INTEL® XEON® 600 PROCESSORS FOR WORKSTATIONS

A full architecture overhaul. Xeon 600 brings Granite Rapids to the workstation for the first time — up to 86 cores, native FP16 AI acceleration, and DDR5 memory speeds that double the previous generation. This isn't an incremental refresh. Build a turn-key agentic AI workstation with local, private compute — configured and ready from day one.

MAX CORE COUNT

86 CORES

↑ 43% vs. prior gen (60)

DDR5 MRDIMM SPEED

8,000 MT/s

↑ 67% vs. prior gen (4,800)

PCIe 5.0

128 LANES

NATIVE AI ACCELERATION

FP16 AMX

New — PyTorch/TF native

WHAT'S NEW IN XEON 600

NOT A REFRESH — A NEW ARCHITECTURE

Our platforms support every phase — from early experimentation on a
workstation to production-grade training and inference on GPU servers.

GRANITE RAPIDS ARCHITECTURE (REDWOOD COVE+)

First time the server-class Granite Rapids die has been brought to single-socket workstations. Intel 3 process enables higher core density with better power efficiency. Up to 9% single-thread and 61% multi-thread improvement vs. prior generation.

FP16 IN INTEL AMX

Previous Xeon workstation chips only supported BF16 and INT8 in AMX. Xeon 600 adds native FP16 — the default datatype in PyTorch and TensorFlow — eliminating type conversion overhead. Intel claims 17% faster AI workloads from this alone.

DDR5 MRDIMM AT 8,000 MT/s

Previous gen topped out at 4,800 MT/s with standard RDIMMs. Xeon 600 adds support for Multiplexed Rank DIMMs (MRDIMM) at 8,000 MT/s — a 67% increase in memory bandwidth. Standard RDIMMs now run at 6,400 MT/s. Max capacity remains 4TB.

NEW W890 CHIPSET & LGA4710 SOCKET

Not a drop-in upgrade — Xeon 600 requires new W890 motherboards on the LGA4710 socket. The platform also adds native CXL 2.0 support for memory expansion and USB4 connectivity. The previous W-3500/W-2500 line used W790 on LGA4677.

FEATURED SYSTEM

BUILT ON INTEL XEON 600

The FlexPrime V80X and R80X bring the full Xeon 600 platform to your choice of form factor.
Configure either as a turn-key agentic AI workstation — local, private, and ready from day one.

FlexPrime V80X

FORM FACTOR: Full Tower
CPU: 1x Intel Xeon 600 Workstation Series
GPU: Up to 4x Double-Wide GPU
MEMORY: Up to 8x DDR5 ECC/R

CONFIGURE REQUEST QUOTE

FlexPrime R80X

FORM FACTOR: Full Tower / Rackmount Convertible
CPU: 1x Intel Xeon 600 Workstation Series
GPU: Up to 4x Double-Wide GPU
MEMORY: Up to 8x DDR5 ECC/R

CONFIGURE REQUEST QUOTE

AI ACCELERATION

NATIVE FP16 — FORMAT YOUR FRAMEWORKS ACTUALLY USE

Intel AMX on Xeon 600 now handles FP16 matrix operations natively, alongside BF16 and INT8. This matters because FP16 is the default precision for most PyTorch and TensorFlow inference workloads — previous Xeon chips required type conversion, which added overhead.

Real-world results: Topaz Labs Video Upscaler runs 29% faster, and SPECworkstation 4.0 AI benchmarks show a 17% improvement over the W9-3595X. Combined with multi-GPU PCIe 5.0 bandwidth, this platform handles both the CPU-bound and GPU-offloaded portions of modern AI pipelines.

MEMORY

8-CHANNEL DDR5 WITH MRDIMM SUPPORT

Xeon 600 supports octa-channel DDR5 across all X-series SKUs. Standard RDIMMs now run at 6,400 MT/s — already a 33% increase over the previous 4,800 MT/s ceiling. But the real upgrade is MRDIMM support at 8,000 MT/s, which Threadripper doesn't yet offer.

This directly impacts memory-bound workloads like fluid dynamics simulation, large-dataset manipulation, and AI training data preprocessing. ECC remains standard across the lineup for data integrity during multi-day jobs.

EXPANDABILITY

PCIe 5.0 LANES FOR MULTI-GPU WORKLOADS

The Xeon 600 platform delivers 128 lanes of PCIe 5.0, giving motherboard designers room for multiple x16 GPU slots alongside NVMe storage and high-speed networking — all at full Gen 5 bandwidth. Depending on your board, that can mean up to 7x PCIe 5.0 x16 physical slots.

For AI training, rendering, and simulation workloads that scale with GPU count, this I/O budget means you're not bottlenecked at the bus level when running multi-GPU configurations.

PROCESSOR LINEUP

INTEL XEON 600 SERIES FOR WORKSTATIONS

11 CPU SKUs to choose from 12 to 86 cores. All on LGA4710, all Q1 2026. X-series
models are overclocking-unlocked.

PROCESSOR	CORES	BASE CLOCK	MAX TURBO	CACHE	TDP
PROCESSOR Xeon 698X OC	CORES 86	BASE CLOCK 2.0 GHz	MAX TURBO 4.8 GHz	CACHE 4.8 GHz	TDP 350W
PROCESSOR Xeon 696X OC	CORES 64	BASE CLOCK 2.4 GHz	MAX TURBO 4.8 GHz	CACHE 4.8 GHz	TDP 350W
PROCESSOR Xeon 678X OC	CORES 48	BASE CLOCK 2.4 GHz	MAX TURBO 4.9 GHz	CACHE 4.9 GHz	TDP 300W
PROCESSOR Xeon 676X OC	CORES 32	BASE CLOCK 2.8 GHz	MAX TURBO 4.9 GHz	CACHE 4.9 GHz	TDP 275W
PROCESSOR Xeon 674X OC	CORES 28	BASE CLOCK 3.0 GHz	MAX TURBO 4.9 GHz	CACHE 4.9 GHz	TDP 270W
PROCESSOR Xeon 658X OC	CORES 24	BASE CLOCK 3.0 GHz	MAX TURBO 4.9 GHz	CACHE 4.9 GHz	TDP 250W
PROCESSOR Xeon 656	CORES 20	BASE CLOCK 2.9 GHz	MAX TURBO 4.8 GHz	CACHE 4.8 GHz	TDP 210W
PROCESSOR Xeon 654	CORES 18	BASE CLOCK 3.1 GHz	MAX TURBO 4.8 GHz	CACHE 4.8 GHz	TDP 200W
PROCESSOR Xeon 638	CORES 16	BASE CLOCK 3.2 GHz	MAX TURBO 4.8 GHz	CACHE 4.8 GHz	TDP 180W
PROCESSOR Xeon 636	CORES 12	BASE CLOCK 3.5 GHz	MAX TURBO 4.7 GHz	CACHE 4.7 GHz	TDP 170W
PROCESSOR Xeon 634	CORES 12	BASE CLOCK 2.7 GHz	MAX TURBO 4.6 GHz	CACHE 4.6 GHz	TDP 150W