Quick Verdict
A strong 32-core server CPU with excellent memory bandwidth and built-in AI acceleration, best suited for dual-socket enterprise and AI inference platforms where per-core performance matters more than raw core count.
Overview
Launch
2025
Status
LaunchedGeneration
6th Gen Xeon Scalable (Granite Rapids-SP)
Market
2S Server / Workstation
The Intel Xeon 6732P is a 32-core, 64-thread server processor based on the Granite Rapids-SP architecture, featuring a 3.8 GHz base clock, up to 4.3 GHz turbo, 144 MB of L3 cache per socket, and 8-channel DDR5-6400 support, targeted at virtualized, database, AI inference, and general-purpose enterprise workloads in dual-socket platforms.
With 32 P-cores running at 3.8–4.3 GHz, 144 MB of L3 cache per socket, and 8-channel DDR5-6400, the Xeon 6732P delivers strong per-core throughput and memory bandwidth for databases, virtualization, and AI inference.
It includes Intel AMX and AVX-512 for matrix and vector workloads and supports dual-socket configurations with 88 PCIe 5.0 lanes, but it draws up to 350 W and is not intended for power-sensitive or single-threaded-only environments.
Specifications
Performance
Strong multi-threaded and per-core performance for database, ERP, and virtualized enterprise workloads.
Excellent VM density and I/O bandwidth in dual-socket configs; NUMA tuning recommended.
Not targeted at gaming; adequate for light or GPU-bound game server workloads but not competitive with modern desktop CPUs for high-refresh-rate gaming.
High performance per socket but high power draw; efficiency improves with consolidation of older servers.
- •No integrated graphics; requires discrete GPU.
- •High single-thread clocks help some game servers, but platform is not optimized for gaming.
- •GPU-bound game servers may still run well depending on title and configuration.
- •Intel AMX (BF16/INT8) and AVX-512 accelerate CPU-based inference.
- •Well suited for small to medium LLMs, embedding models, and classic ML.
- •Not a replacement for dedicated accelerators for large-scale training.
Architecture
Intel 3
Process Node
Granite Rapids-SP
Codename
32C / 64T
Core Config
144 MB
L3 Cache
350 W
TDP
Architecture Overview
Granite Rapids-SP is Intel’s 6th-generation Xeon Scalable server platform, using chiplet-based Redwood Cove P-cores on Intel 3 with integrated memory controllers and accelerators.
CPU Design
32 Redwood Cove P-cores with 2 MB L2 cache per core and 144 MB shared L3 per socket, connected via a mesh interconnect. Each core supports Hyper-Threading for 64 threads.
Memory Subsystem
8 DDR5 channels per CPU with support for RDIMM and MRDIMM at up to 6400 MT/s, delivering high bandwidth for memory-intensive workloads.
PCIe & I/O
88 PCIe 5.0 lanes from the CPU for GPUs, NVMe, and networking, with configurable bifurcation for flexible I/O designs.
Overclocking
No unlocked multiplier; frequency is managed by Intel Turbo Boost and Speed Select Technology profiles.
- Move from Intel 7 to Intel 3 process node.
- Higher per-core frequency and IPC uplift from Redwood Cove.
- Integrated AMX and more advanced acceleration engines.
- Higher DDR5 speed and MRDIMM support.
Key Highlights
- 32 high-frequency P-cores with strong per-core performance.
- 8-channel DDR5-6400 with MRDIMM support for high bandwidth.
- 144 MB L3 cache per socket improves working-set performance.
- Intel AMX and AVX-512 accelerate AI and HPC on CPU.
- 88 PCIe 5.0 lanes for flexible I/O in dual-socket servers.
- Mature RAS and security features (TDX, SGX, total memory encryption).
- 350 W TDP requires robust cooling and raises power costs.
- Dual-socket NUMA topology needs OS and application tuning.
- Higher platform cost compared to previous-gen Xeons.
- No integrated graphics; not suitable for headless or light graphics workloads.
- Core count lags higher-tier SKUs like 6740P/6760P for highly parallel tasks.
History
The Xeon 6732P launched as part of Intel’s Xeon 6700P Granite Rapids-SP lineup on February 24, 2025, following years of speculation about Intel’s answer to AMD’s high-core-count EPYC processors. Granite Rapids-SP moves Intel’s server platform from the Intel 7 process used in Sapphire and Emerald Rapids to Intel 3, and adopts a chiplet design with compute dies that integrate memory controllers and accelerators. The 6700P series is positioned for dual- and multi-socket servers, filling the gap between the high-core 6900P parts and the efficiency-focused 6500P line.
The 6732P, with its 32 cores and relatively high clocks, targets workloads where per-core performance and memory bandwidth matter more than raw core count, such as virtualized databases, AI inference, and in-memory analytics. Early OEM listings from HPE, Lenovo, and Supermicro, plus SPEC CPU2017/2026 results, confirm its role as a mainstream dual-socket enterprise CPU.
Improvements over Previous Generation
- Move from Intel 7 to Intel 3 process node.
- Higher per-core frequency and IPC uplift from Redwood Cove.
- Integrated AMX and more advanced acceleration engines.
- Higher DDR5 speed and MRDIMM support.
