Quick Verdict
A no-compromise, high-core-count server CPU tailored for HPC, AI, and dense virtualization, where its 72 P-cores, huge cache, and 12-channel DDR5/MRDIMM memory deliver substantial throughput, provided you can supply and cool 500 W per socket.
Overview
Launch
2025
Status
LaunchedGeneration
Xeon 6 (Granite Rapids-AP)
Market
Server / Data Center / HPC / AI
The Intel Xeon 6962P is a 72-core, 144-thread server processor based on the Granite Rapids-AP (Redwood Cove P-core) architecture, built on Intel 3 process technology with 432 MB of shared L3 cache and a 500 W TDP, designed for high-performance computing, AI inference, and dense virtualization in dual-socket platforms.
With 72 Redwood Cove P-cores, 144 threads, 432 MB of L3 cache, and 12-channel DDR5/MRDIMM support, the Xeon 6962P targets data center workloads that demand both high core counts and strong per-core performance. Its 500 W TDP and FCLGA7529 socket require robust power and cooling, but in return it delivers significant memory bandwidth and I/O capability via 96 PCIe 5.0 lanes and CXL 2.
0, making it well-suited for HPC, AI inference, virtualization, and in-memory databases.
Specifications
Performance
No standardized desktop productivity benchmarks available for this server SKU.
No official benchmark scores; designed for high VM density in enterprise environments.
Not intended for gaming; no relevant benchmarks available.
500 W TDP indicates low efficiency per watt compared to lower-core server parts; no official efficiency scores available.
- •Server-focused SKU with no integrated graphics or gaming-optimized firmware.
- •No official gaming benchmarks from Intel or independent labs.
- •Not a target use case for this processor.
- •72 P-cores with AMX and AVX-512 for matrix and vector workloads.
- •High memory bandwidth via 12-channel DDR5/MRDIMM benefits AI inference.
- •No official AI benchmark scores; real-world performance depends on framework and model.
Architecture
Intel 3 (compute dies) / Intel 7 (I/O die)
Process Node
Granite Rapids-AP
Codename
72C / 144T
Core Config
432 MB
L3 Cache
500 W
TDP
Architecture Overview
The Xeon 6962P uses Intel’s Granite Rapids-AP architecture, which separates compute and I/O dies and scales up Redwood Cove P-cores to very high core counts within a 500 W power budget.
CPU Design
It features 72 Redwood Cove P-cores, each with a 2 MB private L2 cache and support for two threads via Hyper-Threading, yielding 144 threads. The cores are organized across multiple compute tiles built on the Intel 3 process, connected by a high-speed mesh and sharing 432 MB of distributed L3 cache.
Memory Subsystem
The integrated memory controllers support 12 channels of DDR5-6400 or MRDIMM-8800 memory, with up to 3 TB capacity per socket. This configuration is optimized for bandwidth-hungry workloads like AI and in-memory databases, and MRDIMMs provide higher effective bandwidth than standard RDIMMs at similar capacities.
PCIe & I/O
The I/O die, fabricated on Intel 7, provides 96 PCIe 5.0 lanes per socket and supports CXL 2.0 for coherent expansion and memory disaggregation, enabling direct attachment of high-speed NICs, GPUs, and CXL memory devices while maintaining low-latency access for the CPU cores.
Overclocking
The multiplier is locked, reflecting its server focus; frequency is managed by Intel’s Turbo Boost and Turbo Boost Max Technology 3.0, which prioritize critical workloads on the fastest cores within the 500 W TDP envelope.
- Move from Ice Lake-SP to Granite Rapids-AP with Redwood Cove P-cores and Intel 3 process.
- Roughly 2x core count increase versus typical high-end Ice Lake-SP SKUs.
- Substantially larger L3 cache (432 MB vs typical 60 MB class) and 12-channel memory vs 8-channel.
- PCIe 5.0 and CXL 2.0 support instead of PCIe 4.0 and no CXL.
