PyTorch's GPU-initiated networking with AMD boosts all-to-all performance by up to 30%
PyTorch · x · 2026-09-30
At PyTorch Conference North America (Oct 20-21, San Jose), AMD's Rishi Sinha will present GPU-initiated networking (GIN) integrated into TorchTitan via the RCCL team's work: GPU kernels issue network puts directly to remote GPUs, improving all-to-all kernel performance by up to 30% and end-to-end performance by 12-15%.
More from Infra
- WSL containers are now GA: run Linux containers natively on Windows — unixterminal · 2026-09-30
- SF Compute confirms it's been supplying GPUs to AI startup neolabs — charles_irl · 2026-09-30
- 1,000 NVIDIA R200 NVL 8 systems to be rentable in APAC by mid-2027 — ns123abc · 2026-09-30
- Targon launches Bare Metal servers and second-level Sandboxes for agents — JosephJacks_ · 2026-09-30
- H trains computer-use agents on SkyPilot with thousands of envs and sub-second cold starts — skypilot_org · 2026-09-30
- Professor: Buy 4 Mac Studios and Run a 1T Model Instead of Burning Half Your Salary on Tokens — DimitrisPapail · 2026-09-30