Unreal Engine – Intel Xeon W-3300 Processor Performance

Introduction

Intel has long been a staple in the world of workstation computing, but when AMD released their 3rd generation Threadripper line in late 2019 (and more recently Threadripper Pro), they took over the performance crown for a number of workflows. With the launch of the new Xeon W-3300 series, however, Intel is looking to retake its position as uncontested top dog in the workstation space.

The Intel Xeon W-3300 series of processors include a number of advantages over the previous W-3200 line, including an increase in max core count, 64 lanes of PCI-E Gen 4.0, 8 channel DDR4-3200 memory (up to 4TB max), and up to an 18% increase in IPC (instructions per clock). Threadripper Pro still has the advantage in terms of total core count (64 vs 38) and PCI-E Gen 4.0 lanes (128 vs 64), but these changes – plus a number of other improvements – should make the Xeon W-3300 series a significant upgrade over the previous generation.

If you want to read about what sets the Xeon W-3300 series apart in more detail, we recommend checking out our landing page for Intel Xeon W-3300 Processors.

In this article, we will be examining the performance of the new Intel Xeon W-3300 series CPUs in Unreal Engine compared to AMD’s Threadripper Pro line. If you are interested in how these processors compare in other applications, we also have other articles for several other applications on our article listing page.

One very important thing to note is that we will be performing our testing with both the default “Balanced” Windows power profile, as well as the “High Performance” profile. In the course of our testing, we discovered that the Xeon W-3300 processors can sometimes give significantly lower performance on the default Windows power profile, so we thought it was important to show results for both profiles.

We will be leading with the “Balanced” profile results as that is what most systems will likely be using by default, but since changing the power profile is extremely easy, we will look at the performance with both power profiles.

If you would like to skip over our test setup and benchmark sections, feel free to jump right to the Conclusion.

Test Setup

Listed below are the specifications of the systems we will be using for our testing:

*All the latest drivers, OS updates, BIOS, and firmware applied as of July 20th, 2021

Benchmark Analysis: Compiling Source Code

This test isn’t specific to Unreal Engine but is common enough amongst game developers and machine learning development to include it in this round of testing. For this, we used the “Release” branch (version 4.25) from GitHub and compiled it in Visual Studio 2019.

Intel’s 32-core Xeon W-3365 is within a few percent of AMD’s 32-core Threadripper Pro, while the 24-core is not far behind. At 16 cores, Intel and AMD are in a virtual tie. The real outlier here is Intel’s 38-Core Xeon W-3375. After several reruns of the test, it consistently performed poorly while compiling code in Visual Studio. While observing Task Manager, rarely did more than 12 threads hit 100% utilization at any one time. This must be some processor scheduling issue, either with Windows, Visual Studio, or the CPU itself. Until there is a fix in place, it is impossible to recommend this new 38-Core CPU to anyone working with Visual Studio.

For the Windows Power Profiles, some CPUs from both camps saw a 10% improvement, some saw no improvement. Since this is an easy thing that any user can enable or disable, we recommend trying it out with your workload and seeing if you get an improvement.

Benchmark Analysis: Build Lighting

While more and more Unreal users are turning to Unreal for its real-time lighting features, traditional rasterized rendering still has a lot of use. Anyone that has used Unreal for a while can attest to how much time is spent waiting for lighting to be rebuilt. For this test, we used “Medium” lighting quality in the “Virtual Studio” project from Epic. Also, note that GPU Lightmass is now available, but for this test, we are focusing on the legacy CPU baking.

When it comes to building lighting in Unreal, AMD is considerably faster across the board. On a cores-to-cores comparison, Intel isn’t close at any tier. Even worse, their top-of-the-line 38-core is as fast as their less expensive 32-core.

The Windows Power Profile makes no measurable difference for most CPUs, except for the 16-core chips from both companies. Intel got a little faster, while AMD got a little slower.

Benchmark Analysis: Recompile Shaders

Recompiling shaders is something that can take a lot of time depending on the workflow. Smaller projects or virtual production or architecture may not have to rebuild all their shaders very often, if at all. On the GameDev side, they may need to rebuild all their shaders quite often. Not everyone on the team, but someone will be doing this often.

Intel is able to narrow the gap a little bit but is still being outperformed by AMD across the board. We finally see Intel’s 38-core performing the way it should, though it is barely ahead of AMD’s much less expensive 32-core offering.

In this test, enabling the Performance Profile in Windows actually shows a good improvement in most of the top-end SKUs. AMD’s 32-core manager to leapfrog the 38-core Xeon. As this is a task that everyone using Unreal will experience to some extent, it may be worth trying on your system.

Overall Unreal Engine Performance Analysis

Most development tasks in Unreal Engine, especially outside of gameplay, prefer more cores, but also benefit from those cores being fast. If we are to combine the above results into a scoring system, Intel’s new Xeon W-3300 series just cannot keep up with AMD’s Threadripper Pro line. The Xeon W-3375 38-core CPU’s poor code compiling performance combined with its middling light building results makes it a bad choice for Unreal Engine work.

Comparing the Windows power profiles, we see on average it boosts the performance of most CPUs, with some showing little improvement. The Threadripper Pro 3955WX 16-core actually performed worse with Performance Mode enabled.

How well do the Intel Xeon W-3300 CPUs perform in Unreal Engine?

The new Intel Xeon W-3300 line offers more performance than previous generations but still falls short of AMD’s Threadripper Pro line. On a per-core basis, the Xeon is close but only offers up to 38 cores, while AMD offers up to 64 cores. Then there is the cost. Comparing the two 32-core options, Intel is $950 more expensive, but almost 10% slower. Intel’s 38-core Xeon W-3375 has some serious issues with compiling source code, so should be avoided that is something you spend a lot of time with.

Intel’s biggest selling point is Xeon’s long track record of reliability. Xeon has long been the most stable and reliable CPU in highly demanding workstation and server applications. That alone will be worth the cost to many people. AMD is attempting to compete in that same market with their Threadripper Pro line, but as a new product, it doesn’t have history. These performance numbers may be enough to question the long-term value.

Keep in mind that the benchmark results in this article are strictly for Unreal Engine and that performance will vary widely in different applications. If your workflow includes other software packages, you need to consider how the system will perform in those applications as well. Be sure to check our list of Hardware Articles to keep up to date on how all of these software packages – and more – perform with the latest CPUs.