Testing Philosophy

The objective review goal is to test system memory using a real world configuration and use-case for all types of consumers. These reviews focus on providing an accurate representation and set realistic expectations using a wide range of benchmarks to showcase the best and worst of the product in question. To accomplish this, some accommodations had to be made. To produce repeatable results variables must be accounted for and taken into consideration during testing. For example, the processor is locked to a set frequency to rule out CPU Turbo as a factor. System memory's Primary and Secondary timings are also observed to make sure that it is accurately applied when XMP/EXPO Profile is enabled. If the motherboard in question does not apply these timings correctly, they are manually entered. Random spot check validations are performed comparing previous results to make sure they are still the same as before, which lowers the probability of bad data throughout large data sets.

One variable that cannot always be fully controlled is the graphics card. This is because it uses a clock frequency curve based on temperature, voltage and power draw. However, benchmarks are run in a specific order to ensure every memory kit has the same starting point. Room temperature is also monitored to ensure a minimal impact on performance. Once again the overall goal is to have the least amount of variables while testing, which could skew the results.

What does it mean to be GPU Bound?

One of the most common misconceptions and inaccurate information on how memory impacts gaming performance is the lack of technical understanding of what it means to be GPU bound. It has a very simple core concept, but with many details that need to be factored in for an accurate assessment. First, let's start with a generalized definition of what it is to be GPU bound. The term keep "GPU Bound" refers to when the performance (FPS) in a game or application (GPU task) does not change regardless of the processor and or memory used in the computer because the graphics card in question is the overall systems limiting factor. The statement is broad, which allows for the unattended misuse of the phrase. The complexity becomes more apparent when CPU clock frequency, number of CPU Cores, memory speeds, the total amount of memory available and application in question is combined.

The Breakdown

To break it down to get a better understanding, we can first take a look at the impact of just the CPU frequency on a given game's frame rate. Using the Ryzen 9 7950X and two games for this example, we can see that games are impacted by clock frequency. While Forza Horizon 5 is considered within the margin of error due to being GPU bound, Far Cry 5 is on average down to 15 FPS and has a greater dependence on the CPU.


This next chart shows the impact on the average frame rate compared to the number of cores enabled. Once again, Forza Horizon 5 barely shows any signs of change until we are left with only 2 cores (4 threads). This is when game frame rates become impacted due to not having enough resources for Windows background tasks.

Far Cry 5 on the other hand, did something unexpected. 16 cores enabled came in at 3rd place. I'm unsure why this is the case, and the results are repeatable too. Not only is CPU frequency important to Far Cry 5, but also the number of CPU cores enabled.


Setting the Ryzen 9 7950X to the base frequency (4.5 GHz); just changing the memory ratio had little impact in either game.


Putting the ratio back to 2:1:1, we can see how just the Infinity Fabric impacts these two games. Once again, Far Cry 5 seems to be the most impacted.



With the data collected so far, the final test is to see how much of an impact system memory has on Forza Horizon 5 and Far Cry 5 once the CPU is the bottleneck. On the highest graphical preset, the average frame-rate impact is minimal. However, when switching to medium settings, the CPU bottleneck magnifies. These are only two examples of how system memory can have an influence on a game's frame rate under the right circumstances.

The frame-rate limitation here for Forza Horizon 5 is mainly the graphics card. This is an infinite cycle; a more powerful graphics card raises the frame-rate, which will surpass the current CPU abilities to compute the necessary game logic for each frame rendered. Games that are less optimized, have more world physics, and have larger general compute sets per frame will need more CPU processing power. If the processor in question can also complete those tasks quicker with the aid of faster system memory. Ultimately, it depends on the configuration of the computer, game resolution, and graphical preset. If you find this interesting, try to replicate this breakdown of your favorite game and share your results on the TPU forums!

Why do I get different results?

Besides having different computer hardware, other factors play a role in the overall frame-rate. Things like the version of Windows you are using, current graphics drivers, game version and Windows power saving modes will change the outcome of the results to a certain degree. Motherboard features such as Resize-Bar or memory training algorithms can change the overall frame rate for games as well.

PCIe scaling is not often covered on a deep technical level. TechPowerUp has tested a number of games using a NVIDIA RTX 4090 in thisNVIDIA GeForce RTX 4090 PCI-Express Scaling article for further analysis.


AMD can also suffer from lower memory throughput due to how the CCD chiplets communicate with the system memory. Turn one CCD off, and you lose some bandwidth. Though, this seems to have little to no impact on games once you are GPU bound.