Performance
Performance has been tested on the AMD Ryzen platform, which contains Ryzen 9 5900X processor, MSI X570 Unify motherboard, RX6800 XT graphics card, Silicon Power US70 1TB NVMe SSD, and Abkoncore 850W 80+ Gold PSU.
All results have been made using Silicon Power XPOWER Turbine RGB 32GB DDR4-3600 memory kit. All overclocking results were stable and passed multiple tests, as listed in the comparison. The best setting on the tested AMD Ryzen platform is DDR4-3800 CL16-19-19 1.45V. If we used the latest Intel platform, then we can count that the DDR4-4200 CL18-22-22 1.45V will be the best in synthetic tests.
As usual, we will start with the AIDA64 Memory and Cache benchmark, which is probably the best application to check memory bandwidth and latency.
AIDA64 bandwidth results are about where they should be or even a bit higher. Results at DDR4-3600 CL16 and CL18 are almost the same while the next step, so DDR4-3800 already shows significant improvements. The DDR4-4200 CL18 setting is performing worse than expected. Even though it doesn’t work at a 1:1 ratio with the infinity fabric and memory controller, then on Micron IC tested in last weeks, this difference was lower.
Latency results are showing that sub-timings of the XMP profile are pretty good. The overclocked setting at the DDR4-3600 CL16 is only 2ns better, and it explains not a big difference in other tests.
In PCMark 10, we can see not big differences between all settings. In this test, DDR4-3800 and DDR4-4200 settings are the best. As long as the DDR4-3800 results were expected, then DDR4-4200 is a surprise.
Since this review, I’ve replaced Cinebench R20 with the latest, R23 version. The benchmark is not much different but runs tests for 10 minutes instead of a single pass.
During this quite demanding benchmark, we can see that DDR4-3600 and DDR4-3800 results are optimal. The XMP is not much worse than other settings, which is a good sign.
3DMark series benchmarks are showing similar results in all settings. We can tell which one is higher or lower, but most are closer to the error margin. The only clear difference is in the Fire Strike test, which is the easiest for the graphics card and reacts the best to RAM performance. The XMP performs well and isn’t far behind overclocked settings at DDR4-3600 or DDR4-3800. The DDR4-4200 setting is significantly worse in the last test.
VRMark usually reacts better to memory settings, while now we can see only up to 3FPS difference in tests where is already over 300FPS. Even though it’s a repeatable result, it’s more like an error margin difference.
Superposition 3D benchmarks are not reacting well to RAM settings, and regardless of settings, we are receiving similar results. Final Fantasy XV, on the other hand, shows better results at overclocked DDR4-3600 and DDR4-3800 settings. These results are not much better but still count.
The most interesting performance gain we can observe in new games. Depends on the title, we can see up to 13FPS improvement. The XMP profile is again not far from the best settings, but there is still about a 5FPS difference.
Silicon Power XPOWER Turbine RGB performs well at XMP settings and should be more than enough for gamers. There is not much more to achieve in more demanding games, while if we play at lower display resolution and we miss a couple of FPS, then we can always overclock memory or adjust some timings.
