Intel's 350W D270 claims 5.9x lead over RTX PRO 6000 — but the fine print reveals benchmark games
StefanoGogioso · x · 2026-11-22
Intel's official benchmarks claim its 350W D270 (Xe3P) beats the 600W RTX PRO 6000 Blackwell by 5.9x. Reading the report's fine print shows how:
- Where 5.9x comes from: All headline wins are FP64 option-pricing workloads (Monte Carlo 5.9x, Black-Scholes 5.6x, binomial 4.6x). Nvidia crippled workstation-flagship FP64 to 1:64 to protect B200/H200 margins, while the D270 offers native full-speed FP64. The appendix also shows Intel's rig had 1× D270 + 3× Arc Pro B70 vs 8× RTX PRO 6000, without stating how many cards were active.
- LLM inference efficiency is genuinely competitive: Single-card Qwen3-30B-A3B (FP8) throughput is 1.12x the RTX PRO 6000, with 29% lower latency on Drex 2.0 — at roughly half the power draw.
- The local-LLM crowd's dream spec: Instead of HBM, Intel stacks LPDDR5X — 160GB base at 1.5 TB/s, up to 480GB on custom boards, versus 96GB GDDR7 on the RTX PRO 6000.
Related event: Intel D270 Early Benchmarks: 160GB Memory, Lower Power Draw(2 posts)→
More from Infra
- Gemini 4 Neon autonomously writes GPU kernels, 5.9x faster prefill on AMD hardware — sebkrier · 2026-11-22
- SenseNova-Loop-9B Hits 1,300+ tok/s on a Laptop RTX 5090, Benchmarked vs Qwen 27B — tebjan · 2026-11-22
- A tale of two Nvidias: solid earnings vs. execs calling for Jensen's return — TechCrunch AI · 2026-11-22
- Chollet: GPU spot prices keep rising as HBM sells out through 2029 — AI compute buildout isn't over — fchollet · 2026-11-22
- DeepMind bakes speculative decoding into Gemma 4: up to 3.1x server speedups, 2.2x on smartphone GPUs — dair_ai · 2026-11-22
- Gavin Baker: No investor feels misled by Nvidia's China GPU disclosures — GavinSBaker · 2026-11-22