The speed of RAM and CAS latency can make a noticeable impact on performance.
The speed of RAM and CAS latency can make a noticeable impact on performance.
Hello friends, I'm planning to upgrade my current 2x8 DDR4 2666 RAM and had a query about RAM choices. I remember that for RAMs with the same clock speed, you should pick the one with lower CAS latency. I want to keep using Gskill (the Trident NEO sticks 2x32) and aim for a speed of 3600. My question is, will switching from such slow RAM to a faster version—like 2x16g at 3600 at CAS 18—make a noticeable difference compared to a 2x16g at 16? They’re about a $40 price gap. Thanks in advance!
It depends on your CPU platform. For Ryzen processors, you'll generally face more bandwidth limitations, but tighter timings can help improve performance by around 5%. With CL16 kits versus CL18 kits, the difference is noticeable, though it varies with workload. On Intel systems, memory bandwidth is less constrained, so the benefit is smaller. If you're building a new system, opting for the faster kit adds only about $40 to the total cost, which is usually negligible. But if you're just upgrading, the extra $40 may not justify it.
Thanks! I'm also upgrading my CPU from a Reyzen 5 2600 to a Ryzen 7 5800x, so I think I'll choose the more affordable one.
This idea doesn't quite align, except perhaps by considering Ryzen's higher core count compared to older Intel designs. Ryzen prefers faster RAM clock speeds since the connections between cores work best when synchronized with those speeds, while latency shouldn't impact performance. Computation tasks will rely on a consistent memory bandwidth, regardless of whether they're from AMD or Intel; in fact, AMD could potentially use slightly less bandwidth due to its larger L2/L3 cache.
It works well with higher frequency first, though tighter timing still helps. The bigger cache acts as a buffer, but when tasks need files larger than the L3 size, speed noticeably drops. I own a 5900X and a solid B-Die setup, allowing me to run quick benchmarks—using a 3600CL16-16-16-36 configuration and a 3600MHz CL18-22-22-42 kit. The processor ran at 4.2GHz, though I avoided focusing on core performance. I tested with 7-Zip and Y cruncher; both are fast and memory-intensive. Y cruncher slowed down noticeably, showing a clear improvement from the memory setup. 7-Zip saw a modest gain, but overall scores improved by roughly 2%. The same tests at 3600-16-19-19-39 gave results around the average.