DDR4 CAS Latency Performance Gains / Overclocking

[Tyerker]

Long time, no post. But I found a deal on a B660 motherboard I've been eyeing for months, and I couldn't pass up. It's DDR4 support, and from what I heard at least early on, DDR5 was very finnicky with 4 sticks, and has been crazy overpriced.

Speaking of overpriced, as I'm looking for DDR4 memory to pair with this new motherboard, I noticed a pretty stark price difference:
G Skill Ripjaw V 2x16GB DDR4000 18 CAS - $110
G Skill Ripjaw V 2x16GB DDR4000 16 CAS - $200

I know in the past, memory manufacturers would upcharge for MHz quite a bit, but this has me wondering if there is an engineering / performance reason a couple CAS Latency points is almost double the price. Does RAM still behave like in the past where you can increase voltage and manually set timings? Or is it pretty much buy the speed you want and trust XMP to make it work?

I bought my current PC towards the launch of DDR4 platforms, and it has honestly held up perfectly. But I couldn't pass up this motherboard deal, and figure it's time to properly research the rest of the components.

 

[Blaylock]

Yes, RAM still overclocks the same as in the past, though the gains are less noticeable in game applications. If you're looking for a kit that will run stable at 4000 without overclocking then the 18CAS will be fine. You will not be able to overclock much on that kit though as it is at the top of the Hynix-M die limit (assuming I have the IC correct).

TLDR: If you're looking to set it and forget it the CAS18 is fine. If you are looking to tinker I would look at different kits.

 

[Haider]

It all kind of depends look at 5800X3D; allow to explain when your CPU has to go memory for data that's when RAM speed & latency will come into play to give you an analogy you notice pension when you retire. During your working life it's there but does nothing for you till you retire and you need it. Fast RAM will help lower stuttering or make it less worse...

 

[Zerileous]

Generally specific benchmarks are preferred to sweeping generalizations. Each platform and application will behave a little differently. The difficulty of course, is, has the title you like to play been benchmarked and if not what similar titles can be an accurate representation.

I believe 1% lows is a decent representation of stuttering in benchmarks.