AMD prépare l'accélérateur de calcul MI250X avec 110 UC et 128 GB HBM2E

AMD is preparing an update to its compute accelerator lineup with the new MI250X. Based on the CDNA2 architecture, and built on existing 7 nœud nm, the MI250X will be accompanied by a more affordable variant, the MI250. According to leaks put out by ExecutableFix, the MI250X packs a whopping 110 unités de calcul (7,040 processeurs de flux), running at 1.70 GHz. The package features 128 Go de mémoire HBM2E, and a package TDP of 500 W. As for speculative performance numbers, it is expected to offer double-precision (FP64) throughput of 47.9 TFLOP/s, ditto full-precision (FP32), et 383 TFLOP/s half-precision (FP16 and BFLOAT16). AMD’s MI200 “Aldebaran” family of compute accelerators are expected to square off against Intel’s “Intel a abandonné son projet de construire commercialement des GPU pour serveurs Xe-HP” Xe-HPC, and NVIDIA Hopper H100 accelerators in 2022.