RX 7900 XTX beats R9700 by 30% on GPT-OSS-20B local inference tests
glenbeer · x · 2026-09-16
Enthusiast 1337hero revisited the RX 7900 XTX for local LLM inference: it ran GPT-OSS-20B about 30% faster than R9700s, though Qwen3 8x27B (dense) numbers looked off, while Gemma4-26B-A4B Q80 came in fast — useful reference points for AMD local deployment.
More from Infra
- Japan's chip exports catch up to autos as AI boom lifts the cycle — PAstynome · 2026-09-16
- Dev rebuilds Radish: a Redis server running entirely inside a Cloudflare Durable Object — whoiskatrin · 2026-09-16
- Anthropic lands its first Australian datacentre, a $31.9bn project in western Queensland — nordicinst · 2026-09-16
- Manchester trains UK-wide air pollution model on NVIDIA Earth-2 in just two days — NVIDIA Blog · 2026-09-16
- SemiAnalysis: Vera Rubin NVL144 hits ~7x tokens per MW vs Blackwell, over 2x profit per GW — sudoraohacker · 2026-09-16
- Musk explains why Terafab must exist: Taiwan chip risk plus capacity ceiling — elonmusk · 2026-09-16