Performance has been tested on the Intel Skylake-X platform which contains the 7900X, 10-core processor, EVGA X299 Dark motherboard, EVGA GTX1080Ti FE graphics card and Crucial P1 1TB SSD with installed Windows 10 x64 and the latest updates.
If anyone wishes to compare memory bandwidth at XMP settings on the AMD Threadripper and the Intel Skylake-X then below I’m adding a screenshot from AIDA64 Cache & Memory Benchmark.
AIDA64 Cache and Memory benchmark is probably the best software to check synthetic memory speed.
Results are great regardless of used platform, and I can even say that on the AMD, it performs slightly better.
As you can see, our comparison includes also overclocking results and settings at which the Ballistix Sport LT DDR4-3200 was stable. Maximum stable settings on our motherboard were DDR4-3466 CL16-16-16 at default 1.35V. Higher clock, even up to DDR4-3733 CL17 was possible but was causing some instability and I guess it’s more related to the platform itself than the memory. On the next pages, you will find out why I think that.
PCMark 10, which base on popular applications is showing that performance is scaling better with memory frequency than latency. It clearly looks that large and fast CPU cache is covering delays which could happen. Ballistix’s XMP profile shows optimal performance which is about as high or barely lower than that of overclocked memory. Excellent work with the memory profile.
Rendering benchmarks like Cinebench series, are showing lower differences between memory settings. This software uses mostly processor power. Rendering benchmarks can’t show the full advantage of RAM. Professional work will show higher differences where high capacity memory kits like our 64GB will be just perfect.
Let’s move to graphics tests based on popular 3D benchmarks from UL(previously Futuremark).
Both benchmark lines, so 3DMark and VRMark are showing higher performance gains related to memory settings at lower graphics details. It’s because less demanding environment uses more CPU power for calculations. In all tests, the XMP profile provides optimal performance or not much lower than the overclocked settings.
More demanding 3D tests at the display resolution up to 8K are not much different. In Final Fantasy XV and Superposition benchmarks results are slightly better at higher memory frequency but nothing that would highly affect our gaming experience. Final Fantasy XV benchmark shows that there is already quite a big difference between DDR4-2666 memory and DDR4-3200. Again, nothing severe but it’s clearly visible on our graphs.
New games like Farcry 5 or Shadow of the Tomb Raider, which is using DirectX 12 show that the difference in performance between various memory settings can be up to 3 FPS. Not much but it counts. The same games will show up to 5-6 FPS difference in 1080p display resolution. Simply if you play in lower screen resolution, then memory performance is more visible in games.