SuperPI failing on an non-OC'ed OEM AMD 64 3200+

[bigpgo]

When I try to run SuperPI for a 4M calculation, I keep getting errors. The thing is, I didn't even overclock my 90 nm athlon64 3200+ yet... Is it bad ? (I got OEM)

 

[bigpgo]

I tried 1M, 2M, and 8M. Those run fine, but 4M keeps getting "not exact in round" error.

 

[imported_whatever]

no idea why it is getting that
try running prime95 torture test. it is a better stability test than SuperPi
have you checked RAM latencies etc? could be that they are set too tight

 

[Gnoad]

I forget if superPI stresses ram at all. Try memtest to single out the ram. Then try prime95. If your cpu fails prime, I'd RMA.

 

[Duvie]

I would say test memtest....Superpi is stil sensitive to flaky memory or memory settings I believe. Prime95 wil fail if you fail superpi...

With my P4 at unstable levels I could pass 32mb superpi with FH going at same time and have prime95 fail out under 10minutes....Superpi IMO will usually mask an unstable system far more then prime95....

 

[bigpgo]

Yup, just ran Prime95 stress test - failed after 50 minutes ("Fatal error: rounding was 0.5, expected less than 0.4").

I'll try memtest to rule out the RAM.

 

[Gnoad]

Looks like your going to RMA something. Although, what is your RAM? You could just have it set up wrong.

 

[bigpgo]

Ram - crucial ballistix pc3200, 2x512, running in dual channel. Everything is at the default (stock) speeds and voltages, I haven't OC'ed anything yet.

 

[Gnoad]

Either run memtest or set ram to 3-3-3 and see what happens. I've seen many more bad sticks of ram than bad cpus, which is why I'm thinking ram. Also try testing each stick seperately and see what happens. If you pass memtest and get errors using each stick of ram, RMA that pos cpu. And for further testing, you could lower the cpu multi a few notches and see if it really is the cpu.

 

[bigpgo]

Okay, I ran 2 passes of memtest - no errors. Switched from slots 3-4 to slots 1-2 (still dual channel), re-ran memtest - 4 passes (about 1.5 hrs), - no errors... Very strange. The one thing I noticed when moving ram sticks around was - there was a tiny carpet hair stuck to the end of one of the sticks (right by the gold pins). Not sure when it got there. I guess I'll start prime95 and leave it running overnight.

 

[bigpgo]

SuperPI (32M run) just gave an SQR error... I guess I'll run memtest for like 10 hrs to be certain that it's not the memory, but after that, I'll prolly have to exchange the CPU

 

[LTC8K6]

Can you underclock the CPU? If you underclock it and it passes, then you have another strong clue that it's the CPU.

 

[StriderGT]

DISABLE SPREAD SPECTRUM in BIOS

 

[bigpgo]

Thanks guys, I'll try both of those (for under-clocking, should I only lower the FSB?)
In the mean time: last night I put the 2 mem sticks back in slot 1-2, and let SuperPI 32M and Torture Test run. It's now been 10 hours since than, Pi is finished, and Prime95 is still running no errors. Now that I think about it, as I kept moving sticks around, I've seen failures when they were in slots 3-4. Go figure. Is it possible that there's something wrong with those slots on the mobo?

 

[bigpgo]

Well, 22 hrs of Prime95, no errors. I guess it's a lesson learned - instability can be due to bad (?) dimm slots 3 and 4.

 

[Duvie]

I was just about to tell you it is the memory....I saw that exact error today when I tried to drop my fsb to run a 2600mhz but use a 166ram divider as opposed to 133...the result would have been 480mhz ddr but instead I couldn't pass 1mb of prime95 without getting that error...I tried the lowest of all timings with cas 2 and no go....I tried cas 3 and she doesn't even post....

Thanks for the tip I forgot tp maybe try slots 3-4 and see if I can get 480 to work in them....

 

[toNka64]

I've heard bad things about dimm slots 3 & 4.
I haven't personally tested it but everyone says use 1 & 2.
I get rounding errors too if i oc pass 2.1ghz.
I get no errors at stock spead but...
It will crash my entire system eventually.
I'd leave it running while i'm at work.
Come home and its off.
If you have rounding erros it will show in the first test.
Change your AGP freq. from 66 to 67.
I did that and Prime95 never crashes my computer anymore.
Still get rounding erros when i OC though.
I think its a power issue for me.
It might be for you too.
People have had problems with Antecs and MSI K8N Neo2s.
I'm RMAing my mobo/cpu and getting a new psu (OCZ).
That way it eliminates all potential issues... if you have the $$$

 

[Duvie]

slots 3 & 4 were crap....

They flat out restarted at windows splash screen.....in slots 1&2 at same settings you virtually could do anything on the desktop short of superpi and prime95...

Back to 1&2 theygo....