Alternatives & Competitors
Should You Buy It?
Recommended for the right buyer
Dual-socket enterprise servers running virtualization, databases, or CPU-based AI inference where per-core performance and memory bandwidth are critical.
Avoid if…
- You need maximum core count for highly parallel workloads (consider 6760P/6740P instead).
- Your environment is very power-constrained or air-cooling only.
- You are building a gaming or desktop workstation.
Use Cases
Interesting Facts
The Xeon 6700P series is codenamed Granite Rapids-SP, the P-core-only counterpart to the E-core Sierra Forest line.
Each Granite Rapids-SP compute die integrates four DDR5 memory channels, moving controllers onto the compute dies rather than a separate IOD.
In dual-socket systems, each 6732P appears as a separate NUMA domain; Intel offers SNC3 and HEX modes to trade off latency vs. NUMA locality.
SPEC CPU2017/2026 results show 144 MB L3 per socket and 288 MB total L3 in dual-socket systems, confirming per-chip cache size.
Technical.City reports a PassMark CPU Mark around 74,849 for the 6732P, indicating strong multi-threaded throughput among server CPUs.
Intel positions Xeon 6 as providing up to 1.9x higher AI performance over 5th-gen Xeon in MLPerf inference benchmarks.
The 6732P’s base frequency is among the highest in the Xeon 6700P lineup, making it attractive for latency-sensitive workloads.
Intel’s SST-PP profiles allow trading base frequency for TDP, enabling ‘server mode’ and ‘compute mode’ configurations within the same SKU.
Supermicro’s CPU matrix lists the 6732P with 144 MB cache and 2S scalability, confirming its dual-socket focus.
HPE and Lenovo list the 6732P as a supported processor in ProLiant and ThinkSystem servers, indicating broad OEM adoption.
People Also Ask
Is the Intel Xeon 6732P good for virtualization?
Yes. Its 32 cores, 64 threads, 144 MB L3, and 8-channel DDR5-6400 make it well suited for hosting many VMs and containers in dual-socket servers.
What socket does the Xeon 6732P use?
It uses the Intel FCLGA4710 socket (also referred to as LGA4710-2), part of the Granite Rapids-SP platform.
How much memory does the Xeon 6732P support?
Up to 4 TB of DDR5 memory across 8 channels at 6400 MT/s, depending on DIMM type and configuration.
Does the Xeon 6732P have integrated graphics?
No. It requires a discrete GPU or BMC for display output.
What is the difference between Xeon 6732P and 6737P?
The 6732P has a higher 3.8 GHz base clock and 350 W TDP; the 6737P runs at 2.9 GHz base with 270 W TDP and slightly different SST-PP tuning.
Can I use Xeon 6732P in a single-socket workstation?
Yes, but you will only access half the platform’s potential PCIe lanes and memory bandwidth; a single-socket-optimized SKU may be more cost-effective.
Is Xeon 6732P built on Intel 3?
Intel ARK does not list the node, but third-party technical documentation and resellers describe Granite Rapids-SP as using Intel 3 lithography.
How does Xeon 6732P compare to EPYC 9354?
Both are 32-core/64-thread server CPUs, but the EPYC 9354 uses Zen 4 on TSMC N5 with 12-channel DDR5 and 256 MB L3, while the 6732P uses Intel 3 with 8-channel DDR5 and 144 MB L3 per socket.
What cooling does Xeon 6732P need?
It requires robust server-grade air or liquid cooling capable of handling 350 W TDP per socket in a controlled data-center environment.
Is Xeon 6732P good for AI inference?
It is capable for small to medium models and batch inference thanks to AMX and AVX-512, but large-scale training or inference is still better served by dedicated accelerators.
Frequently Asked Questions
Does the Intel Xeon 6732P support overclocking?
No. The multiplier is locked; frequency is controlled by Intel Turbo Boost and Speed Select Technology profiles.
What PCIe version does the Xeon 6732P support?
PCIe 5.0, with up to 88 lanes from the CPU in supported platforms.
How many UPI links does the Xeon 6732P have?
Four UPI links at 24 GT/s for dual-socket interconnectivity.
What memory speeds does the Xeon 6732P support?
DDR5 up to 6400 MT/s; MRDIMM speeds depend on platform and configuration.
Is the Xeon 6732P compatible with LGA4189 motherboards?
No. It requires an FCLGA4710 socket and compatible Granite Rapids-SP platform.
What is the launch date of the Xeon 6732P?
Intel launched the Xeon 6700P/6500P series on February 24, 2025, with the 6732P as part of that lineup.
Does the Xeon 6732P support Intel TDX?
Yes, Intel Trust Domain Extensions (TDX) are supported for confidential computing use cases.
Can I mix Xeon 6732P with other Xeon 6 SKUs in the same system?
Typically no; multi-socket configurations usually require matching SKUs and steppings. Check vendor validation for any exceptions.
What is the recommended customer price for the Xeon 6732P?
Intel lists a recommended customer price around $5,930 for tray processors, but actual pricing varies by channel and volume.
Is the Xeon 6732P suitable for gaming?
Not as a primary gaming CPU; it can run game servers or GPU-bound titles, but desktop CPUs offer better gaming performance and value.