GPUDatacenterReleased —

Intel GPU D270 (Crescent Island)

480 GB LPDDR5X

Intel's first data center GPU built purely for AI inference, developed under the codename Crescent Island and named the Intel GPU D270 in October 2026 as the first model of the GPU D-Series. It pairs 32 Xe3P cores with LPDDR5X instead of HBM: 160 GB on Intel's own card and up to 480 GB on partner designs, so large models fit in memory at a far lower cost per gigabyte. Intel rates it at up to 1.3 PFLOPS of FP8 in an air-cooled 350 W PCIe card. Customer sampling began in 2026, with volume availability expected in 2027.

General

Manufacturer: Intel
Model: GPU D270 (Crescent Island)
Segment: Datacenter
GPU chip: Crescent Island
Architecture: Xe3P (Celestial)
Release date: —
Price: Search on Amazon

Memory

Memory size: 160 – 480 GB
Memory type: LPDDR5X
Bus width: —
Bandwidth: —
Memory clock: —

Clocks & silicon

GPU clock: —
Boost clock: —
Shader units: —
Tensor / AI cores: 256
RT cores: —
Transistors: —
Technology (nm): —
TDP / TBP: 350 W

Interface & APIs

Bus interface: PCIe 5.0 x16
SLI / multi-GPU: —
DirectX: N/A (compute-only)
OpenGL: N/A (compute-only)
Vulkan: N/A (compute-only)
Other APIs: oneAPI Level Zero, SYCL, OpenCL

AI support

Precisions: FP8FP16BF16INT8INT4TF32FP64
AI accelerator: Xe Matrix Extensions (XMX), 256 engines with 16-deep systolic array
Peak AI TOPS: 1,300 TOPS
Notes: 32 Xe3P cores in four compute slices, each core with eight vector engines and eight XMX engines, plus 32 MB of unified L2. The XMX units use a 16-deep systolic array, add FP8, FP4 and microscaling formats, and reach up to 1.3 PFLOPS of FP8. FP64 runs at full rate. Models are served through vLLM and PyTorch.

All Intel GPUs →

Datacenter gpus from Intel announced around the same time as the Intel GPU D270 (Crescent Island). Add any of them to your compare list to see the specs side by side.

gpu