EX38-DQ6 -- memory divider Q

[mneblett]

When in the MIT, I can set the memory ratio to 2.00D (i.e., 1:1). When running at 400 MHz FSB, that's DDR2-800. If I go to, say 425MHz FSB, the memory goes to 850 MHz -- no surprise there.

Is there any way with this board to decouple or otherwise alter the memory speed?

The reason I ask is that I'd like to be able to ramp up the FSB to 445 MHz (to get to 4.0 GHz on the E8400), but keep the memory down in the 800 MHz range. This is because one of my four memory modules (Corsair DDR2-800 4-4-4-12) appears to have a weak cell which fails after a few hours of Orthos when at higher speeds (~880-890), even when the timings are relaxed to 5-5-5-18. If I can't change the ratio between the memory and the FSB, I likely will not be able to exceed about 3.8 GHz on the E8400.

Any (constructive!) thoughts appreciated.

Thanks,
Mark

[Lsdmeasap]

Re: EX38-DQ6 -- memory divider Q

RMA the bad RAM!

But no, there is no way to lower the ram under 1:1. Using a higher multi, but I assume you already are using the highest plus the .5 fine clock correct?

[mneblett]

Re: EX38-DQ6 -- memory divider Q

RMA the bad RAM!

But no, there is no way to lower the ram under 1:1. Using a higher multi, but I assume you already are using the highest plus the .5 fine clock correct?

I probably should, but in good conscience I'd have a hard time doing so -- it runs fine at spec (4-4-4-12-2T, 800), and only gave me a memtest error when OC'd. Plus, I'm now not sure it's the memory that causes Orthos to punt.

In the course of isolating components, I reduced the E8400's multiplier, set the board to 425 MHz (DDR2-850) and ran Orthos overnight -- so far, 8 hrs with no errors.

I was not aware that the E8400 would use the 1/2 multiplier -- if it can, that would help by allowing the CPU to reach 4 GHz with a 422 MHz FSB (i.e., DDR2-844, instead of DDR2-890 at a 445 MHz FSB). I'll give this a try tonight -- thanks!

Mark

[Lsdmeasap]

Re: EX38-DQ6 -- memory divider Q

Yes it can use the .5, it should be in your MIT page called fine clock ratio. You however cannot use it above your CPU's max, incase you were unaware. I think you can do 8.5 but not 9.5, 8400 is 9 max multi right? If so, 6.5, 7.5, 8.5 is how you may use it

To be sure it is not the ram causing orthos to fail, take the memory out of the equation and set orthos on small FFt's and run for X hours, then large FFt's if no fails then but it fails on blend it is indeed your ram. Blend is the only test that stresses the ram

[mneblett]

Re: EX38-DQ6 -- memory divider Q

Yes it can use the .5, it should be in your MIT page called fine clock ratio. You however cannot use it above your CPU's max, incase you were unaware. I think you can do 8.5 but not 9.5, 8400 is 9 max multi right? If so, 6.5, 7.5, 8.5 is how you may use it

Yep -- 9's the max on an E8400.

To be sure it is not the ram causing orthos to fail, take the memory out of the equation and set orthos on small FFt's and run for X hours, then large FFt's if no fails then but it fails on blend it is indeed your ram. Blend is the only test that stresses the ram

Perfect -- a definitive test. I'll let you know how it comes out.

[mneblett]

Re: EX38-DQ6 -- memory divider Q

Update --

Well, I'm happy to say I'm stable at 4.0 GHz on the E8400

I tested the 4x2GB Corsair 4-4-4-12 at 5-5-5-15 900 MHz (450 MHz bus, cpu multiplier at 6x), and it went 20+hrs with no problem.

I then set the bus to 445 MHz (9x445=4005), upped the cpu voltage to 1.4V (shows 1.376 or 1.392V in CPU-Z), and it passed Orthos for a bit over 12hrs.

*EDIT* I don't know whether it made any difference, but before running the 4 GHz test, I disabled legacy USB support in the bios.

I'd like to run at lower voltage, so I'm now testing progressively lower voltages until I find the lowest stable level.

Thanks again for the help!

[Lsdmeasap]

Re: EX38-DQ6 -- memory divider Q

Yeap, that is the way to do it. Progressively lower and test

Do you use loadline enabled? If not, have you tried it, it helps with the initial drop from bios to windows, and the Vdroop from idle to load.

I have noticed that it does make a different amount of difference at every different voltage setting, sometimes better, sometimes not so much help