Need Overclocking Help... Please!

[GML3G0]

I have my 3200+ clocked at 2.4GHz and my RAM at 2.5-2-2-7 @2.8 Volts. It runs fine for a few minutes (the highest I go is 1 hours)... then I get a fatal error saying it rounded ).5 and expected something less than 0.4. If I bring it down to 235MHz HTT, then it goes for about an hour and 40 minutes before I get this error again. My vcore is 1.5, but my board makes it flunctuate anywhere between that and 1.55. It's running at max load 59*C.

Now my question. Is that error a RAM error that I have to loosen up timings... or could it be the CPU with not enough juice or cooling?

 

[DerKaiser]

Now my question. Is that error a RAM error that I have to loosen up timings... or could it be the CPU with not enough juice or cooling?

Lower your multiplier to eliminate the CPU as the cause.

 

[CaBoOse999]

Run memtest first and if it passes that then you can narrow it down to being cpu related. Prime95 doesn't test the memory as much as it does the cpu so most likely its the cpu.

 

[MTSteel]

That is the same error I got on my 3000 when I oc it to 2.25Gig within the first 2 minutes of the test. I then upped the CPU voltage from 1.425 to 1.475 and it ran for 30 minutes before I simply turned the test off. I then played 2+ hours of CS:S without a problem.
I will try running Prime95 for a couple hours tonight and see what happens. I am totally new to this A64 OC thing, and look forward to seeing some answers. That is how I got it to Prime for more than 3 minutes anyways, so it liked some more juice

 

[MrControversial]

Loosen up those timings a bit. Try a lower CAS latency and then up the rest to 3, 3, 6 respectively.

 

[GML3G0]

Originally posted by: MrControversial
Loosen up those timings a bit. Try a lower CAS latency and then up the rest to 3, 3, 6 respectively.

CAS is the first one correct? the 2.5? if i'm right, then no, this RAM will not do 215MHz FSB with the 2 CAS. The system won't boot.

I don't think it's the juice. It already has between 1.5 and 1.55 volts. I'll try loosening up the timings. Might be that. But I passed MemTest no problem. I'll try your suggestions though when I have some time. Still waiting for the XP-90 to come in. Thanks.

 

[cbehnken]

9/10 out of ten it's simply your cpu can't clock that high

 

[GML3G0]

ok, i tried lossening the timings to 2.5-3-3-8. well it lasted a little longer, but ultimately, I still got that same rounding error again. You think I should try to bump up the vcore a little bit higher, to say... 1.525? I might try that, see how it goes.

 

[DerKaiser]

LOWER THE MULTIPLIER TO SEE IF IT'S YOUR CPU OR NOT.

 

[GML3G0]

ALRIGHT! lol
that is what im going to do now. bumping up the voltage to 1.5250 seemed to prolong the test.
1). running memtest86 off floppy with memory at 240FSB 2.8V 2.5-2-2-7. according to zebo, it can do this speed up to 245MHz (yes I know everyone's is different)

RESULTS: It's not the memory. Ran MemTest86 off floppy. Ran test with 2 passes - 0 erros @ 2.5-2-2-7 240MHz FSB 1:1. Next...

2). lowering divider to 8 or 9 to see if its cpu

RESULTS: Motherboard not letting me lower multiplier.

3). ...

 

[DerKaiser]

BTW, is that a Winchester 3200?

Also, can you select 1T vs 2T on that board? On my A8N-SLI, I can't get over 220 mhz memory at 1T but I can get over 250 at 2T.

 

[GML3G0]

Originally posted by: DerKaiser
BTW, is that a Winchester 3200?

Also, can you select 1T vs 2T on that board? On my A8N-SLI, I can't get over 220 mhz memory at 1T but I can get over 250 at 2T.

yes it is. see sig.

P.S. view my previous post for updates.

 

[Amaroque]

Most people wouldn't consider your overclock stable unless it can run Prime95 for at least 24 hours (48 recommended). Your RAM timings are too tight.

Crucial Ballistix 3200 @ 240FSB 2.8V 2.5-2-2-7

FYI Zebo was using Ballistix PC 4000.

2.5-3-3-8 1T will work much better for that RAM speed. Also, that is a Via chipset. Via doesn't OC nearly as well as an nVidia chipset. At 240 you are also running into the upper limitation for the Via chipset. It will be hard to troubleshoot your problem if it's a chipset issue. Memtest most likely won't pick up a chipset problem either. Prime95 will error out if your chipset isn't stable.

 

[GML3G0]

well then, I guess it's my chipset. but games and everyday tasks run very smooth and stable. Anyways... I'll continue testing to isolate the problem. I wanted a Neo2, but everyone had it out of stock. And it's been that way for a while...

 

[GML3G0]

houston, we have a problem. my motherboard is not allowing me to lower the multiplier. won't even let me touch it.

Is a 32M Super Pi calculation enough to test stability? I doubt it, but just asking.

EDIT:Crap, even that craps out in rounding after 20 minutes. What if I increase my AGP lock to 74MHz, up from 66MHz? There has to be something that would get it stable. It shouldn't need for than 5.5 vcore right?

