Hi there I’m experimenting with OpenCL on GTX1050ti in Linux. The following picture shows, that the tiles rendered by the GTX are wrong. (The others where rendered by a Vega 64). In Win10 this doesn’t happen.
I already tried all versions of the NVIDIA driver available from their page.
The question is, is this a known problem?
If yes, then is it a problem in the drivers or blender?
Please, if you report this problem, you inform to developers that you also have an AMD card, and you share benchmark results of 2.79 using AMD only, nvidia only (CUDA), and both cards at the same time (OpenCL), to note that it is really important in your case be able to have nvidia card working properly with OpenCL.
Edit:
A question just out of curiosity if you have time available. How do you do in Linux to be able to have AMD and Nvidia working at the same time? Is it the only way to achieve it, having nvidia handling the display with proprietary drivers? Or is it also possible AMD card as primary display handling the display?
Hey @YAFU , thats the plan. As I said, I’m experimenting with this. If I will file a report I want to make sure, that it’s not useless.
Adressing your edit: I am using the Vega 64 as display driver and adding in the GTX 1050ti to render. I am perfectly willing to share what i have done. But I first wanted to gather some (at my side) missing information.
Edit: Is there interest in explaining what I’m doing here? If so, I will put up a somewhat longer story…
I’m just curious about how to configure drivers (open source/propietary) and primary display in Linux when you have nvidia and ATI/AMD cards at the same time. But it’s just curiosity since I do not have AMD card, and it’s mostly to point to your explanation when some other user has a problem related to this. Just when you have time available to do so.
Regarding the first thing I wrote in other message… I perfectly know how to render with OpenCL in nvidia, even before split kernel existed. The issue is that developers label as “very low priority” any report in the tracker related to nvidia and OpenCL. Basically the answer is: you use CUDA
So if in addition to reporting the problem you show that you have good benefits in render time with the two cards at the same time using OpenCL, that would be good to show to developers that it is useful/important to have OpenCL working correctly in nvidia.