LaunchedXeon 6 (Granite Rapids-SP)

Intel · Xeon 6700P

Intel Xeon 6781P

80 P-cores, 160 threads and 136 PCIe 5.0 lanes for dense, AI‑accelerated single‑socket servers and workstations.

AI Inference & TrainingVirtualization & CloudIn-Memory DatabasesHigh-Performance ComputingStorage & HCI

Cores / Threads

80/ 160

Base / Boost

2/ 3.8 GHz

PCIe Lanes

136

L3 Cache

336MB

TDP

350W

Socket

FCLGA4710

Verdict

8.7/ 10

87

Quick Verdict

A very strong single-socket server CPU with excellent core count, memory bandwidth, and integrated AI accelerators, best suited for AI, virtualization, and data-center workloads where its 350 W TDP and platform cost are justified.

Best for:AI Inference & TrainingVirtualization & CloudIn-Memory DatabasesHigh-Performance ComputingStorage & HCI

Overview

Launch

2025

Status

Launched

Generation

Xeon 6 (Granite Rapids-SP)

Market

Server / Workstation

About this CPU

The Intel Xeon 6781P is an 80-core, 160-thread server and workstation processor based on Intel’s Granite Rapids-SP architecture, built on the Intel 3 process and targeting high-core-count, AI-accelerated workloads in single-socket platforms.

Intel’s Xeon 6781P brings 80 performance cores and 160 threads to the LGA4710 platform, with a 2.0 GHz base and up to 3.8 GHz turbo, 336 MB of L3 cache, 8-channel DDR5/MRDIMM memory up to 4 TB, and 136 PCIe 5.

0 lanes. It includes Intel AMX for AI acceleration, QuickAssist, DLB, DSA and IAA engines, and is tuned for single-socket servers and workstations running AI, virtualization, databases and storage workloads. It is not a consumer or gaming CPU, and its 350 W TDP requires robust cooling and power delivery.

Specifications

ArchitectureGranite Rapids-SP (P-cores only)
Manufacturing ProcessIntel 3
Cores / Threads80 / 160
Base Clock2 GHz
Boost Clock3.8 GHz
L3 Cache336 MB
TDP350 W
SocketFCLGA4710
Memory TypeDDR5, MRDIMM
Memory SpeedUp to 8000 MT/s (DDR5-6400 / MRDIMM-8800)
Memory ChannelsOcta-Channel (8)
Max Memory4096 GB
PCIe Version / LanesPCIe 5.0 × 136
Integrated GraphicsNone
Octa-Channel136 PCIe Lanes
Target Audience
GamersStreamersContent CreatorsDevelopersWorkstation UsersOffice UsersStudents

Performance

Productivity
N/A

Very strong for parallel server and workstation workloads; real-world performance depends heavily on memory and I/O configuration.

Virtualization
N/A

Excellent for VM density and I/O-heavy virtualization, especially in single-socket platforms with 8 memory channels and many PCIe lanes.

Gaming
N/A

Not intended for gaming; no integrated graphics and limited per-clock gaming relevance compared to consumer CPUs.

Efficiency
N/A

Competitive performance per watt for its core count at 350 W TDP, but total platform power is high, so efficiency matters only in dense deployments.

GamingNot applicable
  • No integrated graphics; requires a discrete GPU.
  • Optimized for server and AI workloads, not gaming clock rates or latency.
  • Gamers should choose mainstream desktop or workstation CPUs instead.
CreatorVery Good
Blender (CPU rendering)V-Ray / Arnold (CPU rendering)FFmpeg / video transcoding (with QAT/DLB)Scientific simulationsIn-memory analytics
AI / MLExcellent
  • Intel AMX (Advanced Matrix Extensions) accelerates INT8 and BF16 matrix operations.
  • Good for CPU-based inference and training where GPUs are not available.
  • For large-scale training, GPUs or dedicated accelerators still dominate.
Industry Impact
Gaming
Minimal
Workstations
High
Content Creation
Moderate
Virtualization
High

Architecture

Intel 3

Process Node

Granite Rapids-SP

Codename

80C / 160T

Core Config

336 MB

L3 Cache

350 W

TDP

Architecture Overview

Granite Rapids-SP is the performance-core (P-core) variant of Intel’s Xeon 6 family, designed for compute-intensive and AI workloads, as opposed to the E-core-focused Sierra Forest line.

