Demystifying Intel’s Range of Gen-6 Xeon Processors

A blog by Allan Kaye, Managing Director at Vespertec

—

Release date: 20 January 2025

If you’ve been keeping an eye on Intel’s Xeon 6 processor family, which began rolling out in Q2 2024, you may have been slightly confused by the difference between the Xeon 6500, 6900P Granite Rapids, 6700E Sierra Forest, and so on.

Intel has now introduced some tailored variants of Xeon that cater specifically to hyperscale and high-performance computing environments. This means launching products with two types of cores: E-cores (efficiency-focused) and P-cores (performance-focused), aimed at different use cases.

So, in this blog, I’ve laid out exactly what you can expect from the two types of cores, as well as the 6900, 6700, 6500, and 6300-series processors, each designed for a specific performance tier.

E-cores vs. P-cores: who are they for?

First, it’s important to note that both E-cores and P-cores use a compatible x86 instruction set architecture (ISA) and a common hardware platform, including CPU socket type. So, to help you understand the differences between them, and which variant of processor suits your workloads, let’s break down the technical differences:

Granite Rapids and Sierra Forest

If you’ve been looking into the Xeon 6 series, you’ll likely have seen the terms ‘Granite Rapids’ and ‘Sierra Forest’ around. Essentially, these terms are Intel’s way of classing processors fitted exclusively with P-cores (Granite Rapids), or with E-cores (Sierra Forest).

P-cores

P-cores, as you’d suspect from the name, are designed for high-performance workloads that require substantial per-core processing power. This makes them ideal for compute-intensive tasks like AI, high-performance computing (HPC), machine learning, and large-scale simulations. For instance, the Xeon 6900, the most advanced processor in the series, is powered by up to 128 P-cores per socket.

These cores support Intel AVX-512 and Intel AMX instructions for boosting vector-based and AI workloads, especially those requiring heavy floating-point operations. Also, the higher L3 cache sizes help to reduce latency and increase throughput in tasks that benefit from large data sets.

Finally, memory-wise, P-cores use DDR5 6400 and can support MRDIMM for up to 8800 MT/s, allowing for faster memory bandwidth, especially in HPC and AI environments.

E-cores

E-cores, on the other hand, are all about efficiency and high core density. Smaller than P-cores, they’re optimised for applications where power and cooling are limiting factors. Having been optimised for high throughput per watt, E-cores consume significantly less power, allowing data centres to consolidate workloads and manage background tasks efficiently, contributing to overall system energy savings.

To give you a sense of the difference in density, the Xeon 6900 can fit 288 E-cores per socket, but only 128 P-cores.

E-cores have been built to tackle task-parallel, scalar workloads that don’t require massive compute power per thread. That makes them best suited for cloud-native applications, microservices, and scale-out workloads like containers and task-parallel environments.

Finally, these cores support Intel AVX2 with VNNI (Vector Neural Network Instructions), which enhances AI processing capabilities without the power overhead of P-cores.

Now, having broken down the differences between their core types, let’s get into the features and strengths of the Xeon 6900, 6700, 6500, and 6300-series.

Intel Xeon 6900-series

The 6900-series is Xeon’s most advanced set of processors, whether 6900P Granite Rapids or 6900E Sierra Forest. This architecture is designed for maximum performance environments such as cloud, AI, and high-performance computing (HPC).

Its key features include the highest memory bandwidth and throughput of all the Xeon 6s, support for ultra-high density compute performance, and up to 128 P-cores or 288 E-cores per socket. It’s ideal for demanding workloads like AI inferencing and scientific simulations.

Intel Xeon 6700-series

Where its bigger brother, the 6900, is designed specifically for AI, cloud, and HPC, the 6700-series excels in a more diverse set of environments, including data centre and telco.

With up to 86 P-cores (or 144 E-cores), the 6700 is optimised for a balance between performance and efficiency. You’ll find it suitable for a broad range of data centre workloads, including transactional databases and HPC applications.

Intel Xeon 6500-series

The workhorse of the Xeon 6 series, the 6500 is designed for mainstream server and edge environments where power efficiency and cost are the primary considerations. Unlike the 6900 and 6700, which are aimed at high-performance and diverse data centre environments, respectively, the 6500-series focuses on balanced performance within standard data centre power and cooling limits.

With a mix of core counts and enhanced I/O capabilities, it’s well-suited to general-purpose applications such as virtualisation and moderate-sized databases, providing a solid foundation for environments where predictable performance at an efficient power envelope is paramount.

Intel Xeon 6300-series 

At the entry level, the 6300-series targets small to medium-sized businesses that need reliable performance on a budget. The 6300 is optimised for lower-demand compute tasks, making it a strong fit for use cases like web hosting, small databases, and general IT infrastructure.

Though it lacks the high core densities and memory bandwidth of its larger counterparts, the 6300-series offers a cost-effective solution for organisations looking to deploy scalable, basic server capabilities without the expense of higher-tier models. This makes it particularly useful for businesses prioritising affordability, without sacrificing scalability later on.

Real-world performance insights

Now, let’s see what that looks like in the rack.

ServeTheHome found that Intel’s 6700E Sierra Forest processors, particularly the 144-core Xeon 6780E, can deliver up to a 9:1 consolidation ratio when replacing legacy systems built on older Xeon Gold 5218 processors. This means that data centres can consolidate up to nine older systems into a single new server powered by the 6780E, achieving over 2.5x power savings at the server level. This makes the 6700E series highly effective for hyperscale cloud environments aiming to modernise their infrastructure while reducing energy costs.

On the other hand, the 6900P processors, especially the 128-core Xeon 6980P, mark Intel’s return to x86 leadership. With support for 12 memory channels and 96 PCIe Gen5 lanes per CPU, they excel in AI and high-performance computing workloads. According to ServeTheHome, these processors surpass some well-known competitors in core count and memory bandwidth, delivering superior performance in demanding environments​​.

Suited for any workload

So, whether you’re upgrading for AI, HPC, or efficient cloud scaling, the Xeon 6 series offers you a variety of options, tailored to meet the specific demands of modern data centre environments.

For more breakdowns of hardware specifications, see my last blog decoding NVIDIA’s reference architectures and product categories.

If you’d like to know about the availability of the top-end Xeon 6 processors, or want to chat about which use cases these would be a perfect fit for, get in touch.