ROCm 7.0.0 vs. ROCm 7.2.3 Performance On The AMD Radeon AI PRO R9700
While all the excitement and chatter these days is around "AI", it was refreshing to see there were actually some ROCm-OpenCL performance improvements with the updated stack from ROCm 7.0.0 to ROCm 7.2.3. Rare we see much in the release notes or media attention to ROCm common OpenCL improvements compared to the hyper focus on AI workloads.
It was nice seeing some steady improvements to the OpenCL performance with the ROCm 7.0 to 7.2.3 milestones that went under the radar until now. Again, this is just with the updated user-space components with sticking to the same AMDGPU/AMDKFD kernel drivers. If switching those out as well there would be the possibility of seeing greater changes.
