AMD Ryzen 7 3700X review: gaming performance tests

How memory clocks affect performance.

Review by Richard Leadbetter Technology Editor, Digital Foundry

Updated on 21 Jul 2019

Historically, the importance of memory bandwidth in gaming was totally overlooked - mostly because the GPU defined so much of the gameplay experience, but I'd say that two new technologies have changed that radically. First of all, there's the arrival of high refresh rate monitors, with gamers now looking to target 144fps or even 240fps from their games. Pushing frame-rate this hard brings us to CPU limits - which are defined in many cases not just by core frequency, but also by bandwidth. Secondly, there's the arrival of Ryzen. The component core complexes communicate at a speed defined by system memory frequency - meaning more performance is unlocked by running with faster RAM.

Things have moved on significantly though since the first-gen Ryzen launched. Back then, a 3200MHz kit based on Samsung B-Die memory modules was the best way to run AMD's processor and to avoid compatibility issues, but this came with a significant price premium. The situation has changed in more recent times, thankfully. Ryzen is now capable of comfortably running on many more memory types, while the cost of 3000MHz and 3200MHz kits has dropped massively.

For our review, AMD supplied 16GB of 3600MHz GSkill Trident Z Royal - which isn't cheap. So what happens to game performance when you use cheaper, slower memory? Here you can see how performance is impacted by running at 3000MHz instead of 3600MHz. Three tests produce very different results. There's a circa seven per cent drop in frame-rates on all processors in the Ashes of the Singularity CPU test, while Far Cry 5 shows Intel holding onto more of its performance compared to AMD. Meanwhile, Crysis 3 shows no degradation at all on the Core i7 9700K, a minimal three per cent on Ryzen 7 3700X, and a wide 10 per cent delta on the last-gen AMD offering.

More data is needed for firmer conclusions, but again, the results here suggest that AMD's reliance on fast memory has reduced, while Intel still seems to cope best with lower bandwidth RAM. One thing worth pointing out though is that AMD allows faster, overclockable memory on both mid-range and high-end boards, while Intel demands a higher end Z board to run memory above the CPU's bandwidth limitation - which in the case of the 9700K is a mere 2666MHz. Food for thought based on some of these results...