Intel Xeon 6761P 64c/128t 2.5GHz-3.9GHz 350W (PK8072006270400)

Intel Xeon 6761P — 64-core Granite Rapids CPU with MRDIMM support for bandwidth-critical enterprise, AI, and technical workloads

The Intel Xeon 6761P stands out inside the Granite Rapids Xeon 6 family due to its combination of high compute density, expanded memory technology support, and a performance envelope tuned for workloads that must sustain high throughput without interruptions. While structurally similar to the 6760P, this model introduces MRDIMM capability and a higher turbo ceiling, making it an especially strong choice for organizations building next-gen data engines, AI inference pipelines, and multi-tenant compute clusters with strict QoS requirements.

At its core, the 6761P delivers 64 cores, 128 threads, 2.5 GHz base frequency and up to 3.9 GHz turbo, providing a more responsive top-end compared to its non-MRDIMM sibling. This benefits real-time analytics, low-latency financial workloads, and inference services where short-duration bursts of elevated performance significantly impact user experience and SLA consistency.

A major architectural highlight is its 336 MB of L3 cache, one of the largest cache configurations within Granite Rapids. This enormous shared pool enhances stability in workloads with unpredictable memory access patterns—large graph engines, distributed object stores, simulation frameworks, columnar DBMS scans, and Python-based data science pipelines. When multiple high-pressure operations run concurrently on the same socket, this cache capacity reduces cross-service interference, improving determinism and reducing tail latency.

Memory capabilities are where the Xeon 6761P differentiates itself. Alongside standard DDR5-6400 support, it introduces MRDIMM (Multiplexed Rank DIMM) compatibility, enabling even higher memory throughput (up to 8000 MT/s depending on configuration). This matters for workloads such as vector databases, real-time recommendation engines, LLM-powered microservices, and large-scale scientific datasets where bandwidth saturation is a frequent bottleneck. With up to 4 TB of RAM per socket, memory-bound jobs can run fully in-memory, significantly reducing compute stalls and I/O pressure.

I/O performance is equally modern: 112 PCIe 5.0 lanes allow dense configurations of NVMe SSDs, SmartNICs, GPGPU accelerators, FPGA cards, and high-speed networking adapters. For AI and HPC-lite applications that depend heavily on accelerator topologies, PCIe 5.0 ensures predictable bandwidth allocation even under multi-device load.

At 350 W TDP, the Xeon 6761P is engineered for sustained high utilization in modern thermal envelopes. Server builders targeting 1U and 2U designs can rely on predictable thermals, making it viable for cloud compute nodes where thousands of machines must maintain uniform performance profiles without thermal drift.

In real-world deployments, the 6761P excels in:

• AI inference clusters supporting large numbers of concurrent queries
• High-bandwidth analytic engines (ClickHouse, BigQuery-style systems, Spark SQL)
• Memory-intensive microservices including vector search, caching layers and JVM workloads
• High-density virtualization clusters with mixed application footprints
• Compute-heavy R&D environments such as simulation and genome analysis
• Multi-GPU servers relying on PCIe 5.0 throughput

Overall, the Xeon 6761P is designed for organizations requiring a high-core, high-bandwidth CPU that remains stable under pressure and supports emerging memory innovations shaping modern computing.