LaunchedXeon 6 (Sierra Forest)

Intel · Xeon 6700E

Intel Xeon 6740E

96 E-cores in a 250W envelope for dense, efficient cloud and networking compute.

Cloud HostingContainerized MicroservicesNetwork Function VirtualizationScale-Out StorageVirtualization Infrastructure

Cores / Threads

96/ 96

Base / Boost

2.4/ 3.2 GHz

PCIe Lanes

88

L3 Cache

96MB

TDP

250W

Socket

FCLGA4710

Verdict

8.2/ 10

82

Quick Verdict

A strong choice for operators needing huge core counts and high memory bandwidth within a 250W TDP, but its E-core design and modest clocks make it less suited for legacy single-threaded or floating-point-heavy HPC codes.

Best for:Cloud HostingContainerized MicroservicesNetwork Function VirtualizationScale-Out StorageVirtualization Infrastructure

Overview

Launch

2024

Status

Launched

Generation

Xeon 6 (Sierra Forest)

Market

Cloud / Scale-Out Server

About this CPU

The Intel Xeon 6740E is a 96-core, 96-thread E-core-only server processor in Intel’s Xeon 6 (Sierra Forest-SP) family, focused on high core density and performance per watt for cloud, scale-out, and networking workloads rather than maximum per-core performance.

Intel’s Xeon 6740E sits in the middle of the 6700E Sierra Forest lineup, combining 96 Crestmont E-cores, 96 MB of L3 cache, eight-channel DDR5-6400, and 88 PCIe 5.0 lanes in a 250W LGA4710 package. It targets cloud and networking deployments that benefit from many threads and high memory bandwidth rather than ultra-high per-core performance, offering a compelling alternative to both older Xeon generations and AMD’s Bergamo EPYC CPUs in density-optimized environments.

Specifications

ArchitectureSierra Forest-SP (Crestmont E-cores)
Manufacturing ProcessIntel 3 (5 nm-class CPU compute node, 10 nm-class I/O node)
Cores / Threads96 / 96
Base Clock2.4 GHz
Boost Clock3.2 GHz
L3 Cache96 MB
TDP250 W
SocketFCLGA4710
Memory TypeDDR5
Memory SpeedDDR5-6400
Memory ChannelsOcta-Channel (8)
Max Memory4096 GB
PCIe Version / LanesPCIe 5.0 × 88
Integrated GraphicsNone
Octa-Channel88 PCIe Lanes
Target Audience
GamersStreamersContent CreatorsDevelopersWorkstation UsersOffice UsersStudents

Performance

Productivity
88Very Good

Very strong for parallel integer and data-centric workloads (key-value stores, web services, logging), but limited by per-core performance for legacy or lightly threaded tasks.

Virtualization
92Excellent

Excellent for consolidating many small VMs or containers; 96 threads and high memory bandwidth allow dense packing without heavy overcommit.

Gaming
40Limited

Not targeted at gaming; modest single-core clocks and E-core architecture result in significantly lower gaming performance than mainstream desktop or P-core Xeons.

Efficiency
86Very Good

Strong performance per watt for scale-out workloads compared to older Xeons, though still a 250W part requiring robust cooling and power delivery.

GamingPoor
  • Low base and boost clocks (2.4–3.2 GHz) and E-core IPC
  • Not intended for client or gaming workloads
  • Modern desktop CPUs and P-core Xeons deliver far higher frame rates
CreatorModerate
Blender (CPU rendering)V-Ray (CPU rendering)Video Transcoding (FFmpeg, HandBrake)Batch Image/Video ProcessingCompilation / Build Farms
AI / MLModerate
  • Good for CPU-based inference where batch size can be scaled across many cores
  • No dedicated matrix or AI accelerator blocks
  • Better suited to data preprocessing and lightweight inference than heavy training
Industry Impact
Gaming
Negligible
Workstations
Low
Content Creation
Moderate
Virtualization
High

Architecture

Intel 3 (5 nm-class CPU compute node, 10 nm-class I/O node)

Process Node

Sierra Forest-SP (SRF-SP)

Codename

96C / 96T

Core Config

96 MB

L3 Cache

250 W

TDP

Architecture Overview

Sierra Forest-SP is Intel’s first Xeon generation to use only density‑optimized E‑cores (Crestmont), organized in a chiplet design with Intel 3 compute dies and separate I/O dies connected via EMIB. The 6740E uses a single compute die with 96 Crestmont E‑cores and two I/O dies, which integrate memory controllers, PCIe 5.0, and accelerators like QAT, DLB, and DSA. This architecture prioritizes high thread count and power efficiency over the large caches and high single‑thread performance of Granite Rapids P‑core parts.

