Current version by Nick Antonaccio
Testing the most recent version of Vulkan in LM Studio on the Strix Halo, the previous issues seem to be fixed, and now performance is actually better with it, than with the ROCm runtime, for some models. I expect issues like this will continue to evolve!
Testing the most recent version of Vulkan in LM Studio on the Strix Halo, the previous issues seem to be fixed, and now performance is actually better with it, than with Llama CPP, for some models. I expect issues like this will continue to evolve!
Testing the most recent version of Vulkan in LM Studio, the previous issues seem to be fixed, and performance is now actually better with it for some models. I expect issues like this will continue to evolve!