Ive always been curious as to why people want to overclock their RAM frequency so much, when realistically, the greater benefit would be lowering timings, which brings me to my question.
Is it possible to gain greater timings (by greater I mean lower, not higher, for those who don't understand) by Underclocking high frequency ram with stable timings? IE: bring a 2400 kit with 9/9/9/24 down to 1600, and reduce its timing to say 6?
And please, before alot of people jump in and harass me with regards to "Oh hes such a n00b who would want slower ram" - Even with 1600MHz, we do not saturate the full bandwidth in gaming or day to day use, that's why the gains from 1333 to 1600 are miniscule, ~1 fps (exc amd on-board gpus, or rendering), however, timings are of greater benefit.
Thanks for your time anyone that reads and contributes to this. And anyone else that would like to play around with their lower timing by all means do so at your own risk of course.
Really depends on the workload. Because the memory controllers are integrated into the CPU itself, it makes RAM latency less of an issue as well. RAM latency used to be big back when we had FSB.
So, in general, especially on Intel platforms, RAM is pretty unimportant past a certain point (which I'd say is 1333-1600 CL9). But you're right, if all you were doing is FPS gaming and basic tasks, and wanted to squeeze every last bit of performance out of your system, RAM latency would be of greater benefit. It's really not worth the fuss IMO because the gains are hardly noticeable either way.
I used to have 1866 CL9, but had a very noticable change in loading and also ~3fps from changing to 1600 CL6. And in comparison, i havent tested in my pc yet, but my brothers 1333 CL7 performed better than my old CL9 so even he doesnt want it. I was just wondering why everyone has a huge hype on frequency and not so much on timing.
Gaming: Timings give larger boosts per point of performance. The only thing I can think of frequency performance gains beating this are Heavy rendering loads or very large scale CAD files
Changing RAM speed and timings is a double edged sword.
When increasing the frequency of RAM, you are increasing the transfer rate to and from the memory.
Lowering the timings, changes how quickly it responds.
So which is better. Lowering your speed, but lowering response or Increasing speed and higher response times. Effectively doing either of this could cancel each other out. (okay I know 1333 memory can transfer at 10GB/s which is still F**king fast)
For example (low speed, low response) I can deliver a parcel in 4 days, but it only takes only a day for pick up or (high speed, high timings) I can deliver the parcel in 1 day, but it takes 4 days for it to be picked up.
If you are an FPS gamer, then yea you may see a benefit in doing this. But in reality, increasing frequency and lowering timings is the best option. (Pick up and delivery in 1 day :P)
My DDR3 2400 MHz RAM with timings 11,13,13,31 CR2 runs at 57.6 nanoseconds latency, only 0.1 nanosecond slower than DDR3 1600 at 9,9,9,24 CR2.
Unless you have DDR3 1600 at CL 7, which is insanely expensive and doesn't have fucks worth the bandwidth that 2400 MHz does, you aren't going to see a big difference in latency jumping up to higher frequency RAM because they are already optimized to run as fast as 1600 MHz CL 9
Thanks for posting this, it gives me a yardstick to compare by
This is my aida64 scores for my 32gb of kingston beast ram set to both 2133(loose timings) and the *old* 1600 ram that it replaced (which I used in my brothers build).
I have managed to tweak it a little faster than this but not by much... i was never sure if I was looking at an example of *good* or *bad*?
That's an older version of AIDA64 you are running, I'm running v3.00.2500 and you can't compare scores between versions. Actually if I remember correctly, that version doesn't test RAM the same way the newer version does. The newer version displays higher bandwidth scores. You may want to think about updating if you can then retest your RAM with v3. But yes those scores look about right for that version as I had that version before updating to v3.
Noticeable improvements in some games, like those that use the source engine not sure why. System is slightley more responsive, especially at startup.
Current clocks are 10-11-10 30, 2T. Been playing with subtimings since I hit a ceiling with frequency and I don't want to loosen current timings.
If your system is stable, then your RAM is good. Most don't like messing with RAM oc's because it takes time to find that setting that saturates bandwidth and maintain stability.
To further confuse the argument of RAM speed vs latency timings I thought I would share an interesting variable that the AMD side of the hardware argument has.
This is what 2400 MHz 11,13,13,31,45 RAM looks like tested on paper running at 2133 MHz 9,11,10,28,39 with the Memory controller overclocked to 2400 MHz instead of the standard 2135 MHz you would normally set it at.
Looks a hell of a lot like the ram throughput I posted earlier doesn't it? lol
Also, note that the original 2400 MHz test was done with half of the memory running in the machine as AMD memory controllers cannot run more than 2 DIMMs of RAM installed when the native speed is set to 2400 MHz. With this configuration I was able to run the full 16GB of RAM with similar speeds and latency as 8GB of 2400 MHz RAM.
All RAM settings are verified stable with Memtest86+ v5.00RC1 before testing.
