MI355
A high-end AI accelerator card from AMD's Instinct series targeting data center workloads. It features substantial compute throughput and memory bandwidth, supporting both AI training and inference workloads.
Timeline (last 90 days)
On the AMD MI355, FlyDSL's FP16 GEMM achieved approximately 19% geometric mean speedup compared to Triton.
NVFP4 is originally an NVIDIA proprietary format, and AMD's support for this format on the MI355 means that native hardware-level computation support for this numeric format has been added.
All Facts (2)
On the AMD MI355, FlyDSL's FP16 GEMM achieved approximately 19% geometric mean speedup compared to Triton.
50%UnverifiedNVFP4 is originally an NVIDIA proprietary format, and AMD's support for this format on the MI355 means that native hardware-level computation support for this numeric format has been added.
50%