timing vs OC speed

bewitcher

Distinguished
May 18, 2009
8
0
18,510
ok soooo i'm going back and forth on something and a friend and i both are scratching our heads. Say i have 1600 ddr3 at 9-9-9-24 CAS 9, and then i have 2400 ddr3 at 10-12-12-31 CAS 10. in a gaming machine am i going to see a big difference between the 2 on the ASUS MAXIMUS VI HERO with a 4770k?
 
Solution
Using separately would guess maybe up to 5 FPS depending on the games, DRAM doesn't have much aeffect with gaming as it's mainly just used as a conduit for data, the 2400 however will offer much better performance in multi-tasking and mamory centric apps like video, imaging, VMs, CAD and simply multi-tasking
From what I found, you can take the speed of ram and divide by 2, * 1000 and then * by the CL.

So

1/800 * 1000 * 9 = 11.25ns

1/1200 * 1000 * 10 = 8.3ns

That doesn't take into account all the other timing values or how much the 12-12-31 vs the 9-9-24 affects it, or how that ~3ns difference means anything.

http://www.pugetsystems.com/labs/articles/Breaking-the-Hype-of-High-Frequency-RAM-142/

They didn't seem to think so.
 


You will not see significant difference in terms of FPS in games using both RAMs, though I would like to stick with lower CAS Latency for gaming, which can be done manually for 2400MHz RAM.
 

Tradesman1

Legenda in Aeternum
Using separately would guess maybe up to 5 FPS depending on the games, DRAM doesn't have much aeffect with gaming as it's mainly just used as a conduit for data, the 2400 however will offer much better performance in multi-tasking and mamory centric apps like video, imaging, VMs, CAD and simply multi-tasking
 
Solution


Yup, that's correct.

DRAM latency is measured in IO bus cycles, and data is transferred on the IO pins on both the rising and falling edge of the reference clock.

What's often not mentioned is that DDR3-SDRAM has 8 individual banks to work with, so latency can be masked by a decent memory controller as it will just flip through the banks while waiting for the memory operations to complete..