250 FSB...OK, 251 FSB...Barf? Memory Question

[evil4blue]

This question is to more or less quench my curiosity.

 

I can run my memory at the timings shown in my sig stable, no memtest errors after 10+ hours, no Prime95 errors after 9+ hours, all other benchies run fine except for the 32M calc in SuperPI (which throws out some strange error about iteration 13 or 140).

 

Anyway, if I up the fsb just one stinking MHz, memtest generates thousands of errors, even with loose timings. I'm happy with the 250 FSB, but why is there such a dramatic effect from just 1 MHz.

 

BTW, I'm currently in a memtest burn loop, and plan to let it go for >24hrs.

 

Any info would be appreciated.

[Guest JaY]

that's bh5 for you, i get the same thing with my ocz vx

 

edit: try 2-2-2-6 for your ram, also you may want to try dropping the voltage. SuperPI 32m took longer to finish at 1.5-2-2-5 than it did at 2-2-2-6 for me...also found it performed better when i dropped the voltage to 3.2 after i had been using it for a few weeks/burned it in some more, i used to have it at 3.4v

[evil4blue]

Thanks, I plan on dropping the voltage after this burn-in session.

 

I've tried 2-2-2-6 to 2-2-2-15 but it was all a no go. I have also tried 2-3-2-6 and 2.5-3-3-8 just to see, but nothing works. Thousands upon thousands of errors.

 

I find it funny the the treshold is exactly 250.

 

Could drive strenght have any real effect. Mine is set a 8, but I'm willing to try 6 or something else. Secondly, would bios make any real difference. I can try the 414 bios.

 

I will do more experimentation in a couple of days; I was just looking for ideas.

 

For now my RAM will burn-baby-burn.

[Guest JaY]

i have my drive strength set to 6, i havent tested out to see what difference 8 would make.

 

No idea with the bios, if you do try it though I'd be interested in your results to see if it made any difference.

[evil4blue]

I read somewhere that the newest beta bios lets you use slots 1&3 for dual channel and have seen where people have gained about 10 MHz of stability.

 

Again I don't really know, but I believe the RAM in question was BH-5, so 10 MHz is a pretty big deal.

[evil4blue]

Okay, here's an update, I got home from work and found that memtest had 32 errors in over 12 hours of looping test 5, but none of the errors were generated in the last 300 passes. This was @ 3.4v 250fsb-1.5-2-2-5. I added an 80mm fan directly on the ram and upped the voltage on the vdimm to 3.6 just to see if I could get 260 fsb-2-2-2-5 to run psuedo error free. It was a no go.

 

I decided to try the 414-2 bios and the error count seemed to go down, but it was still in the 80-100 per pass. I decided to try the sticks in slots 1&3 since the new bios allowed dual channel in those slots and memtest is currently running error free @ about 50 passes s far. This is great since I've never gotten 260 to run at all.

 

I don't like the idea of 3.6v on the vdimm, but I'm gonna let it run over night and see if maybe by tomorrow I can drop the voltage a little. I haven't tried any faster fsb, but if this works I may up a little further.

 

If I can't get 260 to run under 3.6v, I'm probably gonna turn it back down to 3.4v and run as fast as that'll let me.

 

Keeping my fingers crossed.

[Knob]

evil, run Prime 95 longer. u will be surprised to find that your system is not as stable as u think it is

[evil4blue]

I hear you about prime95, and now I have run prime for 17+ hours error free! Last night I set up memtest to run overnight for another burn-in session and was able to drop my vdimm to 3.5v from 3.6v and run error free for 10 hours. This is great, I'm hoping that after a few more burns I'l be able to get down to 3.4v .

 

Here's my one problem, I still can't run SuperPI's 32M calculation. All other calcs and benchies run absolutely perfect, but it bothers me that SuperPI craps out. Any suggestions?

 

Here's a couple screenies of the new timings:

 

 

[Exar_333]

thats BH-5....it hits a dfinite ceiling, and you really cant do anything about it except try to give it a little more voltage....that might be as high as it will go...