Affect of memory divider

[nvening]

What effect would a memory divider have on my system if the ram worked out to be running at just under 200mhz (say 190). Surely it does not have much affet as A64s for normal use do not need loads of bandwidth?

[BUFF]

http://www.madshrimps.be/?action=getarticle&articID=325

[NightshadowUK]

I ran my system with the RAM at 195MHz for a while and it was fine.

It's now up at 205MHz though which I suppose is better so if you can push the overclock further I'd do it.

[kalniel]

What effect would a memory divider have on my system if the ram worked out to be running at just under 200mhz (say 190). Surely it does not have much affet as A64s for normal use do not need loads of bandwidth?

None

[nvening]

If i made the memory run at 200 on a non 1:1 divider then it would still have the same bandwidth as a non overclocked processor, it would just the the clock speed increasing? So if i overclock to 2.6GHz which made my memory run at 200mhz then it would be exactly the same performance as an fx55 (same cache and core)?

[kalniel]

Yes

[nvening]

Ok i now have it @ 2.6GHz (237mhz htt) which makes the memory run @ 200 dead on.

Im getting temps of 40/41 after and hour of prime95. OK? (AC freezer 65)

Also i got an error in prime95 at 57mins, is that any good? I have heard prime95 can be unreliable though.

Also my PWM is hitting 51 degrees however after a quick google it seems thats fine?

I have it on 1.525 volts, apparently thats on the edge of safe, moving onto "some risk".

[NightshadowUK]

Aim for 55 degrees under load, anything less is a-okay.

As for Prime being unreliable, it isn't, but stability means different things to different people - some people consider playing BF2 for a few hours a worthy stability test. 24 hours of Prime however, is generally considered to be 100% stable.

I wouldn't like to go above 1.525v if I'm honest, so maybe your only option is to lower the overclock?

[Zak33]

clock speed is king...memory divider is not crucial.

Knock it down a tad to a nice stable 195 (if that's what you get as 100% stable)

then go play some games and stop fiddling

[nvening]

Well its not the memory thats holding me back its the the cpu (well actually me unwilling to push the voltage up) so why should i put it down to 195?

How stable do you think my system is then? As long as it does not crash every few hours its fine for me, it should be ok for games for long periods of time?

I have also overclocked my x800xt to a x850xtpe

EDIT: wow that gpu OC raised my 3dmark05 score by 700 points!

[Zak33]

sorry fella....read Nightshadows psost and thought it was yours

If it's stable, bingo. The importance is that the divider goes where it needs to go, in accordance with the highest multiplier you can get to make the highest core clock that's stable, and then it wouldn't matter too much if the divider made your ram a tad lower than it's maximum...so long as you'd got a nice high clock speed.

It's in effect very different to , say, an Athlon XP on an nForce 2 mobo, where, once you';d got the potentially highest CPU clock you could get, you'd knock the FSB as HIG as it would go, and if need be, drop the multiplier and even lose a fraction of a clock speed to improve buss speed.

[nvening]

Well the ram is at 200mhz which really makes the htt 200mhz and so running on a divider is almost quite like multiplier overclocking?

[NightshadowUK]

Have you tested the card using ATiTool's 'Scan for Artifacts'?

Minimum CPU stability test should really be a 32M run-through using SuperPi... better still a couple of hours Prime. An extra few hundred MHz isn't worth having an unstable system for.

[nvening]

I scaned for about 40mins and it found none.
Ill see about my result in super pie.

EDIT: 32m superpi had no errors

[NightshadowUK]

Have another bash using SP2004 set to 'Large' test - it's basically a more user friendly Prime95.

My last overclock was fine with SuperPi but would restart randomly after long sessions of BF2, so I tested using SP2004, it managed 18 minutes.

[kalniel]

Hehe that would be a tad too far then

Even though you're not overclocking the htt, try reducing the hypertransport multiplier to 4x rather than 5x (800 rather than 1000) - may help stability with no real loss of performance. Alternatively loosen up ram timings a bit. Even though you're not overclocking the ram, the CPU is getting info to them faster