Specifications
Vendor
Intel
Product
Xeon 7 "Diamond Rapids"
Cores
Up to 256
Last-Level Cache
1.28 GB
Process
Intel 18A-P
Foveros 3D stacking
Memory Channels
16× DDR5-8000
+ MRDIMM-12800
Memory Bandwidth
1.6 TB/s
I/O
128× PCIe Gen6 + CXL 3.0
+ UPI 3 + 8× Gen4
Accelerators
AMX + AVX 10.2
QAT / DSA / IAA
Architecture
Fabric Hub + compute
building blocks
Availability
2H 2027
Stock
Available
Price
Quote
Intel Diamond Rapids — The 2027 Xeon Platform
Diamond Rapids is Intel's next-generation Xeon 7 server platform, detailed at Hot Chips 2026. It is the follow-on to Xeon 6 "Granite Rapids" and Intel's entry into the high-core-count, high-memory-bandwidth lane against AMD's EPYC. A flagship configuration combines four compute building blocks for up to 256 cores and 1.28 GB of last-level cache.
At the heart of the design is a modular Fabric Hub that centralizes memory and I/O, with compute building blocks connecting through a unified memory fabric and a flexible I/O fabric. Core chiplets (up to 16 cores each) connect to a base tile via a 3D crossbar using Foveros 3D hybrid bonding, built on Intel's 18A-P process node with an expanded set of low-power and high-performance devices (including a dual-contact PowerBoost option and tighter skew corners).
Memory & I/O
Sixteen memory channels feed the platform with up to 8000 MT/s DDR5 and up to 12800 MT/s MRDIMM (Gen 2 MRDIMM), delivering 1.6 TB/s of memory bandwidth. I/O spans 128 lanes of PCIe Gen6, CXL 3.0, and UPI 3, plus 8 lanes of PCIe Gen4. Each Fabric Hub carries four x16 ports (each configurable as PCIe Gen6, CXL 3, or UPI 3), up to 16 MB of I/O cache, on-die snoop filtering, and two accelerator complexes housing QAT, DSA, and IAA.
ISA & Acceleration
Diamond Rapids adds spill-and-fill optimization instructions to the ISA — 16 new general-purpose registers (32 integer registers total) and three-operand instructions that remain compatible with existing x86 software. Matrix-multiplication acceleration rides on AMX and AVX 10.2 with expanded vector support, and coherency moves on-die with a snoop filter replacing DRAM-resident directory state for lower latency and reduced coherence traffic.
QS Compute supplies Intel Xeon 7 Diamond Rapids platforms and processors for high-core-count, memory-bandwidth-bound AI and HPC servers.
Related Products
- Arm AGI — Arm's First In-House CPU
- Fujitsu Monaka — Arm Server CPU
- AMD EPYC 9006 Venice
- NVIDIA Vera CPU
Request a Quote — Intel Xeon 7 Diamond Rapids
QS Compute — Global B2B AI computing hardware supply, 15-day sample lead time.
Get Your Quote →