CPU Design

The Xeon 6781P has 80 P-cores with Hyper-Threading, yielding 160 threads. A subset of cores can be designated as high-priority (30 cores at 2.10 GHz base) and low-priority (50 cores at 1.60 GHz base) using Intel Speed Select Technology (SST-PP and SST-BF), allowing different frequency/power points for mixed workloads.

Memory Subsystem

An 8-channel integrated memory controller supports DDR5-6400 and MRDIMM-8800, with up to 4 TB capacity. MRDIMM multiplexes ranks to increase effective bandwidth, which benefits memory-bound AI and HPC workloads.

PCIe & I/O

136 PCIe 5.0 lanes are routed from the CPU, enabling multiple GPUs, NVMe SSDs, and CXL/SmartNIC devices. The PCIe controllers can also run CXL 2.0 for coherent memory expansion and pooling.

Overclocking

The multiplier is locked; Xeon 6781P is not an overclocking part. Tuning is done via SST-PP profiles and power limits rather than BCLK or multiplier adjustment.

Generation Comparison
Intel Xeon w9-3495X (Sapphire Rapids HEDT)Intel Xeon 6781P
  • Much higher core count (80 vs 24–36) and threads (160 vs 48–72).
  • Intel 3 process and higher IPC plus AMX for AI workloads.
  • 8-channel DDR5/MRDIMM and 136 PCIe 5.0 lanes vs 4-channel DDR5 and fewer PCIe lanes on older HEDT.
  • Integrated accelerators (QAT, DLB, DSA, IAA) not present in older HEDT CPUs.

Key Highlights

80 Performance Cores, 160 Threads
High core density aimed at parallel AI, virtualization, and database workloads rather than E-core throughput.
8-Channel DDR5 + MRDIMM
Supports DDR5-6400 and MRDIMM-8800 with up to 4 TB capacity, delivering very high memory bandwidth for in-memory workloads.
136 PCIe 5.0 Lanes
Provides massive I/O bandwidth for GPU accelerators, NVMe storage, and SmartNICs in a single-socket system.
Integrated AI & Accelerator Engines
Includes Intel AMX for matrix math, plus QuickAssist (QAT), DLB, DSA and IAA accelerators for AI, analytics and networking.
High-Priority / Low-Priority Core Split
30 high-priority cores at higher base frequency and 50 low-priority cores at lower frequency, with SST-PP profiles for tuning.
Strengths
  • 80 cores and 160 threads for highly parallel workloads.
  • 8-channel DDR5/MRDIMM with up to 4 TB memory capacity.
  • 136 PCIe 5.0 lanes for GPUs, NVMe, and CXL devices.
  • Intel AMX, QAT, DLB, DSA, IAA accelerators for AI and I/O.
  • Strong single-socket performance for virtualization and databases.
  • CXL 2.0 support on the Xeon 6 platform for memory expansion.
Weaknesses
  • High 350 W TDP and demanding power/cooling requirements.
  • Premium pricing; overkill for SMB or light server workloads.
  • No integrated graphics; not suitable for basic desktop use.
  • Locked multiplier; no enthusiast overclocking.
  • Platform is new and may have early BIOS/firmware maturity considerations.

History

Launch Date
2025
Status
Launched
Generation
Xeon 6 (Granite Rapids-SP)
Market
Server / Workstation
The Story

The Xeon 6781P emerged as part of Intel’s Granite Rapids-SP rollout in 2025, which introduced the performance-core-only Xeon 6700P series to complement the earlier Sierra Forest E-core parts. These CPUs were positioned to regain competitiveness against AMD’s EPYC 9004/9005 lines in enterprise and AI workloads by offering higher per-core performance and new AI accelerators, while sharing a common platform with Sierra Forest.</br>Intel announced the Xeon 6 family in 2024–2025, with Granite Rapids-SP focusing on P-cores for AI and HPC, and Sierra Forest on E-cores for cloud-native density.

The 6781P specifically targets single-socket servers and workstations needing very high core counts, large memory, and many I/O lanes without moving to a dual-socket design.</br>Early listings and SPEC submissions from vendors like HPE, Cisco and Supermicro showed the 6781P running at 2.0/3.

