6” and not “Radeon RX Vega 64”, then indeed it is not powerful enough to give any benefit.
When OpenCL is enabled, it is used for image demosaicing with no denoising or standard denoising only.
Thanks for your quick answer.
It is indeed an APU with Vega 6, no dedicated RX Vega 64 card.
So does this mean that when I use PRIME on a photo, OpenCL is automatically disabled? I actually enabled PRIME for my tests in order to get maximum processing. But it looks like a 300W graphics fast car is required to benefit from OpenCL anyway. Thought that an entry-level 384 ALU thingy could help quite a bit too…
For the processing that needs PRIME, OpenCL is indeed not used. This doesn’t prevent its use for other processing without PRIME that may happen in the meantime though.
Basically a recent (< 5 years old) discrete GPU is indeed needed to see a benefit with OpenCL. I don’t know about power consumption as it’s hard to get the information, but there’re many great graphics cards below 300W TDP.
Also I can’t tell more for now but please note that for the next version of PhotoLab, having a good GPU will make a significant difference.
Looking forward to that and to finding out what “good GPU” actually means .
The AMD Renoir APU (Ryzen 4500U) with its 6 core CPU in my new ultrabook processes a 16mp RAW file in 3s which is quite nice though. Only PRIME cuts in badly.
To make it simple, let’s say a gamer GPU
PL writes log files in your Documents folder which include the results of CPU and GPU benchmarks.
Yes. I checked that already. Here are the results for the AMD Ryzen 5 4500U:
cpu perf : 30.0699
PerfIndex = 26.5045 Status = OpenCL device is slower than the main CPU BinOffset = 0 BinSize = 1812120
My GTX980Ti has a performance index of 242. If you look at opencl or GPU benchmarks for example here https://browser.geekbench.com/opencl-benchmarks you can use that as one data point to roughly predict the index for other GPUs.
It would take something like a AMD Radeon HD 7800M to double the performance index of your CPU for example.