DeepSeek job posting reveals 100k-accelerator cluster challenges amid Ascen 950 training rumor
teortaxesTex · x · 2026-11-20
A DeepSeek job posting for supercomputing cluster ops reveals stark challenges: making a 100,000-accelerator cluster as stable as a single computer (near 300 failures to handle daily), tuning new nodes to peak performance out of the box, building observability for GPU utilization/NCCL latency/checkpoint signals, and automating cluster self-healing so on-call engineers sleep through the night. The poster notes reliably running such clusters is something DeepSeek — and nobody in China — has done with these chips; getting hardware is only step one. A quoted tweet claims (unverified) that DeepSeek V5 will train entirely on Huawei Ascend 950, despite multi-fold costs.
More from Infra
- GTA VI hacked to 60fps with DLSS 5 and first-person within 48 hours, footage looks lifelike — LinusEkenstam · 2026-11-20
- Celesto CI claims 90% cheaper than GitHub Actions with benchmark table — aniketmaurya · 2026-11-20
- Blogger caps Anthropic at $200B revenue, predicts CoreWeave collapses first — sidjustice_ · 2026-11-20
- Cerebras Hardware Deep Dive: How Wafer-Scale SRAM Cuts Data Movement for Fast LLM Inference — draecomino · 2026-11-20
- LoRA merge + requantize fix boosts Qwen3.8B speed ~50% with zero syntax fails — yb2698 · 2026-11-20
- Neon's MCP server now executes SQL as a selected Postgres role, constraining agents via DB grants — JeremyCMorgan · 2026-11-20