Questions regarding RAM CAS Latency

RainOfPain125 · October 17, 2017

I know I don't need RAM with over 2400 MHz or so unless I wanna be some pro mlg 5000 fps gamer, but I'm just on a small budget and I'm aiming for a good 60 fps constant in any game with the lowest settings, BUT, I also want to host servers in the background so I've settled my mind of getting a slightly overkill CPU (overkill in the sense I just wanna play games at the lowest settings) (Ryzen 5 1600) and I've settled my mind on getting 2x8GB's of RAM (I'll need the extra GB's for the servers)

But from what I've heard the higher MHz for RAM is good for games to relieve bottlenecks amongst the CPU / GPU, but from what I was told the CAS Latency is a deal when it comes to workloads (like me hosting servers)

And since RAM prices are kinda skyrocketed I'd like to spend my money as best as possible. Should I get low CL RAM or.. whats a good in-between (preferably 2400 MHz and 2x8GB DDR4) between CL and price that I should buy? Is it wrong that CAS latency effects workloads more? etc..? Would it effect my FPS a lot if I got 2133 MHz and low CL instead of 2400 MHz and higher CL?

edit: here ya go https://pcpartpicker.com/user/RainOfPain125/saved/8vNPsY the RAM I picked was purely random, so just ignore that

Edited October 17, 2017 by RainOfPain125
Added linkkkkk

Cyracus · October 17, 2017

with ryzen you want fast ram, preferably at least 3000mhz because core cpu functionality is linked to ram speed. ram speed and cas latency are the two main factors in your rams response time. Well worth the extra $9 https://pcpartpicker.com/product/ZskwrH/gskill-memory-f43000c15d16gtzb

Carclis · October 17, 2017

Both low CL and high frequency are just as important for CPU performance. The only think that might be important is that frequency adds with it extra bandwidth which scales with infinity fabric but then again that may only help you in use cases where you are bandwidth limited. If your budget allows I would opt for a 3000Mhz+ kit of RAM just because it's easy but if you opt for one of the cheapest kits you could probably overclock the RAM to 2666/2933Mhz with decent timings.

The kit you have selected actually looks decent. If you settle on that one I'd bump the voltage up to 1.4v and see how far you can push it. Otherwise a decent alternative is this one which should be a bit faster assuming you can get it working at its rated speeds.

Evanair · October 17, 2017

I have some a 8GBx2 cl12 2400mhz kit sitting in a drawer I originally got for Ryzen (day 1 purchase) which I replaced for 4266mhz c19 (running 3333mhz at c14) because it runs a ton faster, even at 3200 c19, it's faster.

I'm working on benchmarks (going to run both sets and compare numbers from Ryzen launch) on the c12 vs c19 at different speeds and see what numbers I get this week. Also OC stuff and see how bios changes have effected it.

Anyway, anecdotal but hope it helps.

Santo22 · October 17, 2017

I always try to aim for the Lowest CL I can find in my price range and the tighter timings the better. A CL14 is probably the best bang for buck. Most of what you will find will be CL16 though. It's not going to make a HUGE difference, but just stick with something 3000 or up. I am running TridentZ 3000 CL15 at 16-16-16-35 and actually increased my Cinebench score by almost 100 point vs the default settings.

w/Ryzen 7 1700x

SteveGrabowski0 · October 17, 2017

CAS Latancy is not the number to base your purchase on. CAS Latancy is the number of clock cycles of delay, but latency as you actually experience in doing the operation CAS latency measures is a time delay. And that time delay is inversely proportional to the clockspeed. So CAS 12 DDR4-2400 runs at 1200 MHz = 1200 Mega cycle per second = 1200 * 10^6 cycle per second and the latency would be

t = (12 cycle)/(1200 * 10^6 cycle/second) = 10^(-8) second = 10 nanosecond

While for CAS 15 DDR4-3000 (which runs at 1500 MHz) the latency would be

t = (15 cycle)/(1500 * 10^6 cycle/second) = 10^(-8) second = 10 nanosecond

The real number to base your RAM purchase on is CAS latency divided by clockspeed. Lower is better since CAS latency divided by clockspeed is proportional to the time delay (specifically, the time delay doing the column address strobe).