7800x3d/9800x3d - ram latency/performance

[Ichigo16]

Hello

 

I wanted to ask someone in the know if the gaming performance of 7800x3d/9800x3d will respond to ram timings specifically CAS latency.

From my understanding there's little benefit going beyond 6000/6400Mhz as far as frequency goes, what about CL? Will going from CL30 to CL26 impact performance at all in avg or 1% lows? 

 

 

[SorryBella]

1 minute ago, Ichigo16 said:

Will going from CL30 to CL26 impact performance at all in avg or 1% lows? 

The price of CL26 is still too high and most CL30 kits (particularly those that currently only use SK Hynix as their DRAM vendor like G.Skill Flare X5) can still go down to C28-27 with some of your own overclocking if you so inclined. But even without it, CL30 is the spot for productivity and gaming consistency, not that much more expensive than a C40-36 kit, but provides you with far better experience.

 

 

[Somerandomtechyboi]

cl is a garbage timing performance wise doesnt do jack shit for performance

 

its only use is for ic binning and most of the performance diff going from 6000c36 to 6000c30 is in tighter subs auto set by board as hynixes are just better than current micron/samsung trash, they pull even farther ahead with tuning and theres already easy to follow guides like buildzoid hynix timings

 

6000c30/32 are unbinned as far as i can tell (similar oc abilities to oem hmcg78agbua081n 5600c46 a die) with 6000c28 being an idiot trap and 6000c26 being properly binned rams for those who want to push some crazy ocs (ex 8000c28 >1.7v) though performance diff is probably not gonna be too noticable even when maxing them out so its more of an enthusiast thing then a practical thing

 

just buy the cheapest 6000c30/32 and run buildzoid timings and thats 90-95% of the way there performance wise

[Nocte]

Only look for first word latency. CL doesn't mean anything without considering the speed as well.

Consider:

(CAS Latency * 2000) / MHz

and get those that are at least equal or below 10ns. I wouldn't go below 6000MHz since the cheapest models usually get up to that speed.

 

For example, 6000MHz 30 CL are the same in terms of first word latency to 7600MHz 38 CL.

Zen 5 performance is mainly tied to first word latency. Using CAS latency is a wild over generalisation, otherwise DDR4 - or even moreso DDR3 - would be better than DDR5.

[Somerandomtechyboi]

1 hour ago, Nocte said:

Only look for first word latency. CL doesn't mean anything without considering the speed as well.

Consider:

(CAS Latency * 2000) / MHz

and get those that are at least equal or below 10ns. I wouldn't go below 6000MHz since the cheapest models usually get up to that speed.

 

For example, 6000MHz 30 CL are the same in terms of first word latency to 7600MHz 38 CL.

Zen 5 performance is mainly tied to first word latency. Using CAS latency is a wild over generalisation, otherwise DDR4 - or even moreso DDR3 - would be better than DDR5.

i havent really gotten into performance tuning yet as ive mainly been screwing around with high frequency stability and wont bother tightening subs till i hit imc wall which i still havent managed to do with quad stick ddr3 3300c13 on my currently out of commision z97x soc and my rams all suck so i cant loosen the cl that much compared to the tightest cl i can run so any data id have would be within margin of error with a meager 2-3cl difference

 

even then i still know damn well that its the subs that matter and not the mostly useless primaries most important of which being trcd and not the useless cas latency (i still hunt low cl anyways for the fun of it and as an excuse to run high voltage but if i were more practical in my overclocking i wouldnt)

 

cl first word latency all buzzword nonsense that really only helps when you are benching and only by a tiny bit at that so real world performance wise you wont even see the difference, ask any other competent mem ocer and theyll say the same thing with the person in the video i linked being a very competent mem ocer considering the entire channel is just mem oc videos =p

 

btw the 7600c38 will be slower by default, heck even if it were 7600c28 (doable with those binned 6000c26 sticks albiet >1.6v) itd still be slower due to the desync uclk latency penalty, only reason to run high freq on the cpus is for lower vsoc as the soc does draw quite abit of power