Tagged hbm
2 entries
2026-08-27
Brief · 27 August 2026NVIDIA unveiled NVLink Fusion, a new interconnect that merges NVLink with HBM to create NVHBM, promising up to 2.5 TB/s per‑socket bandwidth for next‑gen AI servers. The announcement came alongside Amazon’s pledge to add 2 million more Nvidia GPUs to its data‑center fleet. [1][2]
2026-08-26
Brief · 26 August 2026An arXiv pre‑print introduced the Maia 200 accelerator, claiming 10,145 TFLOP/s FP4, 5,072 TFLOP/s FP8, 7 TB/s HBM bandwidth and a 750 W TDP, positioning it as a new class of software‑defined dataflow AI chips. [arXiv 2026‑08‑25]