CPU Design

96 Crestmont E‑cores without Hyper‑Threading; each core is optimized for throughput and area efficiency rather than single‑thread speed. The chiplet layout allows Intel to mix and match compute and I/O dies for different SKUs and to improve yield by using smaller Intel 3 compute dies alongside older I/O dies.

Memory Subsystem

Eight‑channel DDR5‑6400 with support for RDIMMs and MRDIMMs, delivering up to 409.6 GB/s of theoretical bandwidth. The memory controllers reside in the I/O dies, so the number of channels is fixed by the platform rather than by the core count.

PCIe & I/O

88 PCIe 5.0 lanes are exposed directly from the CPU, enabling direct attach of NICs, GPUs, and NVMe drives without requiring a pluggable switch. This is especially valuable in networking and storage appliances where slot count and bandwidth are critical.

Overclocking

The multiplier is locked, and the 6740E is not intended for overclocking. Intel Speed Select Technology instead provides defined profiles (e.g., Server vs Networking modes) to adjust base frequency, turbo behavior, and TDP within safe limits.

Generation Comparison
Intel Xeon Gold 6530 (Emerald Rapids, 32 P‑cores)Intel Xeon 6740E
  • Much higher core count (96 vs 32) for throughput workloads
  • Significantly better performance per watt for scale‑out services
  • Modern I/O with PCIe 5.0 and DDR5‑6400 vs PCIe 4.0 and slower DDR4 on older platforms
  • New accelerators (QAT, DLB, DSA, IAA) integrated into the I/O dies

Key Highlights

96 Crestmont E‑cores
Delivers massive thread density for cloud and container workloads, with no SMT to keep scheduling and licensing simple.
Intel 3 Process Technology
Built on Intel’s 5 nm-class compute node with chiplet packaging for improved density and efficiency compared to earlier 10 nm Xeons.
8‑Channel DDR5‑6400
Provides very high memory bandwidth (up to 409.6 GB/s) for bandwidth-bound data and network workloads.
88 PCIe 5.0 Lanes
Supports many high-speed NICs, accelerators, and NVMe drives in a single socket, reducing the need for pluggable switches.
Intel Speed Select Technology Profiles
Configurable base/turbo and TDP profiles (Server vs Networking modes) let operators tune for throughput or latency within the same SKU.
Strengths
  • 96 E‑cores and 96 threads for high VM/container density
  • Intel 3 process and chiplet design improve density and efficiency
  • 8‑channel DDR5‑6400 with up to 4 TB capacity and high bandwidth
  • 88 PCIe 5.0 lanes for NICs, accelerators, and storage
  • Integrated accelerators (QAT, DLB, DSA, IAA) offload common data path tasks
  • Speed Select Technology profiles allow tuning for server vs networking use cases
Weaknesses
  • Modest base and boost clocks limit single‑thread performance
  • E‑core IPC is lower than P‑core Granite Rapids or AMD Zen 4c
  • No SMT; thread count equals core count, which can be a disadvantage for some licensed workloads
  • 250W TDP requires robust cooling and power delivery in dense configurations
  • New platform (LGA4710) means limited installed base and potential early‑adopter risks

History

Launch Date
2024
Status
Launched
Generation
Xeon 6 (Sierra Forest)
Market
Cloud / Scale-Out Server
The Story

The Xeon 6740E launched in mid‑2024 as part of Intel’s Xeon 6 Sierra Forest generation, the first Xeon family to exclusively use density‑optimized E‑cores. Sierra Forest addressed the growing demand from cloud providers and telecom customers for higher thread counts and better performance per watt, competing directly with AMD’s EPYC Bergamo processors. The 6700E series, including 6740E, brought up to 144 E‑cores to the LGA4710 platform, with Intel emphasizing core density and efficiency over the large caches and high clocks of earlier P‑centric Xeons.

