Asus P5WDH - RAM DDR1000 stable, CPU 3.52 GHz stable, both cannot be stable together

[Thasp]

I recently got a set of g.skill DDR1000/4-4-4-12 binned D9.

It works with my e6400 with stability, before even touching memset at 333 FSB.

It works barely at 400 FSB, same DDR1000, before loosening out memset, and even then it's not 100% stable.

At 445 FSB, which is what I use for my stable CPU clock(I tested CPU clock with memory at under DDR680, for over six hours of prime), the RAM won't even run windows for more than 20 seconds at DDR1000. This is at 5-5-5-15, and loosening out TRFC, TRRD, and a bunch of other memset things.

I tried MCH 1.75 and have a fan on the chipset, I tried following the instructions in the p5wdh gskill thread. This really does seem to be a motherboard issue.

What I'm interested in knowing is whether this a limitation of this motherboard, this chipset, or my motherboard in particular(defective maybe, it was powered by a failing PSU for a few days before I realized it)?

I don't want to use my CPU at 2.6 GHz, and I don't want to use my RAM at DDR677.

If I had to choose before buying a new board, I'd choose the CPU at 3.52, because the CPU clock is important for what I use the computer for, whereas the RAM clock only helps applications I use a different computer for.

Thanks in advance for any responses. :)

[Revv23]

so have you tried 4:5 at 400fsb? what ram volts?

what about 1:1 445 fsb? again what ram volts/ timings?

[Thasp]

2.3 ram volts. the RAM says 2.2-2.3, I haven't seen reason to go with more. If the RAM is stable at 2.3v for DDR1000 with 333 FSB, but not at 400-445, it is my understanding that's not the RAM's fault.

I haven't tried 1:1 445 FSB, but I guess I can. Should I really expect it to even POST at those speeds? If it won't work at lower.. how does going higher fix it? I'm not saying it won't, but I'm curious about the logic behind it, because I am new to Intel overclocking. I've been working with an opteron 165 and a venice 3000+ for the past 18 months.

btw, I tried 375 MHz with the divider that gets me to DDR1000 - I forget the number. But it wouldn't POST, but it would POST and go into windows with 400 FSB and the divider to get me to DDR1000.. this breaks all rules of logic for me, that faster speed was more stable.

[Thasp]

Surprisingly, at 1:1, it didn't like being run at DDR 1600. And that was only 3.2 GHz CPU.

[GrimReaper]

try going a bit over 400fsb like 402fsb and ram at 1004 or so

[Thasp]

It turns out it's not really stable at 400 FSB/DDR1000, with loose memory timings either.

[^don.k's^]

Its the chipset, it doesnt ussually run high FSB + higher dividers on RAM. To get high fsb stability you need lower divider.

what fsb was the stability test @ddr1000 ? 250 ?? 333 ? ;) That's why it's not stable. the chipset wont run high ddr+high fsb ( unless 1:1 )

[justin_c]

How about other volts? vfsb, etc. crank those all the way up.
or your droop is real bad?
try other async dividers.
has to be bios bugs, if not prolly bad ram.

[syne_24]

Its the chipset, it doesnt ussually run high FSB + higher dividers on RAM. To get high fsb stability you need lower divider.

what fsb was the stability test @ddr1000 ? 250 ?? 333 ? ;) That's why it's not stable. the chipset wont run high ddr+high fsb ( unless 1:1 )

yup, i'm experiencing the same thing when I try to run high divider with high fsb. The chipset seems power hungry, atleast on my board. It needs atleast 1.75v just to boot @ ddr1000. My cpu is stable and memtest shows my ram is fine. It's definitely the board that is the problem. Even with max vmch, 1.85v, I still have trouble getting Orthos stable @ ddr1000.

[^don.k's^]

It's the problem of 975x, it has better performance, but OC's lower... So it needs higher voltage. I think it also has diferent die shrink than 965p.

[Thasp]

I just tried one RAM stick, I was memtesting with it. I stopped at 40%, that was enough to tell me it was different than the not-even-stable-to-get-the-windows-splash-screen of 5-5-5-15 DDR900 at 450 FSB, with two RAM sticks.

I looked for a 1T/2T option in the BIOS... there is none? :( Is there a good way to get it to work with two RAM sticks as well as it did for one? DFI made this easy, even with their budget model, all the RAM settings are in one menu and described clearly, here they're nonexistant.

I've heard of more than two RAM sticks killing your overclock, but never of more than one having much effect. I tried the two black slots together, then black + orange.

[Thasp]

Interesting, I was incorrect.

At FSB 450, where the CPU is stable if I take RAM out of the equation(make it DDFR677, 5-5-5-15), it works.

With two sticks, I tried memtesting at 450 FSB, 3-4-4-12, DDR677. Definition of frustrating - it failed after six hours, eleven minutes, and 15 passes.. it froze.

Is there anyway you can overclock memory in this board with two sticks/ Is there a 1T/2T option in the BIOS that I am missing? Is there a way to turn off the chipset's internal suck machine?

[Thasp]

Ok, last question.

Does the 965 chipset have the same issue, where the lower dividers are useless? Only the first two dividers after 1:1 work for overclocking, and obviously 1:1.

I need dividers to run the CPU at its full speed, unless I can find RAM that can do anything at 800 MHz.. which is impossible.

edit: I found out, the "divider" that is ddr800 at 266 fsb is what I was using to get ddr1000 at 333 fsb. The weird thing is that there are two options over it.. so I can do upto ddr1000 at 266 FSB?

Confusing board. Still useless for RAM overclocking at the moment, so I am looking to sell this and get another one since the problem is impossible to fix.

[Revv23]

965 has no dividers lower then 1:1 unfortunatly.

However 965 is great at clocking ram, so maybe 445 1:1 will be easy for you.

[Thasp]

that would mean 610-620 RAM... I doubt it can do that :(

I think I'll wait for RD600 and sell this later.