8 GHz with 336 MB cache and 350 W TDP, confirming it as one of the flagship 80-core Granite Rapids-SP parts for AI-optimized servers.

Improvements over Previous Generation

  • Much higher core count (80 vs 24–36) and threads (160 vs 48–72).
  • Intel 3 process and higher IPC plus AMX for AI workloads.
  • 8-channel DDR5/MRDIMM and 136 PCIe 5.0 lanes vs 4-channel DDR5 and fewer PCIe lanes on older HEDT.
  • Integrated accelerators (QAT, DLB, DSA, IAA) not present in older HEDT CPUs.

Alternatives & Competitors

AMD EPYC 9565
72 Zen 5 cores with 12 memory channels and higher per-clock performance, often with better performance per dollar in many server workloads.
Intel Xeon 6980P
128 P-cores, 504 MB cache, 12 memory channels and 2S scalability if you need more headroom than 6781P.
Intel Xeon 6780E
144 E-cores at lower TDP for cloud-scale, task-parallel workloads where core density and efficiency matter more than single-thread performance.
AMD EPYC 9255
24-core, 48-thread Turin CPU with 12 memory channels and higher clocks, better for workloads that don’t need 80 cores but benefit from strong per-core performance.
Intel Xeon w9-3495X
Older HEDT platform with fewer cores but lower platform cost and still strong for workstation-class rendering and virtualization.
Direct Competitors
AMD EPYC 9565AMD EPYC 9255Intel Xeon 6980PIntel Xeon 6780EIntel Xeon w9-3495X

Should You Buy It?

Recommended for the right buyer

Single-socket servers and workstations for AI inference, virtualization, in-memory databases, or HPC where you need many cores, high memory bandwidth, and strong AI acceleration without going dual-socket.

Avoid if…

  • Building a gaming or light-use desktop PC.
  • Your workload does not scale beyond ~32 cores or cannot use the AI accelerators.
  • You cannot accommodate the 350 W CPU TDP and associated cooling/power supply.
  • You need dual-socket scalability for future expansion.

Use Cases

AI Inference & Training
Excellent
Virtualization & Cloud Infrastructure
Excellent
In-Memory Databases & Analytics
Excellent
High-Performance Computing (HPC)
Very Good
Storage & Hyperconverged Infrastructure
Very Good

Interesting Facts

The 6781P uses Intel’s first server P-core-only Granite Rapids-SP die on the LGA4710 platform, rather than the E-core-focused Sierra Forest design.

Its 30 high-priority / 50 low-priority core split with SST-BF allows latency-sensitive workloads to run on faster cores while others idle at lower frequency.

SPEC CPU2017 results published with this CPU show it reaching 3.8 GHz turbo and 8-channel DDR5-8000 memory in real server configurations.

It supports CXL 2.0, enabling memory expansion and pooling when paired with CXL type-3 memory modules.

Intel positions the 6700P series as the “performance core” line for enterprise and AI, complementing the E-core 6700E/6500E parts for cloud-native workloads.

With 136 PCIe 5.0 lanes, a single 6781P can connect more high-speed devices than many dual-socket previous-gen platforms.

MRDIMM support allows effective memory speeds up to 8800 MT/s, significantly increasing bandwidth for AI and HPC versus DDR5-6400 alone.

The 6781P is a 1S-only SKU; Intel offers 2S-capable 6900P/6980P parts for dual-socket servers.

Intel’s product brief highlights the 6700P series as optimized for general-purpose, AI and telco workloads, with up to 136 PCIe lanes and 8 memory channels per socket.

Early SPEC submissions using the 6781P demonstrate its use in major OEM servers from HPE, Cisco, Supermicro and others.

People Also Ask

Is Intel Xeon 6781P good for gaming?

It can run games with a discrete GPU, but it is not optimized for gaming. Its value is in server and AI workloads, not gaming frame rates or latency.

What socket does Intel Xeon 6781P use?

It uses the Intel FCLGA4710 socket (also called LGA4710), part of the Birch Stream platform for Xeon 6 Granite Rapids-SP and Sierra Forest CPUs.

How much memory does Intel Xeon 6781P support?

