DDR2 RAM timings?

Status
Not open for further replies.

Alexexex

Posts: 105   +0
I'm just wondering what an ideal DDR2 RAM timing setting would be with my computer. Currently I'm running them at 4-4-4-12.

Thanks, Alex.
 
just go with what it says on the sticker; use the corresponding voltage too. if the modules don't have stickers just use the by SPD option or Auto setting in BIOS. there is no ideal settings, memory kits differ by brand and speed.
 
If you are talking about the XMS2 Corsair's you have in your sig then 4-4-4-12 is about as good as it's going to get - I'll assume they are PC2-6400 (DDR2-800).
 
Thanks for the response.
So... I checked the timings online and they come up as 5-5-5-18... and I'm just wondering since they're "supposedly" optimized to operate at this timing, would 5-5-5-18 be faster than
4-4-4-12? (Despite the higher latencies.) Or does that not matter? Oh and would you recommend overclocking the memory to get some more preformance cause I would like to make the most of my aging machine.

Thank you very much, Alex.
 
If its supposed to be 5-5-5-18, then they're already overclocked.
This is just a guess, but you could probably get to 1066 with 5-5-5-15 timings.
 
its kinda rare to overclock RAM without overclocking the CPU.

Higher latencies will mean a slower machine, but I doubt anybody will be able to tell the difference. Any difference will only show up on benchmarks, and it will be quite slight.

You'd do better going with as loose timings as you can possibly get, and increase the speed instead. 1066, 1200, whatever you can get, as long as its stable. Run Prime95 or memtest as long as possible just to be sure (at least overnight, I'd be happy with 24 hours).
 
I kinda did overclock my CPU.

I tried increasing the speed and everytime, I had to reset bios... which meant flipping the computer and jabbing at the jumpers. Kinda annoyed me so I just set the latencies.
If you know how to increase the speed without making me reset bios, I would really like to hear it, but I guess I should put the latencies back to the stock since I don't think RAM does much to preformance. I ran stress tests "Everest" for about 6 hours, then I was REALLY bored so I stopped it and started playing.

Uhm does anybody know the difference between lowering the latencies and setting higher speeds? This is all so confuzzling! :X


Thanks, you guys are awesome. Alex.
 
RAM performance is a sliding scale- setting for increased bandwidth (speed) will invariably call for more relaxed timings. How much the timings need to be loosened depends upon the quality of the chips under the heatspreader.
Any DDR2 RAM running faster than PC2-6400 (DDR2-800) is in effect overclocked. You can obviously overclock the RAM without altering the CPU settings via memory dividers, although they offer very limited options.
Performance of speed versus latency at the speeds (and RAM) you have is going to marginal. XMS is Corsair's standard RAM offering. Better binned (performing) memory chips usually end up in the Dominator series.
For all intents and purposes DDR2-800 at 4-4-4-12 is going to offer much the same performance as DDR2-1066 at 5-5-5-15. Getting the best out of thr RAM usually means running the RAM at 1:1 ratio and/or getting as much out of the CL4 latency as possible.
i.e. if you're running at 3.7GHz then your FSB is 352. Running in synchronous mode will produce DDR2-704 -not great as I sincerely doubt you could get the latency down to make the RAM perform. Running on the 5:6 divider should however net you DDR2-845. If you can produce this at 4-4-4-12 then this should offer the best compromise between bandwidth and latency. Bear in mind that because the RAM you are using is fairly mainstream that you will more than likely need to relax tRFC and possibly tRD (sometimes called Static Read Value or Performance Level). If you have everything set to Auto then I'd suggest using Memset to find out both the optimal (by Corsair's standards) settings and what you are actually running - here's an old post/screenshot I posted that shows the information. https://www.techspot.com/vb/post839493-2.html
I would not recommend changing RAM values using Memset -use the BIOS.
Running SuperPi will give you an indication of RAM performance.
 
hmm.... there is quite a significant difference between 1:1 and some weird ratio apparently. You might be better off running stock 1:1, than overclocking it a little and running 5:6 or something.

Go ahead and try it out, and run some benches and see. Get back to us so we know :D


And yeh, go ahead and run SuperPi, its a small enough program to download and run :D


Btw, whats the overclocks and ratio at?
 
