Intel 5th Gen Xeon CPUs Official: Emerald Rapids Compatible With Sapphire Rapids, Up To 64 Cores, 320 MB Cache, Prices Detailed

Intel Sierra Forest & Granite Rapids-D Xeon CPUs To Accelerate 5G Networking & Edge Computing, 2.7x Faster With 288 Cores 1

Intel has officially launched its 5th Gen Xeon CPU family codenamed Emerald Rapids, bringing higher performance & efficiency with lower TCO.

As expected, Intel is launching its 5th Gen Xeon "Emerald Rapids" family today which replaces the 4th Gen Xeon "Sapphire Rapids" family. These new chips are designed for new and existing servers since they offer a drop-in upgrade path to 4th Gen customers and also offer lower TCO (Total Cost of Ownership), making them a viable solution for those looking to expand their compute & memory capabilities on the same platforms and at the same power.

In terms of what 5th Gen Xeon CPUs have to offer, Intel has stated that Emerald Rapids chips are designed to offer a 21% average performance uplift, 36% average performance per watt uplift & a 77% reduction in TCO (total cost of ownership from typical 5-year refresh cycle) versus the 4th Gen chips. We will get into more details on performance in a bit but we have to look at the specs & the SKUs.

The Intel 5th Gen Xeon "Emerald Rapids" family will be the last of the chips designed for the Eagle Stream platform before Intel splits its lineup into a P-Core and E-Core-only ecosystem similar to its competition.

The lineup relies on the Raptor Cove P-Cove architecture and is based on the Intel 7 (10nm ESF) process node. The CPUs see a moderate increase in core counts from 60 cores to 64 cores on the top 5th Gen SKUs. One huge upgrade is the addition of very large pools of L3 caches which are accessible to the dual chiplet arrays. These chips also support faster DDR5-5600 speeds and Intel UPI 2.0 which provides 20 GT/s (Giga Transfers) of bandwidth.

Each chiplet is connected through a modular die fabric which sits between the two core and cache arrays. Each array features a total of 35 cores of which 3 cores are disabled per die. Each tile has 2 memory controllers supporting DDR5 DIMMs with up to 5600 MT/s speeds, three PCIe controllers (6 total), 2 UPI (4 total), and two accelerator engines (4 total). According to Intel, the Intel Emerald Rapids CPUs will offer:

Increased Performance & Performance/Watt

Increased Core Counts with larger shared L3 cache

Compute Express Link (CXL) Type 1,2, & 3

Increased Intel Ultra Path Interconnect Speeds (UPI)

Up To 80 PCIe Gen5 Lanes

Intel Accelerator Engines

Increased Memory Speeds

Some of the features to expect within the 5th Gen Xeon "Emerald Rapids" CPUs include:

Up To 3x larger LLC cache

Increased Memory Speed

Higher CPU Core Count (Up To 64 Cores)

CXL Type 3 high bandwidth interface

Workload optimizations (Intel AE)

Optimized Power Mode

Up To 17% General Purpose Perf/Wat Increase

Offload CPU cores with Intel Accelerator Engines for Power Efficiency

Huge performance boost across inference and training

Intel AMX for built-in AI Acceleration

Out-of-box deployed with Optimized SW stacks

Moving to the SKUs, it is interesting to see that Intel's 5th Gen Xeon CPUs only scale across 1S & 2S systems. The 4S & 8S configurations are missing from the lineup. The flagship Xeon Platinum 8592 chip comes in three variants, "+" for the General Purpose solutions, "Q" for the Liquid-Cooled and General Purpose solutions, and "V" for the Cloud Optimized systems. This chip racks in a total of 64 cores, 128 threads, a boost clock of up to 3.90 GHz, 320 MB of L3 cache, and a TDP of 385/350/330 Watts across the three variants, respectively. The prices range between $11,000 to $12,000 for the top chip.

For comparison, AMD's top EPYC 9654 "Genoa" CPU with 96 cores, 192 threads, 384 MB L3 cache, and up to 3.55 GHz clocks costs $11,805 US (1kU) while the EPYC 9754 "Bergamo" with 128 cores, 256 threads, 256 MB L3 cache, up to 3.1 GHz clocks costs $11,900 US (1kU). So in terms of core per dollar, you are getting more out of the EPYC 9004 lineup than Intel's Emerald Rapids chips.

The Intel 5th Gen Xeon Platinum lineup also includes the 8581 with 60 cores, 8580 with 60 cores, 8570 with 56 cores, 8568Y+ with 48 cores, 8558 with 48 cores, and the 8562 with 32 cores. These SKUs range. Only the 8562Y+ drops the cache below 100 MB to just 60 MB.

Next up, we have the Intel 5th Gen Xeon Gold SKUs which feature the 6558Q as the fastest SKU which should be expected since it is a liquid-cooled variant of the 6548Y+. This chip has 32 cores, 64 threads, a boost clock of up to 4.1 GHz, 60 MB of cache, and a 350W TDP. The Xeon Gold 6554S and the 6530 remain the only SKUs in the Gold segment with a cache exceeding 100MB with 180 & 160 MB, respectively. The lineup ranges from 8, 16, 24, 28, 32, 36 cores.

Lastly, we have the Xeon Platinum Silver family which includes five SKUs with the 4516+ leading the pack. It features 24 cores, 48 threads, a boost clock of up to 3.7 GHz, 45 MB of L3 cache, and 185W TDP. All 5th Gen Xeon Silver SKUs are listed with DDR5-4400 support and range from 8, 12, 16 & 24 cores.

Coming back to the performance numbers, Intel is sharing some data based on different workloads and comparing them against the 4th Gen Sapphire Rapids Xeons. The 5th Gen Emerald Rapids CPU offers:

Up To 42% Uplift In Image Segmentation (AI)

Up To 24% Uplift In Image Classification (Edge AI)

Up To 42% Uplift In Modeling / Simulation (HPC)

Up To 24% Perf/Watt Uplift In NVMe over TCP (Small Packet Random Read)

Up To 69% Uplift In Network Security Appliance

Up To 33% Perf/Watt Uplift In Server-Side Java (Within Given SLA)

The company also shares some performance figures comparing its flagship Xeon 8592+ 64-Core CPU against AMD's EPYC 9554 64-Core chip. Both chips are tested in a wide spectrum of HPC, Web, Storage, and Data Service-oriented benchmarks. The Xeon scores up to a 2.51x uplift in general server performance and a 1.92x performance per watt gain which is impressive if true.

There are also comparisons against the top EPYC 9654 96-core CPU but here, Intel switches the gears and compares the AI performance of its chips. Intel having various AI accelerators within its Xeon family gets the edge over AMD's EPYC with up to 2.6x gains. The Emerald Rapids CPUs also offer a 1.92x performance boost over the 4th Gen Sapphire Rapids CPUs with their larger L3 cache.

Today's launch marks the step into 2024 where we are expected to get two brand new Intel Xeon families. First up, we have the Sierra Forest chips with up to 288 E-cores launching in the first half of 2024 and then we have the P-Core Granite Rapids (6th Gen Xeon) family coming soon after that.

Following are the full list of SKUs along with their listed specifications: