memory, what settings are genuinly faster?

SooTyLaD

SooTyLaD

SooTyLaD

Soldato
Joined
19 Oct 2002
Posts
3,480
been doing a fair bit of reading on memory in terms of what is better but there are so many conflicting opinions it just aint funny...

could someone tell me with my cpu at 378x9 what would be the optimum memory configuration i should be going for? lots of people say its best in 1:1, some people say as long as its linked its loads better than unlinked, some people say speed maters much more than timing...

what is the actual case? 4-4-4-12 @ 756MHz? 5-5-5-15@ 945MHz? 5-5-5-18 @ 1000MHz?

much more complicated that the cpu... please advise :)

I suppose i could bench the different variations but apparently it seriously depends on application also that that aint much good...
 
Intel systems tend to prefer mhz over timings.

5-5-5-18 @ 1000MHz should be the faster setting.

Tho a couple of runs of SuperPi or Everests memory benchmarks at each setting will confirm it.
 
how do you work out how low you can go? there are so many combinations being 4 numbers so which are the ones that if lower would really make a difference?

i've keyed in 3-3-3-8 at the mo and its blend testing at the mo so i basically picked those numbers by throwing darts if you know what i mean...

what is the proper methodology?
 
The first makes the most difference, followed by the middle two.

Tho pure mhz usually makes the biggest difference.

Which Cellshock is it? A lot of it will do 4-4-4-12 @1000mhz at 2.2v.
 
"CellShock 2GB (2 x 1GB) DDR2 PC2-6400C4 800MHz Dual Channel Kit"

i can't unlink the mem so its either 945 or 1134... but even with slacker timings of 5-5-5-15 the computer wouldn't post... even went up to 2.35V but no luck...

(*wondering why the hell i bought the best ram i could find*)

so i gave up and tightened the timings instead... if you have any advice on getting the clock higher i'd appreciate it :)
 
Have you tried 945mhz 4-4-4-12? I would have thought that that was possible with the CAS4 Cellshock.

And try raising your PCI-E Freq to 105mhz. I find it allows a higher overclock than 100mhz on my Rev 1.0 DS3.
 
blend testing 945 @ 4-4-4-12 now... with the PCI-E freq. @ 105MHz (why does this make a difference?)

the computer restarded itself after a few seconds over and over again but i managed to get into windows... but i've recently read this may well be the mobo and not the overclocking settings so if anyone knows a way to stop the dreaded cycle please let me know :)

i'll post it the above o/c is stable :) thanks Cob ;)
 
QuiKsiLVeR said:
blend testing 945 @ 4-4-4-12 now... with the PCI-E freq. @ 105MHz (why does this make a difference?)

the computer restarded itself after a few seconds over and over again but i managed to get into windows... but i've recently read this may well be the mobo and not the overclocking settings so if anyone knows a way to stop the dreaded cycle please let me know :)

i'll post it the above o/c is stable :) thanks Cob ;)
Click to expand...

I'm hoping it's a BIOS issue. The original DS3 suffered from boot issues with various BIOS's. It seems to be solved these days.

Arcane said:
Also, try running at 1t .
Click to expand...

The DS3, like the majority of P965 boards, can't run at 1T.
 
the blend test has failed twice now, once on stock volts (2.2) and again on 2.25... just upped it to 2.3 so we'll try again :(

i thought i was just my cpu but my whole system seems really voltage hungry - could this be all becuase of a bad mobo or summat?
 
failed again at 4 hours blend :( :( - gonna have to go for 2.4 volts now!

this really doesnt seem right... anyone know whats going on? i feel like i need frikkin mains voltage to run my components!
 
Are you certain its the memory thats causing blend to fail. It could be your cpu or even your memory controller.
 
QuiKsiLVeR said:
what is the actual case? 4-4-4-12 @ 756MHz? 5-5-5-15@ 945MHz? 5-5-5-18 @ 1000MHz?
Click to expand...

It's actually surprisingly easy to work out, as most things are.

The CAS figure is the rate limiting step and to go from CAS 4 to CAS 5 at the same memory frequency it will take 25% longer to look up the columns in memory.

At 756MHz CAS 4 you will get 189,000 opportunities to read the RAM every second, and with 945MHz CAS 5 you will also get 189,000 opportunities to read the RAM so on that basis they are equal. However, at the 945MHz setting you also gain bandwidth so that will be faster in real terms, no question.

At 1000MHz you will get the bandwidth, plus you will get an extra 11,000 opportunities per second to read or write data, so that will be fastest.

In practice, only CAS really matters as it is the rate limiting step, however setting everything else as low as possible will give performance enhancements when everything else is optimal.
 
w3bbo said:
Are you certain its the memory thats causing blend to fail. It could be your cpu or even your memory controller.
Click to expand...
the cpu was 12 hours orthos stable so i'd have thought so... the 2.4V blend just failed and i'm not really willing to go higher so i guess thats it... i bought the best ram i could find and am gonna be runnning it below stock at 756MHz :(

but cheers for that post WJA very interesting :)
 
At what speed and test was it 12hrs orthos stable? Is it just blend you are having problems with?
 
w3bbo said:
At what speed and test was it 12hrs orthos stable? Is it just blend you are having problems with?
Click to expand...
it was orthos stable at 3.4gig (378x9) at 1.5V - this was with the mem at 756MHz 5-5-5-15 (2.2v)

had some real problems just, the computer kept restarting to begin with, then when it finally went in it was freezing just before or after it posts the memory size... i managed to get in to the bios eventually and put it all back to stock while i think about it :(
 
Have you tried memtest? - I always find it a better test of the RAM as it takes everything else out of the loop - never had much luck with the blend test.
 
You must log in or register to reply here.
Back
Top Bottom