Uhm, sorry but I don't have a 5:6 ratio... I have 1:1, 5:4, and 3:2
I forgot to mention my settings were 4-4-4-12 with two commands per clock.
I used CPU-Z and the "optimal" settings came out as 5-4-4-17
How much do you think adjusting these setting will affect the hardware itself?
Right now I'm about to try the 1066 5-5-5-15 settings recommended by HelloKitty.
AHH I tried the 1066, and I got the "no-go" beep!!!!!!! Jumper here I come for you. I'm assuming that I am not able to change memory clock speed.
So... what timings would you say is optimal without the ability to change speeds? At auto ratio, the RAM comes up at 1057.5 Mhz.
I'll be back in the afternoon gotta go to school.
Thanks, Alex
 
I got this feeling that you might not be setting 1066, but inadvertently setting it at double that instead....

Okay.... at auto ratio, what is your CPU speed?
What happens at 1:1 ratio? (CPU speed and RAM speed)

I wouldn't worry about timings at all, they basically don't do much at all. I spent a considerable time mucking about with timings when overclocking my current computer, but benching showed that between slowest, and fastest setting that I can get, I don't even get 0.1% reduction in SuperPi time.

Given this info, decide if its worth the effort. Might as well be trying to get a better overclock on your CPU....
 
AHH I tried the 1066, and I got the "no-go" beep!!!!!!! Jumper here I come for you. I'm assuming that I am not able to change memory clock speed
I got this feeling that you might not be setting 1066, but inadvertently setting it at double that instead....
Maybe I didn't quite make myself crystal clear...
1. Your RAM is bog-standard-bin. 1066 is a big ask on mainstream/budget sticks- made worse by the fact that...
2. Your board Asus P5N73-AM only officially supports DDR2-800, and that through overclocking. This is why I suggested a 1:1 divider or as close as you can get. At the moment you seem to be using a 3:2 divider
 
FSB Mhz, [1410]
MEM Mhz [705]

1:1 gives me 705 mhz? Isn't that underclocking?

Would I be better off leaving my RAM at stock? :O

My new settings are 845mhz - 5-5-5-15 I'll just wait for a response before jacking around with the settings anymore.

@dividebyzero - Thanks for pointing out to me clearly the limitations of my RAM, Its good 845.

Thanks peoples.
 
1:1 by definition is CPU running at the same speed as your RAM.
Your FSB (1410) is what is known as "quad pumped" (Quad Data Rate or QDR- 4 bits of info per CPU cycle)-so your Core frequency is a quarter of the FSB value - 352. Your RAM frequency is expressed as DDR, or Dual Data Rate, which is double the actual clock rate of the RAM ( 705 / 2 = 352).
DDR2-800 at 4-4-4-12 would be slightly better performing than DDR2-845 at 5-5-5-15. If you could keep the 845 at 4-4-4-12 you would probably have the best of both bandwidth and latency, although as I pointed out earlier, you may need to tweak the tRFC and tRD values (it may also need you to slightly raise the RAM and/or MCP voltage for stability) assuming the RAM could get to those timings.
Sorry to be the bearer of sad news, but not all RAM chips are created equal. The good news is that good RAM chips cost more for what is largely a very small gain in performance.
 
Oh okay, thank you very much.
Maybe I'll revert back to the 800 mhz at 4-4-4-12 since it ran stable tests fine.
Although if it isn't too much to ask,
What does Commands Per Clock mean? My settings had it at two, although I'm not entirely sure what it means. It sounds self explanitory but... not really.

Thanks man, Alex.
 
Command Rate is actually the delay (in clock cycles) between when a RAM memory chip (IC) is selected to when the command is issued to the chip. 1T is usually associated with DDR RAM (DDR-266, -333, -400) which has an low overall latency (CL2-3). DDR2-533,-667, (and rarely -800) can also achieve 1T- although this is the province mainly of nVidia chipsets. DDR2 and 3 RAM in general has looser timings so in general needs a 2T rate to properly execute the instructions.
 
Just a little factoid to remember as well about RAM delay, it is measured in clock cycles.

This means the delay for DDR 400 RAM 1T is the same as the delay in DDR2 800 RAM.


For some reason, knowing this does make me feel better buying DDR3 with their "incredibly high" 9-9-9-21 clocks or something....
 
Status
Not open for further replies.
Back