Performance
Performance has been tested on the AMD Ryzen platform, which contains the Ryzen 9 5900X processor, MSI X570 Unify motherboard, and ASUS RTX3070 TUF OC graphics card. Used OS is Windows 10 x64 and the latest updates.
Our comparison includes overclocking results and settings where the Viper Steel RGB 64GB DDR4-3600 memory kit was stable. Our maximum stable settings were DDR4-3800 CL18-22-20 at 1.40V. Higher voltages or more relaxed timings were not helping to pass this memory clock. Maybe it’s a matter of luck as the test platform could make higher frequency with the previous Patriot memory series.
As usual, we will start with AIDA64 Cache and Memory benchmark, which is probably the best software for synthetic memory speed tests.
AIDA64 Memory and Cache benchmark likes high memory frequency. It’s clearly scaling well with memory frequency but also reacts well to memory latency. The most important result in daily work is memory copy. It’s also a similar result to what Windows 10 shows in its internal memory speed test. Even though Steel RGB runs at really relaxed timings, our results do not show that so much. We could still achieve over 51GB/s read and write bandwidth, while memory copy was not much worse than at DDR4-3600 CL16-19-19 settings.
If we take a look at latency results, then we can see significant differences. The difference between the XMP profile and DDR4-3800 CL18-22-20 setting is nearly 6ns, and this is a lot.
In the Cinebench R20 benchmark, results are close to each other, and we can’t clearly say if one setting is better or it’s a margin error.
PCMark 10 is reacting well to memory performance. Our results are not much different, but we can clearly tell which setting is the best. The Productivity test is the only one that is not really showing performance gain because of memory settings. In two other cases, the best are DDR4-3600 CL16 and DDR4-3800 CL18 settings.
3D benchmarks from UL are showing almost the same performance differences as the PCMark 10 benchmark. However, here we can see that the more demanding benchmarks are affected less by memory settings. In Time Spy Extreme, the difference between the XMP and DDR4-3600 CL16 can be counted as an error margin, while in the Fire Strike test, it’s already a significant difference. It’s, of course, hard to translate into performance in games.
In VRMark, we can see a similar connection between memory settings and the test’s difficulty. Except for the Orange Room test, all other results are about the same in all settings.
The Final Fantasy XV benchmark shows that latency is not as important as memory frequency. Both DDR4-3600 results are similar and significantly better than DDR4-3200. On the other hand, the same performance gain is at DDR4-3800.
Superposition benchmarks are showing similar results, regardless of memory settings.
Modern games on the new AMD Ryzen platform like fast memory. The performance gain is scaling well with memory frequency, and results in all these games look about the same as in the Final Fantasy XV benchmark. The difference between DDR4-3200 and DDR4-3800 can go up to even 8FPS. Games using more graphics card power will show a lesser gain, but it’s still visible in benchmarks. On the other hand, I highly doubt that gamers will see the difference between 160FPS and 168FPS. So high FPS difference could be important at about 60FPS, but not above 120.
Our comparison says that the Viper Steel RGB 64GB DDR4-3600 memory, even though it has really relaxed timings, isn’t slow. Its performance is high enough for gaming or anything else, but it could run at tighter timings to improve results. The memory is capable of a better specification or could use a different IC, which shouldn’t cost more.