Key Highlights
- 72 high-performance Redwood Cove P-cores with SMT for massive throughput
- 432 MB shared L3 cache reduces memory bottlenecks in data-intensive workloads
- 12-channel DDR5/MRDIMM memory with up to 3 TB capacity and very high bandwidth
- 96 PCIe 5.0 lanes plus CXL 2.0 for flexible accelerator and storage expansion
- Dual-socket UPI support for coherent 144-core platforms
- Strong platform features (AMX, AVX-512, RAS, Intel TDX) for AI and enterprise
- 500 W TDP requires robust power delivery and cooling, increasing TCO
- FCLGA7529 platform is expensive and limited to server vendor platforms
- No integrated graphics and no client-focused use cases
- High acquisition cost typical of top-bin server SKUs
- Efficiency per watt is lower than lower-core or newer-process alternatives
History
The Intel Xeon 6962P emerged from Intel’s effort to regain leadership in high-core-count server CPUs with the Granite Rapids-AP family. Announced as part of the Xeon 6900P series, Granite Rapids-AP introduced Redwood Cove P-cores and a tiled multi-die architecture on Intel 3, scaling up to 128 cores per socket. After Intel struggled to match AMD EPYC’s core counts for several generations, the 6900P series finally brought Intel to parity and beyond in the P-core segment, with SKUs like the 6962P offering 72 cores and 432 MB of L3 cache in a 500 W envelope.
Launching in 2025, the 6962P filled the mid-to-high core-count niche in the Granite Rapids-AP stack, positioned below the 96-core 6972P and 128-core 6980P. For data centers deploying GPU-accelerated AI and HPC servers, the 6900P series, including the 6962P, represented Intel’s return to competitive high-density compute, emphasizing not just raw core count but also platform features like 12-channel memory, 96 PCIe 5.0 lanes, CXL 2.
0, and AMX for AI acceleration.
Improvements over Previous Generation
- Move from Ice Lake-SP to Granite Rapids-AP with Redwood Cove P-cores and Intel 3 process.
- Roughly 2x core count increase versus typical high-end Ice Lake-SP SKUs.
- Substantially larger L3 cache (432 MB vs typical 60 MB class) and 12-channel memory vs 8-channel.
- PCIe 5.0 and CXL 2.0 support instead of PCIe 4.0 and no CXL.
Alternatives & Competitors
Should You Buy It?
Recommended for the right buyer
New dual-socket server deployments for HPC, AI inference, or dense virtualization where 72 high-performance P-cores and 12-channel memory bandwidth are fully utilized.
Avoid if…
- Building a general-purpose office or small business server
- Seeking maximum performance per watt or low noise
- Needing only modest core counts (8–32 cores)
- Lacking robust power and cooling infrastructure for 500 W CPUs
Use Cases
Interesting Facts
The Xeon 6962P is part of Intel’s Granite Rapids-AP family, which brought Intel back to core-count parity with AMD EPYC in the high-end server segment.
Each Redwood Cove core includes a dedicated 2 MB L2 cache, giving the 6962P 144 MB of L2 alone in addition to its 432 MB L3.
Granite Rapids-AP uses a disaggregated design with Intel 3 compute dies and Intel 7 I/O dies, connected by EMIB and a high-speed fabric.
Technical City lists the 6962P as launched on April 1, 2025, aligning with Granite Rapids-AP availability in 2025.
The 6962P supports MRDIMMs up to 8800 MT/s, which offer significantly higher bandwidth than DDR5-6400 RDIMMs for bandwidth-sensitive workloads.
Intel positions the 6900P series as “accelerator-optimized,” with features like SST-TF and Priority Core Turbo for GPU-heavy AI servers.
The same Granite Rapids-AP architecture scales up to 128 cores in the Xeon 6980P flagship, sharing the same basic tile and I/O design.
The 6962P’s 96 PCIe 5.0 lanes and CXL 2.0 make it suitable for systems with multiple high-speed NICs, GPUs, or CXL memory/expansion cards.
