Question Lower CAS Rating vs higher RAM Speed?

[Super Spartan]

I have 2x32GB 5600MHz DDR5 Kingston FURY Impact memory on my Alienware m18 AMD which runs at 5200 MHz only but at a lower CAS Rating of 38 vs 40 as it shows in CPU-Z.

I just saw a post on Reddit from a guy who said:

I have an Alienware M18, it came with 13900Hx, rtx 4080, 64gb ram 4800 Mhz and 1tb ssd. I have upgraded the ram to G.skill ripjaws 64 GB 5600 Mhz CL-40 and it works at rated speed of 5600Mhz on XMP 1, stock. Only people having trouble running their 64Gb 5600 kits I have seen are those that bought Kingston Fury impact 64gb 5600 or crucial ones, max they could run it was 5200mhz.

This got me thinking, what would be better for snapinness/speed, a lower CAS rating of 38 or would one feel the speed difference if they got RAM that run at 5600MHz albeit at a slightly higher CAS Rating?

 

[WelshBloke]

This I think!

Absolute latency = (CAS latency*2000) / Frequency

 

[dave_the_nerd]

Latency vs. Throughput is a real tradeoff that shows up in synthetic benchmarks.

A 5% drop in latency vs. a 10% increase in clock speed (or some middle-of-the-road compromise between the two extremes) for RAM is unlikely to be noticeable by a human.

If you have specific types of workloads that are bottlenecked by memory access speed, then it makes sense to worry about it.

Also, Threadripper owners - the CPU dies talk to each other over a bus that's clocked the same as memory speed, so higher throughput tends to offer better multiprocessing performance.

 

[VirtualLarry]

I've got DDR-6000 expo CL 30-36-36 32GB kit in my Ryzen 7600 rig, well, used to be mine, lol.

 

[dave_the_nerd]

I remember back in the Athlon XP days - single digit CL numbers, as low as 2, iirc. That actually did make a noticeable difference for some things.

I suppose with the comparatively tiny on-chip caches back then, memory bandwidth was a bigger bottleneck for more applications.