 

[Regulator07]

i had a similar problem with my a8n and 3500, it would fail prime blend after either a minute or an hour. i thought it was hardware, so i upped voltaged to cpu and ram, but still it would crap out. found out now that it was a bios issue, i updated the bios and now it prime blends for over 12 hours. you should run memtest86+ first for about 10 passes, to make sure the memory is not causing the problem. then run the small and large fft tests in prime to see if only the cpu is causing the problem. then run blend to test ram and mem and (i think) more specifically the memory controller on the cpu. if that fails then up the voltage to the cpu, the mem controller may need some more juice to work stable. i have mine raised up a notch anyway just to make sure i dont get errors, my temps are never over 40C so it doesnt hurt the cpu. if that does not do it, then you should set everything back to stock and see if you can pass these tests, especially blend, if it does not pass you may want to try different bioses, and if all else fails, rma the cpu.

 

[MarkHark]

Why should he RMA the CPU only because it does not overclock all the way up to 2.4 GHz stably? I'm pretty confident he'll experiment no problems once he runs it closer to spec speeds

 

[Regulator07]

Originally posted by: MarkHark
Why should he RMA the CPU only because it does not overclock all the way up to 2.4 GHz stably? I'm pretty confident he'll experiment no problems once he runs it closer to spec speeds

no i agree, if he cant overclock but he runs fine at stock then he is just out of luck, but if it cant run stable at stock THEN maybe he should rma it.

 

[MTSteel]

I tried running Prime 95 Last night at 250x9 1.475V, memory using the 166 Divider and failed after 2.5 hours. I have upped the V to 1.50 in BIOS(1.47 in Smart Guardian reads), temps max at 40C, I left it running at home, will check when after work.
I am leaning to a poor CPU.

Here is what I did this morning before work.

1.475V RAM divider 100 HTT 3

300x5 Prime started
300x6 Prime started
300x7 failed load windows
280x7 Prime started
280x8 Prime started
280x9 locked up at black screen

1.525V RAM divider 100 HTT 3

260x9 Prime failed after 1 minute

More things I want to try, but decided to let it run at 1.50v RAM 100 HTT 4 at 250x9 to see if I could get 8+ hours prime while at work.


Last night I was failing memtest 6 quite a bit, sometimes on the 1st path, other times on say the 6th, this was all at Load optimized defaults, and trying more Vdimm, no matter what Vdimm I used it would fail in test 6 only. I forced the RAM to 2.5-3-3-7 1T rather than running it in Auto and I was able to run 13 passes all tests last night with 0 errors all at 2.6V. I hope forcing the RAM with those settings has cleared the memtest problem.

Silly CPU does not like to go over 2.2 without Prime failing in first minute. Highest I have tried was 266x9, booted to XP, but failed prime immediately.
If it fails, I will turn it back to the 200x9, V back to auto and just see if I can prime overnight. Now I am curious if this CPUT can make a night of prime at default speed!

 

[Jeff7181]

Run the Prime95 test that says it stresses the CPU and uses little RAM... if it fails that, it's not the memory.

 

[Regulator07]

Originally posted by: MTSteel
I tried running Prime 95 Last night at 250x9 1.475V, memory using the 166 Divider and failed after 2.5 hours. I have upped the V to 1.50 in BIOS(1.47 in Smart Guardian reads), temps max at 40C, I left it running at home, will check when after work.
I am leaning to a poor CPU.

Here is what I did this morning before work.

1.475V RAM divider 100 HTT 3

300x5 Prime started
300x6 Prime started
300x7 failed load windows
280x7 Prime started
280x8 Prime started
280x9 locked up at black screen

1.525V RAM divider 100 HTT 3

260x9 Prime failed after 1 minute

More things I want to try, but decided to let it run at 1.50v RAM 100 HTT 4 at 250x9 to see if I could get 8+ hours prime while at work.


Last night I was failing memtest 6 quite a bit, sometimes on the 1st path, other times on say the 6th, this was all at Load optimized defaults, and trying more Vdimm, no matter what Vdimm I used it would fail in test 6 only. I forced the RAM to 2.5-3-3-7 1T rather than running it in Auto and I was able to run 13 passes all tests last night with 0 errors all at 2.6V. I hope forcing the RAM with those settings has cleared the memtest problem.

Silly CPU does not like to go over 2.2 without Prime failing in first minute. Highest I have tried was 266x9, booted to XP, but failed prime immediately.
If it fails, I will turn it back to the 200x9, V back to auto and just see if I can prime overnight. Now I am curious if this CPUT can make a night of prime at default speed!

yeah, you should check to make sure that it is stock stable, after that you can begin overclocking. at all stock settings, you should be able to run memtest86+ for a night and prime all three tests for over 12 hours, after that it is up to you if you want to keep it going,obviously it may not fail until 18 hours or something but in my opinion, 12 hours is at least enough to make sure it is running good (i know people will only call it stable after 24 or 48 hours but i dont fold or do any 100% cpu usage programs for days, but i do understand where people are coming from who do do that). if it can pass that, then at least you know the chip is not necessarily deffective, but if you cant oc then you may have just gotten a bad ocing chip, nothing you can really do.

 

[GML3G0]

ok, im 95% sure it's not the memory. it's probably my CPU. now is increasing the AGP/PCI lock from 66MHz to 74MHz worth it? Might that possibly help? I also took someone's advice and updated BIOS from version 17 to 19. running tests again to see if that might have been the problem.

 

[Jeff7181]

Increasing the PCI/AGP speeds isn't recommended. That's why they put locks on them, so they stay put. Raising them too high can cause instability.

 

[GML3G0]

ok, so, why won't my bios let me lower the multiplier?

 

[DerKaiser]

According to this thread at Abit's forums, you have to use pagedown to change the multiplier.