X2 4200 Overclock being a pain to stabilize

[thebove]

ok, i've been out of overclocking for quite a while, but i've been fussing with this chip all night and haven't really been able to get anywhere (see hardware in sig below - system 1).

what i do now thus far, is that according to memtest 86+, the mushkin ram i've got is good for 273Mhz, 3-4-3-8 1T @ 2.7v. this was tested in memtest using a Ht multiplier of 4, so the ht bus was running just shy of 1100Mhz.

here's the issues i'm encountering. i'd like to take advantage of some of this memory's potential, so i need to use the 9x multi to put me in a position to do that. BUT, no matter what i try, i cannot get ANYTHING stable using a 9x multiplier. even running essentially at stock x2 4200 speeds but with a pumped fsb (so 9x245, 4x HT versus the stock 11x200, 5x HT), prime 95 will fail on both cores in under a minute. even running slightly lower than stock causes issues. i haven't tried using other multipliers yet, or messing with dividers, because i'm too green with A64 overclocking to understand how the dividers affect the overclock. but so far the night has been quite frustrating.

anyone have some insight to offer? I'm a bit discouraged at this point.

and if anyone's curious, i'm running the 1011 beta bios for the mobo.

[James]

With the Athlon64's, the memory controller is on die, meaning as you increase the FSB, it has to work harder, even if the rest of the cpu is staying at "stock" speeds.

[thebove]

right i'm well aware of that, but wouldn't it crap out on memtest if it was the mem controller causing issues?

[ComputerManiacs]

Don't really trust Memtest anymore...

now I use SuperPI with the 32M calcalution to test my computer ofr stability and speed.



Computer Maniac

[Almost Famous]

TB,
I just finished building a new NForce4 rig with my XP4200+ & 2Gb ram.
I'm not having any problems running a 9X multi @250mhz.
I'm running the GSkill PC4000 Extreme Series 2GB (2 x 1GB) @ [email protected]
It's strange that you can't even run 200Mhz @ 9X multi.
What's the fastest stable OC that you've attained running another mult?

P.S.
As for the comment about SuperPi being the best test for stability..blah
It only gets my proc up to 50%@32mb...not much of s stress test if you ask me. In fact I'm running the 32Mb SuperPi as I'm typing this ..he he

[thebove]

as i write this, i'm running 10x240, 3-4-3-8 1T @ 2.7v. processor is getting 1.4625v. I'm not entirely stable here either, although i still need to try a little more voltage for the processor. but with these settings i'm prime stable for about 45minutes to an hour, and then both cores will conk out. i've had firefox crash once also.

tomorrow i guess i'll try and push it to the highest speed i can attain, using a high mult and lower fsb and see where it falls.

[James]

P.S.
As for the comment about SuperPi being the best test for stability..blah
It only gets my proc up to 50%@32mb...not much of s stress test if you ask me. In fact I'm running the 32Mb SuperPi as I'm typing this ..he he

That's why you run two processes. It's not a threaded program. Each CPU rates 50%, run two instances and you will get 100% load.

[OS-Wiz]

I seem to get my best OCs when I keep the HT/LDT at or below 1000mhz. Does your mobo allow you to bump up the HT/LDT voltage? If it does you might try bumping it a notch or two. Is the Mushkin RAM TCCD?

[Almost Famous]

That's why you run two processes. It's not a threaded program. Each CPU rates 50%, run two instances and you will get 100% load.

No....really?
It still does not test all your systems components. It's not a complete system stability test...that's my point

TB: See what your highest OC is with a higher mult.
Then at least you'll know what the proc is capable of.
After that, you may find out that your ram is limiting you.

[LAG]

I bet you're always going to have problems when that HTT is over 1000. You're going to have to ratchet that down a bit I think.

[OS-Wiz]

No....really?
It still does not test all your systems components. It's not a complete system stability test...that's my point

TB: See what your highest OC is with a higher mult.
Then at least you'll know what the proc is capable of.
After that, you may find out that your ram is limiting you.

Dah, why didn't I mention that I guess we all sometimes skip the basics. Reduce the memory divider a couple two or three notches and first find max CPU speed. Back off that speed 5mhz and then start adding in some memory divider. HT freq = HTT x HT multi, e.g. 250 x 4 = 1000. Memory frequency is determined by both CPU multiplier and RAM divider, unlike any other system. The Memory Frequency (f) = CPU Frequency (c) / [Multiplier (m) + Divider (d)] f = c / (m + d) where d is 0, 2, or 5 (1:1, 6:5, 3:2) m is an integer, always rounded up (therefore .5 multi's will round up) so if you run 2.2ghz 1:1 on a 10x multiplier your memory frequency is 220mhz 220 = 2200 / (10 + 0) but if you run 2.4ghz on a 9x multiplier (267x9) 6:5 your memory frequency is NOT 222 (267 x .83), but rather 218.
218.18 = 2400 / (9 + 2).

[thebove]

alright, i think i may have discovered what's causing the issues - too much voltage for the memory. the specs say it'll do 3-4-3-8 at 2.8v, and i had started off using about 2.65 and when things didn't pan out like i wanted i had upped to 2.7v, and initially thought they were stable there. but, not so. after about an hour and a half of memtest at 10x250 they crapped out, and only took a few minutes at that voltage under prime 95. so i went down to 2.65v, and it again, was memtest stable for about an hour and a half, but it also lasted about an hour and a half at prime.

so, i set it back down to 2.6v, went out to the bar and watched the pistons game, and 4 and a half hours later they're still memtest stable. so i'll run prime95 tonight and see if it makes it though, and i think i'll be set after that.

still doesn't explain the 9x multi thing, although i'll have to investigate further with low voltages on the memory.

but for the record, the mushkin xp4000 does NOT like extra voltage. it seems to do just fine with stock or near stock power.

[Almost Famous]

....
but for the record, the mushkin xp4000 does NOT like extra voltage. it seems to do just fine with stock or near stock power.

Neither does the GSKill..stock voltages only

[OS-Wiz]

Its all aout what kind of memory chips are used. If they're TCCD then 2.65 - 2.75 is best. If they're UTT (BH-5) then the sky's the limit (3.3 - 3.5), just don't burn out the memory controller.

[rabidmoose171]

Those arent bh-5 or UTT, not at those timings at least