GPU Test Bench and Benchmarking Refresher

The FPS Review may receive a commission if you purchase something after clicking a link in this article.

Introduction

With NVIDIA’s GeForce RTX 30 Series graphics cards right around the corner and AMD’s Radeon RX 6000 Series graphics cards right around the corner, I wanted to detail our main GPU test system configuration. I also wanted to talk about how we will be testing here at TheFPSReview.com for all these exciting new GPUs. If you have not checked out our previous article on this subject, more than a year ago we wrote about our methodologies, and our benchmarks, and finally our testing format. Today’s new article, now more than a year later refreshes this topic with new information and gives you a run-down on our test system build.  

A big question that has been on people’s minds is if PCI-Express 4.0 is going to make a difference in this generation.  AMD has already supported PCI-Express 4.0 on its GPUs for over a year now.  The Radeon RX 5700 series introduced this in July of 2019. Along with the introduction of AMD’s X570 based motherboards in that same year, we have had PCI-Express 4.0 platform support. However, we haven’t really seen the fruits of this bandwidth for gaming performance, except in very specific VRAM limited situations on the low-end Radeon RX 5500 XT 4GB variants where running out of VRAM in a game meant stuttering gameplay.

NVIDIA’s GeForce RTX 30 series now also supports PCI-Express 4.0.  The question is will we see benefits of this bandwidth given the high-bandwidth and performance nature of the RTX cards like the RTX 3080 and RTX 3090?  Any way you slice it, PCI-Express 4.0 is a hot topic and one that we need to be sure we are set up to test.  With both AMD and NVIDIA now supporting PCI-Express 4.0, it is important that our test system also support this and we can put the cards head-to-head on the same platform and give them the best chance possible.

The Goal

Our goal is three-fold.  First, we want to have a system setup that exploits the best performance out of both AMD and NVIDIA GPUs.  To give each GPU the best chance possible at performing as well as it can, and to not hold it back. 

Second, our goal is to use a system setup that a person can obtain easily and real-world to the mainstream gamer.  We don’t want to use some exotic thousands of dollar CPU that a mainstream gamer simply won’t be able to get their hands on.  We want to use real hardware at real-world pricing that is easy to obtain, replicate results on, and that makes common sense. 

Third, our goal is to show you apples-to-apples performance comparisons at relevant game settings and resolutions that fit the scope of the GPU we are reviewing.  We want you to be able to replicate our testing scenarios so you can benchmark your own GPU and compare your results with our own.  In this way, you can see what kind of a performance upgrade the GPU we are reviewing will provide over your own setup.  This means we will be utilizing in-game benchmarks where possible so you can run the same benchmark and compare your results to ours. 

Upgrades – Motherboard

In order to facilitate these goals, we have made some upgrades recently to our primary GPU test bench, and we want to detail that for you now.  We made the choice to go with an AMD X570 chipset-based motherboard.  We will be utilizing an Asus ROG X570 Crosshair VIII Hero (Wi-Fi) motherboard.  This gives us the PCI-Express 4.0 support we want on the GPU and allows us to utilize a PCIe 4.0 NVMe SSD in the future.  It also has the CPU to Chipset bus at PCI-Express 4.0.  This means that AMD and NVIDIA video cards will equally have access to PCI-Express 4.0 graphics pipeline and system I/O is as fast as it gets right now.

CPU and Cooler

We have upgraded our CPU now to an AMD Ryzen 7 3700X (from the 2700X.)  This is an 8 core/16 thread CPU based on Zen 2 and should provide all the threaded goodness we need for mainstream and enthusiast gaming.  We will be running the CPU at stock performance with the latest BIOS and AGESA updates.  In this way, you can compare our performance to your own CPU at stock.  We will also always keep the BIOS updated to utilize the most recent AGESA code.  In the future, we will be able to upgrade to Zen 3 Ryzen 7 4700X, or whatever they call it.  Cooling it is a Corsair H115i Pro RGB AIO.  This keeps the CPU plenty cool so it can operate at its best boost clock.

RAM

One big upgrade we have just made is to upgrade our test system from 16GB of RAM to now 32GB.  It was finally time to make that leap. For the longest time, 16GB has been sufficient, but we are starting to see games now where 16GB is starving performance and introducing stuttering. One of the games that prompted this move was Flight Simulator 2020, where we have found 16GB to be a bottleneck. This game can consume much more RAM, and doing so means a smoother experience.  We didn’t want the game being sluggish or pausing and stuttering simply because it was running out of RAM. 

We have two 16GB sticks of G.SKILL Trident Z Neo (F4-3600C16D-32GTZNC) DDR4-3600MHz 16-19-19-1T RAM installed.  The move up to 32GB from 16GB ensures every game can load all the assets it needs into RAM without bottlenecking. 

Storage

We have also now adopted a 2TB PCI-Express 3.0 x4 NVME SSD as an upgrade from SATA SSD, the ADATA SX8100NP 2TB NVMe SSD.  This will allow us to store all the games we are using, with games like Flight Sim 2020 taking up over 100GB now, the need for 2TB is obvious so we can have many games installed.  In addition, by utilizing a fast PCI-Express 3.0 x4 NVMe SSD game load times and streaming of assets will not slow us down and minimize stuttering in the game.  In the future, we can upgrade to a PCIe 4.0 x4 NVMe SSD to keep up.

Power Supply

In order to make sure the GeForce RTX 30 series GPUs and upcoming Radeon RX 6000 series GPUs have plenty of power and stable voltage, we are using a 1000W Seasonic Prime Ultra Titanium power supply to provide power.  This power supply is only a year old and should provide stable voltage to every video card.  In addition, we have the entire computer plugged into a TrippLite LC1200 Line Conditioner with Active Voltage Regulation.  Clean voltage regulated power will be coming into the system. 

For our display, we are continuing to use an LG 27UK650-W 4K display to test.  Since we test with VSYNC off to show the maximum performance, the refresh rate does not matter.  This display lets us test at the popular resolutions up to 4K.

In addition, we will always be running Windows 10 using the latest major update and build, we keep it updated via Windows Update.  We also keep all of our software and games up to date.  We also enable the “AMD Ryzen High Performance” power profile in Windows and disable background applications.

You can view the entire system on our KIT page.

Brent Justicehttps://www.thefpsreview.com
Former managing editor of GPUs at HardOCP for 18 years, Brent Justice has been reviewing computer components since the late 90s, educated in the art and method of the computer hardware review, he brings experience, knowledge, and hands-on testing with a gamer-oriented and hardware enthusiast perspective. You can follow him on Twitter - @Brent_Justice You can sub to his YouTube channel - Justice Gaming https://www.youtube.com/c/JusticeGamingChannel You can check out his computer builds on KIT - @BrentJustice https://kit.co/BrentJustice

Recent News