Intel Shares First Benchmarks of Its Next-Gen Crescent Island "D270" GPU, Claiming a 5.9x Throughput Over RTX PRO 6000 As Xe3P Goes Big On FP64
Key Points:
- Intel's D270 "Crescent Island" GPU, based on the Xe3P architecture, features 32 cores and native full-rate FP64 double-precision compute, targeting high-performance computing (HPC) and AI workloads that require numerical accuracy.
- The GPU offers up to 480 GB of memory with an LPDDR5X subsystem, supporting configurations from air-cooled 350W models to liquid-cooled dual-GPU setups, emphasizing efficiency and scalability.
- Benchmark tests at SC26 show the D270 outperforming NVIDIA's RTX PRO 6000 Blackwell Server Edition by up to 5.9x in FP64 financial services workloads, while consuming significantly less power (350W vs. 600W).
- In AI benchmarks, the D270 delivers 29% lower latency and 12% higher throughput on large language models compared to NVIDIA's RTX PRO 6000 Blackwell, demonstrating strong performance for decision-making and inference tasks.
- Intel is showcasing the GPU's capabilities in seismic modeling with FP32 compute, highlighting its suitability for demanding industry workloads, though launch timing and pricing details remain undisclosed.