Intel Core i7-1065G7 Ice Lake Linux Performance Benchmarks

Intel's MKL-DNN (or now called DNNL) deep learning workload in its micro-benchmarks did see dramatic speed-ups when using the u8s8f32 data type that is optimized for AVX-512, similar to some of the other dramatic differences within Intel's AVX-512 optimized software. Now if only more software packages were as optimized...

When dealing with f32, the Ice Lake improvement is still there but much less throughout the various MKL-DNN/DNNL configurations.