Up to 4 TB of DDR5/MRDIMM memory across 8 channels, with support for DDR5-6400 and MRDIMM-8800 speeds.

Does Intel Xeon 6781P have integrated graphics?

No, it does not have integrated graphics. A discrete GPU is required for any display output.

What is the difference between Intel Xeon 6781P and 6780E?

The 6781P has 80 P-cores (performance cores) with higher per-core performance and AI accelerators, while the 6780E has 144 E-cores (efficiency cores) optimized for high-density, task-parallel cloud workloads at lower power.

Does Intel Xeon 6781P support CXL?

Yes, the Xeon 6 platform supports CXL 2.0, so with a compatible motherboard and CXL devices, the 6781P can use CXL memory expansion and pooling.

Is Intel Xeon 6781P unlocked for overclocking?

No, the multiplier is locked. Overclocking is not supported; performance tuning is done via power limits and Intel Speed Select Technology profiles.

What AI features does Intel Xeon 6781P have?

It includes Intel AMX for matrix math, plus Intel QuickAssist Technology (QAT), Dynamic Load Balancer (DLB), Data Streaming Accelerator (DSA), and In-memory Analytics Accelerator (IAA) for AI, analytics and networking workloads.

How many PCIe lanes does Intel Xeon 6781P have?

It provides 136 PCIe 5.0 lanes from the CPU, suitable for multiple GPUs, NVMe SSDs, and SmartNICs or CXL adapters.

What process node is Intel Xeon 6781P built on?

Intel lists the lithography as Intel 3 for the 6781P in its official specifications, which is Intel’s 3rd-generation 10 nm class process for high-performance server CPUs.

Frequently Asked Questions

What is Intel Xeon 6781P?

It is an 80-core, 160-thread server and workstation processor from Intel’s Xeon 6 Granite Rapids-SP family, built on the Intel 3 process and targeting AI, virtualization, and high-performance computing workloads.

What are the key specifications of Intel Xeon 6781P?

80 cores, 160 threads, 2.0 GHz base, 3.8 GHz max turbo, 336 MB L3 cache, 350 W TDP, FCLGA4710 socket, 8-channel DDR5/MRDIMM up to 4 TB, 136 PCIe 5.0 lanes, and Intel AMX/QAT/DLB/DSA/IAA accelerators.

Is Intel Xeon 6781P suitable for gaming?

Not really. It has no integrated graphics and is optimized for server and AI workloads. Gamers should choose a mainstream desktop or workstation CPU instead.

What memory types does Intel Xeon 6781P support?

It supports DDR5-6400 and MRDIMM-8800 memory modules with ECC, up to 4 TB across 8 channels.

Can Intel Xeon 6781P be used in dual-socket systems?

No, the 6781P is a 1S (single-socket) SKU. For dual-socket Granite Rapids-SP servers, look at the Xeon 6900P/6980P series.

Does Intel Xeon 6781P have integrated AI acceleration?

Yes. It includes Intel AMX for low-precision matrix operations and dedicated engines for QuickAssist, DLB, DSA and IAA to accelerate AI, analytics and networking workloads.

What cooling does Intel Xeon 6781P need?

With a 350 W TDP, it requires heavy-duty air or liquid cooling designed for server sockets; typical 1U/2U coolers or high-end workstation liquid coolers are appropriate depending on the chassis.

What is the difference between high-priority and low-priority cores on Intel Xeon 6781P?

30 cores are designated high-priority with a higher base frequency (2.10 GHz), while 50 are low-priority with a lower base frequency (1.60 GHz). Intel Speed Select Technology lets you choose profiles that emphasize either more cores at lower frequencies or fewer cores at higher frequencies.

Is Intel Xeon 6781P energy efficient?

For its 80-core performance level, it is reasonably efficient, but at 350 W TDP the absolute power draw is high. Efficiency matters most when the cores are heavily utilized; at low utilization, low-power modes help reduce consumption.

What chipsets and motherboards work with Intel Xeon 6781P?

It uses the Intel Birch Stream platform with the C740/C740Z chipsets and LGA4710 socket. You must choose a server or workstation motherboard that explicitly supports Xeon 6700P/6500P Granite Rapids-SP CPUs.