Over time, Sierra Forest positioned itself as Intel’s answer to cloud‑native and NFV workloads, with the 6740E filling a mid‑range 96‑core niche for operators that needed more cores than 64‑part 6710E could offer, but without stepping up to the 144‑core 6780E or higher‑TDP Granite Rapids parts.

Improvements over Previous Generation

  • Much higher core count (96 vs 32) for throughput workloads
  • Significantly better performance per watt for scale‑out services
  • Modern I/O with PCIe 5.0 and DDR5‑6400 vs PCIe 4.0 and slower DDR4 on older platforms
  • New accelerators (QAT, DLB, DSA, IAA) integrated into the I/O dies

Alternatives & Competitors

Intel Xeon 6731E
Same 96‑core E‑core design at lower cost and slightly lower clocks, if you don’t need the 6740E’s higher base frequency or SST profiles.
Intel Xeon 6780E
Better when you need more cores (144) and can tolerate higher TDP (330W) for throughput‑bound workloads.
AMD EPYC 9734
Higher boost clocks and more threads (112C/224T) with Bergamo’s Zen 4c cores; stronger if your workloads benefit from SMT and higher per‑thread performance.
AMD EPYC 9654
P‑core EPYC with 96 Zen 4 cores and 192 threads; better for mixed workloads that need both strong single‑thread and multi‑thread performance.
Intel Xeon 6700P/6500P (P‑core)
Better per‑core performance and larger caches if your applications are latency‑sensitive or licensed per core rather than per thread.
Direct Competitors
AMD EPYC 9734Intel Xeon 6731EIntel Xeon 6780EIntel Xeon 6952PAMD EPYC 9654

Should You Buy It?

Recommended for the right buyer

Building or upgrading dense cloud or NFV infrastructure where you need many threads and high memory bandwidth per rack unit, and your software scales well across many E‑cores.

Avoid if…

  • You need strong single‑thread or floating‑point performance for legacy applications
  • Your workloads are heavily GPU‑accelerated and you care more about per‑core CPU performance than core count
  • You want a future‑upgrade path within the same platform (LGA4710) and are sensitive to platform longevity

Use Cases

Cloud Hosting / VM Farms
Excellent
Containerized Microservices
Excellent
Network Function Virtualization (NFV)
Very Good
Distributed Storage / Ceph / Object Stores
Very Good
HPC Throughput Workloads (Weather, Finance Kernels)
Good

Interesting Facts

The 6740E is part of Intel’s first Xeon generation to use only E‑cores, branded Xeon 6700E ‘Sierra Forest’.

Sierra Forest launched on June 4, 2024, as the first Xeon 6 generation, targeting cloud and density‑optimized servers.

In SPEC CPU2017 Integer Rate tests, a dual‑socket 6740E system achieved over 100k SPECrate2017_int_base, underscoring its throughput capability.

The 6740E supports Intel Speed Select Technology Performance Profiles, allowing the same CPU to run in Server Mode (2.4 GHz base) or Networking Mode (3.2 GHz base) with different TDP configurations.

Despite having 96 cores, the 6740E does not support Hyper‑Threading; Intel chose 1 thread per core to simplify scheduling and licensing for cloud providers.

The LGA4710 platform used by 6740E also supports P‑core Xeon 6700P/6500P CPUs, letting vendors reuse the same motherboard for both E‑core and P‑core SKUs.

Geekbench 6 reports a single‑core score around 1225 and a multi‑core score near 13600 for a dual‑socket 6740E system, showing strong multi‑thread but modest single‑thread performance.

The 6740E’s maximum memory size is officially 4 TB, eight channels of DDR5‑6400, with ECC support for data integrity in mission‑critical environments.

Intel markets the 6700E series as optimized for ‘cloud‑native’ workloads, directly competing with AMD’s EPYC Bergamo (Zen 4c) CPUs.

