fsb/memory ratio

Fallen · Post by **Fallen** » 09 Oct 2007, 19:15

Are there any pros or cons to having a certain fsb/memory ratio? I have overclocked my Kingston Hyperx Memory to 912Mhz 4-4-4-12 @ 2.1v and the ratio to my fsb is 9:14 (q6600 clocked at 2.64Ghz). Is it better to keep a 3:2 ratio for instance??

badguy · Post by **badguy** » 09 Oct 2007, 19:19

sorry that i am going to hijack this topic

But just need to know who does timings of ram work like 4-4-4-12 ???

I am stupid @ this

psYChowIt · Post by **psYChowIt** » 09 Oct 2007, 21:04

badguy wrote:sorry that i am going to hijack this topic

But just need to know who does timings of ram work like 4-4-4-12 ???

I am stupid @ this

I dunno either...

RobThePyro · Post by **RobThePyro** » 09 Oct 2007, 21:36

the timings of the ram are like, how long it takes to compleat tasks. so cl4= 4clockclycles etc.

And fallen, i would assume having your ram run at an "even" ratio to your FSB would be best, I.E. Ram/Fsb 1:1 or 2:1 or 1:2 etc, where they divide nicely, should give you a small performance boost over something like 1:3 or 9:14 where they dont divide nicely

Rob~!

douglash · Post by **douglash** » 10 Oct 2007, 07:56

Agreed! it just means that everything runs in sync therefore faster...

Post by **maxxis** » 10 Oct 2007, 08:38

Say for instance I run my quad @ 400 x 9. Which will give better results?

DDR2-800 @ 4-4-3-9

or

DDR2-1000 @ 4-4-3-9

The latter should give better memory bandwidth or am I wrong?

douglash · Post by **douglash** » 10 Oct 2007, 11:47

The latter should give better memory bandwidth or am I wrong?

yes - that is correct.

DDR2-1000 @ 4-4-3-9

if you are using the same latencies but higher speed of memory, it will definitely be better/faster...

when you start loosening the timings in order to gain speed(mhz) then it evens out...

so to sum it up, the faster you can run the ram on the lowest possible timings with be the best possible solution...

Post by **maxxis** » 10 Oct 2007, 12:19

You dont want to see my timings at the moment.

Im running

DRAM Frequency: DDR2-1000
DRAM Timing Control: Manual
CAS# Latency: 4
RAS# to CAS# Delay: 4
RAS# Precharge: 4
RAS# Activate to Precharge: 4
TWR: 3
TRFC: 30
TWTR: 3
TRRD: 3
TRTP: 3
DRAM Static Read Control: Disabled

And its stable

douglash · Post by **douglash** » 10 Oct 2007, 12:32

WOW - what voltage are you running it at and what RAM is it?

DoOb · Post by **DoOb** » 10 Oct 2007, 12:40

maxxis wrote:You dont want to see my timings at the moment.

Im running

DRAM Frequency: DDR2-1000
DRAM Timing Control: Manual
CAS# Latency: 4
RAS# to CAS# Delay: 4
RAS# Precharge: 4
RAS# Activate to Precharge: 4
TWR: 3
TRFC: 30
TWTR: 3
TRRD: 3
TRTP: 3
DRAM Static Read Control: Disabled

And its stable

Nice going Maxxis,

@ Doug ---- Maxxis bly met 2gig Team Xtreme PC8000

douglash · Post by **douglash** » 10 Oct 2007, 12:50

Nice - very nice... i must try do better with the timings and stuff on my ram, just haven't had the time lately...

you know, the whole work thing getting in the way again...

Post by **maxxis** » 10 Oct 2007, 12:54

Team Xtreem PC 8000 Micron D9xxx @ 2.2V

Stable with the CPU @ 3.6 400 x 9

Yields a 14.03 1M Pi run

I34z1k · Post by **I34z1k** » 10 Oct 2007, 13:27

Yo maxxis. You gotta remember, I got as much bandwidth 1:1 @ like 450 - 500mhz than I did with my ram @ 620c4. Different dividers yield different results. As a general rule of thumb, a new set of ram should require around 30 hours of testing to find the best results

Post by **maxxis** » 10 Oct 2007, 13:53

Scary thought

RobThePyro · Post by **RobThePyro** » 10 Oct 2007, 14:39

Lol maxis, im running my ram same

but mine is only PC6400 Rated at DDR2-800 CL3 and im running 2.4V LOL!...

Rob~!

Fallen · Post by **Fallen** » 10 Oct 2007, 20:59

Ok, next question: which would give the best performance increase, keeping the timings low resulting in lower mhz or maxing the mhz on standard 5 timings???

RobThePyro · Post by **RobThePyro** » 10 Oct 2007, 21:09

Prob maxing out the mghz, assuming you dont go any higher than CL5 timings. i asked the same question the other day, check it out Here

Rob~!

I34z1k · Post by **I34z1k** » 10 Oct 2007, 21:09

RobThePyro wrote:Lol maxis, im running my ram same but mine is only PC6400 Rated at DDR2-800 CL3 and im running 2.4V LOL!...

Rob~!

Bud, D9s = no go. Lower the voltage a bit. Gates are closing. In a few months those sticks will be dead. Only use v that high for benches.

RobThePyro · Post by **RobThePyro** » 10 Oct 2007, 21:15

ya ya ya.... lol

whatever you say buddy

Rob~!

I34z1k · Post by **I34z1k** » 10 Oct 2007, 21:25

Lol. Funny boy. We'l see how funny it is in a couple months. ROFL. Go look on xs. You will see.

Post by **maxxis** » 10 Oct 2007, 22:33

Without active cooling on the ram stick to the rated Volts.

My set is rated 2 - 2.35 I think.

Need to double check.

Post by **KillerByte** » 11 Oct 2007, 00:46

I34z1k wrote:Yo maxxis. You gotta remember, I got as much bandwidth 1:1 @ like 450 - 500mhz than I did with my ram @ 620c4. Different dividers yield different results. As a general rule of thumb, a new set of ram should require around 30 hours of testing to find the best results

well in my case, i can just ask you. lol

I34z1k · Post by **I34z1k** » 11 Oct 2007, 06:55

maxxis wrote:Without active cooling on the ram stick to the rated Volts.

My set is rated 2 - 2.35 I think.

Need to double check.

Well with d9s even with cooling its dangerous. Check it out

@Kb

Especially on your board

Post by **maxxis** » 11 Oct 2007, 08:47

So even if the ram is rated to operate at 2.1 - 2.3 like my set is, its not a good idea to give it some?

DoOb · Post by **DoOb** » 11 Oct 2007, 10:03

Yip thats what he told me, can kill ram, the most ive put into my ram was 2.4v,