Skip to main content
Intel says its D270 data center GPU beats Nvidia's RTX PRO 6000 by up to 5.9 times in first benchmarks - MLQ.ai

Intel says its D270 data center GPU beats Nvidia's RTX PRO 6000 by up to 5.9 times in first benchmarks - MLQ.ai

NVIDIA GPUs
16 lis 2026, 22:39Hardware News
Oryginał
2 min read

Intel has released benchmark results showing its D270 data center GPU outperforming Nvidia's RTX PRO 6000 by as much as 5.9 times on financial workloads.

The figures, released at the SC26 supercomputing conference in Chicago, come from Intel's own tests conducted on pre-production silicon. A single 350-watt D270 card delivered up to 5.9 times the throughput of Nvidia's 600-watt RTX PRO 6000 Blackwell Server Edition on European Monte Carlo options tests, with similar advantages on other financial benchmarks.

Performance Details

Intel's D270 showed 5.6 times the performance on European Black-Scholes tests, 4.6 times on binomial pricing, and 1.92 times on American Black-Scholes. The GPU is built on Intel's Xe3P architecture with 32 Xe cores and 256 XMX matrix engines, using 160GB of LPDDR5X memory instead of the stacked HBM memory common in AI accelerators.

Intel has marketed the card mainly for AI inference; the new post argues it also belongs in high-performance computing

Testing Methodology

The benchmark comparisons were conducted on November 6, with Intel testing one D270 in a single-socket Xeon 6741P server against an Nvidia server containing eight RTX PRO 6000 cards. For AI inference tests, Intel ran the Nvidia card on an Amazon Web Services EC2 g7e virtual machine rather than bare metal.

AI Inference Results

While financial computing showed dramatic improvements, AI inference gains were more modest. The D270 delivered 1.12 times the output token throughput on Qwen3-30B-A3B and up to 29% lower latency on Drex v2.0. Intel also demonstrated a seismic imaging application ported from HIP and CUDA to SYCL in less than a month.

Market Position

Intel named the D270 at the OCP Global Summit in October, when it also listed 1.3 petaflops of peak FP8 compute. Customer sampling began that same month, though Intel has not announced pricing or general availability. The company recently formed a 23-company Korean alliance to promote the D-Series as an inference card that avoids scarce HBM memory.

Z.ai
Article written by AI
Z.ai GLM-4.5-FlashFree
Free AI articles150 of 150
No limit after signing in with Telegram