Performance
Performance has been tested on the AMD Ryzen 3000 platform, which includes the R9 3900X processor and ASUS Strix B550-I Gaming motherboard, which is one of the best motherboards for memory overclocking on AMD platform. 3D tests were performed with the help of the XFX RX5600XT THICC III Ultra graphics card.
Everything is readily available in stores so our readers can compare results on their computers.
All results were performed on the G.Skill TridentZ Neo 16GB DDR4-3600 CL14 memory kit, so this is what you can expect, but of course, overclocking is never guaranteed as it depends on many factors.
Let’s begin the tests.
Memory bandwidth in AIDA64 is scaling well with memory frequency, which is typical for the Ryzen 3000.
There is also a latency factor that is not included in the above graph. The XMP will give us about 68ns while the highest, DDR4-4800 CL18, achieves about 69ns. In some tests, you will see that similar results are at DDR4-3600 and DDR4-4600 or DDR4-4800. This is how much give the 1:1 infinity fabric divider.
Rendering benchmarks like Cinebench series, don’t react so good to memory settings. As far as we can tell the best result, then all seem not far from each other.
More significant differences we can see in benchmarks that simulate real-world workload like PCMark series. In the PCMark 10, the XMP performs well, achieving close to the best scores. It just proves there is not much more we can make to improve the performance of the Neo memory kit significantly. All our settings, more or less tweaked, are showing or slightly better or slightly worse results.
In UL benchmarks, which base mostly on graphics card performance like 3DMark Time Spy and Fire Strike series, we can count on similar results at all our settings. The XMP is about as fast as any overclocked result, and the difference is within the error margin. However, in these tests, we can see that the DDR4-3733 is about the best, so I assume we will be able to see a significant improvement once the Ryzen 4000 arrives with its higher memory speed support.
Results in the VRMark are all about the same, and I can say that all are within an error margin. There is a 1FPS difference in the least demanding Orange Room. Besides that, we can see exactly the same results in all our settings.
About the same story, we can see in graphics benchmarks at a higher display resolution. The Neo is performing well in Final Fantasy XV and Superposition benchmarks. We can also see that anything at the infinity fabric ratio of 1:1 runs better in the Final Fantasy XV. The Superposition works about the same at all settings, but the higher frequency gives some single points more.
New games at a lower display resolution, which is also the most popular, are already showing some differences. We can see up to 5 FPS at various memory settings in the Tomb Raider. On the other hand, Farcry 5 is not reacting much to memory settings. There would be a higher difference if we added standard speed memory kits, but nowadays, gaming computers are not really sold with standard and slow memory kits.
As I already mentioned, the TridentZ Neo has well-tuned XMP, so even though it has a high overclocking potential, then we can’t count on significant improvements during daily work. The memory is simply already designed to perform great out of the box so the user can enjoy its favorite game and not spend the time on long overclocking and stability tests.
The TridentZ is, of course, one of the most popular overclocking series, so many of those who decide on this memory series are counting on additional overclocking headroom. On the next page, I will show some of our results.
