F5F Stay Refreshed Hardware Desktop Updated Ram "Flck" & adjusted schedule, give it a shot!

Updated Ram "Flck" & adjusted schedule, give it a shot!

Updated Ram "Flck" & adjusted schedule, give it a shot!

Pages (2): Previous 1 2
G
gekkie12345
Junior Member
18
06-11-2016, 08:12 AM
#11
So that's a very loaded question. In most very high FPS games (CS:GO, for example), they want as low memory latency as possible. However, with Ryzen especially but memory in general, memory frequency is directly tied to latency. Basically all the timings are measured in clock cycles, so while 3200MHz CL16 might have a lower CAS latency than 3600MHz CL18, the 3600MHz kit has the same first word latency as the 3200MHz kit but has more bandwidth since it has a higher frequency. Plus with Ryzen, the core interconnect (and by extension a portion of the memory latency) is directly tied to the memory frequency, so even if the effective latency might be the same, the higher frequency kits will have lower latency. Basically, a faster memory kit would improve performance (I can give exact numbers with my 5900X tomorrow if you'd like, I'm currently running a very tuned kit of memory, 3800MHz CL16-13-13-21 with tuned subtimings and can compare the results in CS:GO against 3200MHz CL16-20-20-38 if you want to know just how much it affects performance) Also, the people saying that you want to lower the CAS latency in order to reduce the latency are only 20% correct. The CAS latency is the amount of time it takes to start reading memory, but unless what you're reading it incredibly small, there are so many other subtimings that have a bigger affect on it than CAS latency. You would have a bigger performance improvement by setting some of your subtimings than you would by adjusting CAS latency. There's a reason I'm pretty OK with having my CAS latency at 16 instead of 14 that similar memory kits do run at is because in most scenarios it doesn't affect performance all that much.
G
gekkie12345
06-11-2016, 08:12 AM #11

So that's a very loaded question. In most very high FPS games (CS:GO, for example), they want as low memory latency as possible. However, with Ryzen especially but memory in general, memory frequency is directly tied to latency. Basically all the timings are measured in clock cycles, so while 3200MHz CL16 might have a lower CAS latency than 3600MHz CL18, the 3600MHz kit has the same first word latency as the 3200MHz kit but has more bandwidth since it has a higher frequency. Plus with Ryzen, the core interconnect (and by extension a portion of the memory latency) is directly tied to the memory frequency, so even if the effective latency might be the same, the higher frequency kits will have lower latency. Basically, a faster memory kit would improve performance (I can give exact numbers with my 5900X tomorrow if you'd like, I'm currently running a very tuned kit of memory, 3800MHz CL16-13-13-21 with tuned subtimings and can compare the results in CS:GO against 3200MHz CL16-20-20-38 if you want to know just how much it affects performance) Also, the people saying that you want to lower the CAS latency in order to reduce the latency are only 20% correct. The CAS latency is the amount of time it takes to start reading memory, but unless what you're reading it incredibly small, there are so many other subtimings that have a bigger affect on it than CAS latency. You would have a bigger performance improvement by setting some of your subtimings than you would by adjusting CAS latency. There's a reason I'm pretty OK with having my CAS latency at 16 instead of 14 that similar memory kits do run at is because in most scenarios it doesn't affect performance all that much.

Pages (2): Previous 1 2