The E‑core design focuses on performance per watt and core density, trading away the large caches and high clock speeds typical of P‑core Xeons.

People Also Ask

What is the Intel Xeon 6740E designed for?

It is designed for cloud, NFV, and scale‑out server workloads where high core count, memory bandwidth, and performance per watt are more important than peak single‑thread speed.

How does Xeon 6740E compare to Xeon 6731E?

Both are 96‑core Sierra Forest E‑core CPUs, but 6740E has a higher base clock (2.4 vs 2.2 GHz) and supports Speed Select Technology profiles, while 6731E is limited to 1S only and lower turbo (3.1 GHz).

Does Xeon 6740E support Hyper‑Threading?

No, it uses 96 E‑cores without Hyper‑Threading, so there is 1 thread per core. This simplifies scheduling and licensing for cloud providers but reduces throughput for some threaded workloads.

How much memory does Xeon 6740E support?

Up to 4 TB of DDR5‑6400 ECC memory across eight channels, depending on DIMM type and population rules.

What socket does Xeon 6740E use?

It uses the LGA4710 (FCLGA4710) socket, which also supports Xeon 6700P/6500P P‑core CPUs on the same platform.

Is Xeon 6740E good for gaming?

No. Its E‑core design and modest clocks make it poorly suited for gaming; desktop‑class CPUs or P‑core Xeons offer much better gaming performance.

What is the TDP of Xeon 6740E?

The default TDP is 250W, with configurable profiles via Intel Speed Select Technology to adjust base frequency and turbo behavior within that envelope.

Does Xeon 6740E have integrated graphics?

No, it has no integrated GPU. Graphics output, if needed, must come from a discrete PCIe GPU or BMC.

What accelerators are included in Xeon 6740E?

It integrates Intel QuickAssist Technology (QAT), Dynamic Load Balancer (DLB), Data Streaming Accelerator (DSA), and In‑Memory Analytics Accelerator (IAA) in the I/O dies.

How does Xeon 6740E compare to EPYC 9734?

EPYC 9734 offers 112 Zen 4c cores with SMT (224 threads), higher boost clocks, and more L3 cache, generally delivering higher throughput, but at higher TDP and often higher platform cost.

Frequently Asked Questions

Is the Intel Xeon 6740E still a good choice in 2026?

Yes for density‑optimized cloud and networking workloads where you need many threads and high memory bandwidth; less ideal if you require strong single‑thread or floating‑point performance.

Can Xeon 6740E be used in a workstation?

Technically yes, but it is optimized for server platforms and E‑core throughput. Most workstations are better served by P‑core Xeons or desktop‑class CPUs with higher per‑core performance.

What cooling is recommended for Xeon 6740E?

Heavy‑duty air or liquid cooling designed for 250W+ TDP CPUs; in dense 1U/2U servers, high‑flow fans and large heatsinks are typically required.

Does Xeon 6740E support CXL?

Sierra Forest as a generation supports CXL 2.0, but Intel’s official 6740E spec page does not explicitly list CXL; CXL availability may depend on the platform and BIOS.

What BIOS version is needed for Xeon 6740E?

BIOS support depends on the server vendor; generally, you need a BIOS that specifically lists support for Xeon 6700E Sierra Forest CPUs on LGA4710.

Can Xeon 6740E be overclocked?

No. The multiplier is locked, and the CPU is designed for stable, managed operation in server environments, not for overclocking.

What memory speeds does Xeon 6740E support?

Officially DDR5‑6400; actual speeds depend on DIMM type (RDIMM/MRDIMM), population, and motherboard design.

How does Xeon 6740E handle power management?

It supports Intel Speed Select Technology profiles and power management features to balance performance and power, with configurable base/turbo and TDP settings for different operating modes.

Is Xeon 6740E suitable for AI workloads?

It can handle CPU‑based inference and data preprocessing, but it has no dedicated AI accelerators; for serious training, GPUs or specialized accelerators are still recommended.

What operating systems support Xeon 6740E?

Major server OSes (Windows Server, Linux distributions such as RHEL, Ubuntu, SUSE) support Sierra Forest, but you should verify vendor‑specific support and drivers for your platform.