A chipless NVLink bridge PCB costs $500, derailing a dual-3090 vLLM setup
DjCanalex · reddit · 2026-10-01
A Redditor complains that an NVLink bridge for a dual RTX 3090 setup — a bare PCB with no chips on it — costs $500, killing his plan to get stable tensor speeds with vLLM. He asks the community whether anyone has measured a significant difference from using the bridge under llama.cpp or vLLM.
More from Infra
- Apple quietly added FlexCache to M6, echoing Qualcomm's same architectural bet — clattner_llvm · 2026-10-01
- H100 rental prices up 40% since October, six-year-old A100s still fetching up to $19,000 — r0ck3t23 · 2026-10-01
- An AI Agent Describes the Machine Economy: Micro-Payments via x402, Then Dissolve — RileyRalmuto · 2026-10-01
- Why Agentic Inference Needs P/D Disaggregation: Decoding 300K-Token Codebase Latency — abhijithneil · 2026-10-01
- RTX 3090 throttles on Qwen i2i in ComfyUI: 600s per image vs 40s on a 5070 — NotEnoughVRAM · 2026-10-01
- rakyll: Serverless works until your workload diverges from the baseline assumptions — rakyll · 2026-10-01