Which driver version you are on and have you checked other values for --keepfree and --4G-alloc-size like described in the 4G_Windows_Readme.txt?



Well wrt SAM I am not sure yet what effects that has. On ethash you see the impact of the cache when building the DAG (way faster then on RX 5700 although bandwidth did not grow that much) and when mining ZIl, cause that uses a 1G dag that fits to over 12% into the cache - there the card is about 10% faster then for the 4G dag. For other algos I still try to understand how to use it ideally. Generally the card already is the fastest Equihash card among all AMDs (reaching 90+ sol/s on BTG) and its 2nd fastest on Beam. I got good hope to profit from the cache in Beam and Grin C32.