Intel offers advantages for quicker memory access for everyone.
Intel offers advantages for quicker memory access for everyone.
For those using intel z390 or newer, are you noticing advantages from extremely high memory speeds (>3600mhz)? I'm curious since I've seen several test setups for the latest cards at 4000mhz and above on 10900k boards. Do you have any reliable data showing real gains, or is this mostly a marginal improvement (<5%)?
It all hinges on your CPU's base clock and the software you run. When your RAM is quicker, the CPU can work more efficiently, especially if the program makes good use of cache. If the CPU often waits for data, improving RAM speed helps, but the gain isn't always huge—it depends on how well the app leverages the cache. This principle has been backed by research since 1988 and is known as Gustafson's law. https://en.wikipedia.org/wiki/Gustafson's_law
Here’s a quick example:
- At 5.2GHz with 3600CL15 XMP, using 4000CL17 RAM and 200MHz more cache, performance drops noticeably in multithreaded tests.
- Even with 3600MHz RAM, you’re limited by bandwidth, not just clock speed.
That means a 3600MHz setup might not give you much extra speed over a 5GHz system, especially if the CPU is still bottlenecked by memory. Digital Foundry’s review on their 8700K also touches on this trade-off.
In short, optimizing RAM alone isn’t enough—match it with your application’s needs and hardware capabilities.
Memory speed improves significantly with more CPU cores and cache overclocking. 4000mhz is the lowest setting I use with my 8700K, usually CL16-16-16. This screen shows a 5.2ghz display, 5ghz cache, and 4ghz RAM. The GTX 480 is running 3DMark IceStorm. The Physics score likely reflects CPU power plus memory speed in this image.