Faster RAM is not that helpful
Faster RAM is not that helpful
In another post geofelt mentioned that quicker RAM isn't very useful because it often comes with higher latency and is linked to this study: http://www.silentpcreview.com/article1478-page1.html. The initial sections examine gaming performance with both integrated and discrete GPUs. Generally, faster RAM boosts frame rates when paired with an integrated GPU (which is rare), but not when using a discrete GPU—since the latter should utilize its own faster memory. On page 5, they compare various RAM speeds during CPU-heavy tasks and find no improvement, let alone gains, at higher speeds. In my setup I use two 8GB modules at 3000MHz with XMP 1. It was surprising to see this result. Have others explored the impact of faster memory speeds? I probably just expected better performance.
As clarified, quicker RAM typically results in marginally improved processing speed. In the linked review, performance declines with faster RAM due to either single-channel operation or using four sticks instead of two. Single-channel RAM reduces throughput, whereas additional sticks introduce slight management overhead for task distribution.
It's accurate that with discrete GPUs, faster RAM offers minimal benefit, but suggesting otherwise ignores the actual data.
I am focusing on the 2x8GB 2133MHz C15 and 2x8GB 3000Mhz C15 models. They share the same memory capacity and clock speed, but differ only in processing rate. Research indicates that for CPU-heavy tasks, increased memory speed doesn't significantly improve performance and may even reduce it in certain scenarios.
And even more specifically at the 7-ZIP compression test, which is the only one where the score with ram @ 2133 is ever so slightly above that of ram @3000 (83.0 s vs 83.2): if you have a scientific background, you will inmediatly notice that the difference is irrelevant and could easily be explained as a measurement error, or a random error. In any case, it's just a minor deviation. But indeed, there's little benefit in having faster RAM, for most tasks.
Blas elaborates further on the 7-ZIP compression test, emphasizing that the score with ram @2133 is only slightly higher than with ram @3000 (83.0 s vs 83.2). He points out that for someone with a scientific background, this difference seems negligible and could simply be due to measurement inaccuracies or random variations. He acknowledges that even then, faster RAM offers minimal advantage for most applications.
If you upgrade, performance will improve by about 5%, though it won’t be obvious. Higher bandwidth memory has a greater latency, but that doesn’t matter much. DDR4 3000mhz with a 14ms cache will still outperform DDR4 2133mhz with a 9ms cache. It’s better to choose faster memory from the start, especially when setting up your PC, since it lets you use the CPU’s built-in memory controller optimally.
My DRAM specifications are 2133MHz, 1.2V, 15-15-15-36, but with XMP activated it operates at 3000MHz, 1.35V, 15-17-17-35. Still, during stress testing my CPU temperatures rose by about 7 degrees Celsius on average. This raises the question of whether enabling the XMP profile is worthwhile.
burnhamjs :
my DRAM specs are 2133MHz, 1.2V, 15-15-15-36. with XMP it works at 3000MHz, 1.35V, 15-17-17-35. but during stress tests my CPU temps go up by about 7 degrees. it raises questions about the usefulness of XMP.
this is definitely stress testing, not regular use. don't worry about it.
SgtScream explains the DRAM specifications and performance changes during stress testing. They note the temperature rise under load and suggest caution about enabling XMP unless thermal constraints exist. The discussion highlights normal operation versus potential risks.