Tight Timings VS Higher FSB

realscot

realscot

realscot

Associate
Joined
6 Nov 2005
Posts
2,317
Location
Aberdeenshire
ok I was bored so i thought i'd play with my memory.
It's 4x256 OCZ Premier with stock timings of 2.5-3-3-7-2T.
First off i set the bios DRAM voltage to HIGH, god knows how much volts that is.
I used Everest to run ram benchmarks read/write/copy/latency.

ok here's what i've done so far (the results will be as listed above)

Stock timings with DDR 333 divider (since i'm running 230FSB)
2.5-3-3-7 = 5396/3686/4272/61.9
2.5-3-3-5 = 5399/3686/4260/61.6 (so changing RAS doesn't make much dif)
2-3-3-6 = 5407/3684/4329/59.1 (Best so far though slight decrese in write)
2-3-3-5 = 5397/3680/4314/59.4 (confusing, actually worse with lower ras)
3-4-4-8 = 5335/3673/4212/66.0 (slack timings with DDR333 multi)
3-4-4-10 = 6777/4671/5163/57.5 (slack timings with DDR400 multi =DDR 460)

Right so i've found that my memory doesn't like it's RCD or RP running below the rated 3.. but it's happy to go as low as CAS2 and RAS5 with the DDR333 Multi giving 180MHz..... But when i tried to let the memory run at 1:1 with the CPU FSB of 230MHz (DDR460) i could only get into windows by changing the RAS from 8 to 10. I havn't tried to get the CAS RCD or RP down yet. I better go do some benching first to see if this is stable before I start tightening the timings. Still a massive difference in memory performance with an FSB increase rather than tighter timings. Whether this translates through to the real world of gaming or not has yet to be seen. I suppose I should run 3Dmark2001 at 2-3-3-6 DDR333 and compare it with 3-4-4-10 DDR460 see what the difference is.
I'll do that now, will get back to you...
Also, if anyone wants me to run another benchmark or post some pics let me know.
I DID SAY I WAS BORED lol
 
The Results

Ok I ran 3DMark2001 SE with both settings and the results are as follows.

2-3-3-6 @ DDR 333 = 25015 & super PI of 36s
3-4-4-10@ DDR 460 = 25993 & super PI of 35s

So it does appear that it is more important to have a higher FSB than Tight timings.
I guess the perfect scenario is to have both, if you can afford it.
But if, like me, you are stuck with what you have, then my conclusion would have to be, push your FSB rather than your timings for the best results.

I'd welcome any comments that anyone has.
 
Interesting results, I didn't think the difference was this large. Can you run 1T at all? As all the other tests I remember had a much smaller difference between high FSB/tight timings, but they were all at 1T.

Is your CPU speed constant throughout?
 
I can't run at 1T aparently cos I have 4 ram modules, or maybe it's the asrock board... Who knows. Anyway it seems to be stable and consistant throughout the testing.
 
realscot said:
I can't run at 1T aparently cos I have 4 ram modules, or maybe it's the asrock board... Who knows. Anyway it seems to be stable and consistant throughout the testing.
Click to expand...

Oh yeah, I missed that, you're right you're stuck at 2T :)

What about CPU speed, is it constant?
 
if by consistant you mean did I change the speed at all, then no it was the same through all the tests. Only the memory multi and timings were changed. Still sitting at 230x11 =2530MHz 1.4V 39C right now and running prime testing stability.
When playing with my overclock I was convinced that my CPU needed more volts to get a better overclock cos when i take my HTT above 230 i get random blue screens and that's with everything else locked down. Memory, PCIE Freq etc. Problem is, without a volt mod, i can't go above 1.4V on my board as far as i'm aware.

EDIT~ Just noticed cool and quiet must be on cos CPU-Z is saying my cpu is getting 1.35V
 
Last edited:
It has been known for a long time that 3DMark2001 likes speed > timings but this isn't true for all games/apps.

I think the tRAS needs to be equal or greater than the sum of the rest of the settings that's why you see 2-3-3-8 and 3-4-4-10
 
Can't say I'm overly suprised as I found in my testing that FSB IS important on the A64 platform, despite the common myth that timings are everything.

I think the fact that timings matter more on A64 than they do on Intel/Socket A platforms made some people assume that they are the be-all and end-all of performance tweaking, when they isn't really the case.

Yes, faster timings do give a nice boost in speed, but it depends on how much FSB you have to sacrifice. So while for example 210fsb with tight timings may outperform 220fsb with slack timings, something like 230fsb with 'average' timings may well be faster.

In the example given above, moving from slack timings to tight timings has meant going from 230fsb to 166, which is a HUGE drop. I certainly wouldn't want to be running DDR333 even if I could push my timings down a little.
 
You must log in or register to reply here.
Back
Top Bottom