Just a quick question to start the New Year...Is there any possibility of Intel GPU work units? With Intel getting into the discrete video card field & a large number of systems with integrated GPUs it would seem like a good idea?

This project only has CUDA applications. Intel GPUs cannot be used. Nvidia only. Some of their applications are licensed from larger institutions and might not be able to be ported to OpenCL, and it seems pretty clear that their developers are most comfortable dealing with CUDA, as it’s the industry standard when dealing with AI/ML type work that their other applications are using. It’s possible to use OpenCL devices, but they may not have the time or resources to port their CUDA code.

Even if they could, at least the integrated graphics varieties would struggle to complete work units in the short 5-day deadlines here at this project.
____________

This project only has CUDA applications. Intel GPUs cannot be used. Nvidia only. Some of their applications are licensed from larger institutions and might not be able to be ported to OpenCL, and it seems pretty clear that their developers are most comfortable dealing with CUDA

And, with all this "comfort" on Cuda, the apps are bugged very often.
Great

no different than any other new application. just because something is comfortable, doesn't mean it's perfect the first try.

the app available right now (Quantum Chem) is still in beta phase. errors/issues expected. it's not a production release. but this batch of ~100,000 workunits or so that they've released in the past week seem to be processing with few errors (only errors due to not enough memory on the GPU)

all of the issues they've had have nothing to do with CUDA (or any other platform). it's been mostly configuration and input file issues.
____________