Despite being a server part, the 6962P’s 3.9 GHz max turbo gives it competitive single-thread performance for latency-sensitive HPC and enterprise workloads.
Intel’s own benchmarks show Xeon 6900P series delivering substantial gains vs prior-gen Xeon and competitive EPYC platforms in HPC and AI workloads.
People Also Ask
What is the Intel Xeon 6962P used for?
It is designed for server workloads such as HPC simulations, AI inference and training, large-scale virtualization, and in-memory databases, where high core counts, cache, and memory bandwidth are critical.
How many cores does the Intel Xeon 6962P have?
It has 72 physical cores and 144 threads via Intel Hyper-Threading.
What socket does the Intel Xeon 6962P use?
It uses the FCLGA7529 socket, part of the Intel Xeon 6900P (Granite Rapids-AP) platform.
How much cache does the Intel Xeon 6962P have?
It has 432 MB of shared L3 cache and approximately 144 MB of total L2 cache (2 MB per core), for around 576 MB of combined cache.
What memory does the Intel Xeon 6962P support?
It supports 12 channels of DDR5-6400 or MRDIMM-8800 memory, with up to 3 TB capacity per socket.
What is the TDP of the Intel Xeon 6962P?
The thermal design power is 500 W, requiring robust power and cooling in server platforms.
Is the Intel Xeon 6962P good for gaming?
It is not intended for gaming; there are no integrated graphics or gaming-optimized features, and no official gaming benchmarks exist.
Does the Intel Xeon 6962P support PCIe 5.0?
Yes, it supports 96 PCIe 5.0 lanes per socket plus CXL 2.0 for high-speed expansion.
Can you overclock the Intel Xeon 6962P?
No, the multiplier is locked; frequency is managed by Intel Turbo Boost and Turbo Boost Max Technology 3.0 within the 500 W TDP limit.
When was the Intel Xeon 6962P launched?
It is listed as launched on April 1, 2025, according to Technical City and distributor listings, aligning with Granite Rapids-AP availability.
Frequently Asked Questions
Does the Intel Xeon 6962P have integrated graphics?
No, it has no integrated GPU; graphics output, if needed, must come from a discrete PCIe GPU.
What operating systems does the Intel Xeon 6962P support?
It supports standard x86-64 server operating systems, including Windows Server and major Linux distributions, with driver and firmware support from server vendors.
What chipsets or platforms does the Intel Xeon 6962P work with?
It is designed for Intel Xeon 6900P (Granite Rapids-AP) server platforms using the FCLGA7529 socket and Birch Stream platform architecture.
How much power does the Intel Xeon 6962P draw under load?
The rated TDP is 500 W, but actual power consumption depends on workload, turbo behavior, and platform design; high-load scenarios can approach or exceed TDP.
Can the Intel Xeon 6962P be used in a workstation?
While technically possible in vendor workstations based on the same platform, it is primarily aimed at server and HPC deployments rather than desktop workstations.
What type of memory modules does the Intel Xeon 6962P support?
It supports DDR5 RDIMMs and MRDIMMs, with speeds up to DDR5-6400 and MRDIMM-8800, depending on the memory type and platform configuration.
How does the Intel Xeon 6962P compare to the Xeon 6960P?
Both are 72-core, 144-thread Granite Rapids-AP processors with 432 MB L3 cache and 500 W TDP; differences are minor and mainly in binning or OPN rather than major architectural changes.
Is the Intel Xeon 6962P suitable for small businesses?
Only if the business has HPC, AI, or large-scale virtualization needs; for general file, print, or light application workloads, lower-core, lower-power Xeons are more appropriate.
What cooling is required for the Intel Xeon 6962P?
It requires robust server-grade cooling, typically heavy-duty air or liquid solutions specified by the server vendor, to handle 500 W TDP per socket.
Does the Intel Xeon 6962P support CXL?
Yes, the Granite Rapids-AP platform supports CXL 2.0 on the Xeon 6962P, enabling CXL memory and accelerator devices.