Winchester 90nm and Prime95

[dnottis]

I noticed that failure is more random on the Winchester as well. I can run Torture Test Priority 10 all day long without errors. The next day, I can run it again and error out within minutes.

I was just coming to this thread to complain about the same thing. Last night I let Prime95 run overnight - was about 8 hours of stability. Then today I restarted it and it died within 11 mins. Ran it again and now I cannot break 2 hours. Mind you I made no changes at all to the system. Primes for hours, then can't prime for mins? I'm so frustrated with trying to overclock AMD64, I can't get a good overclocker and when I manage a decent overclock I dunno whether it's stable or not

[Mana]

i have given up on prime, just gonna stick with a 32M superpi until a solution is found.

[trans am]

What about Toast and Hot cpu tester pro? there are many good cpu stabilty testing programs, why does everyone insist on Prime95. Try something else and see what happens. It could be a prime95/winchester software compatiblity bug with prime. Try other software and if the errors remain, we will know for sure it is a hardware problem on AMD's part.

[IvanAndreevich]

What about other A64 CPUs?

[Torin]

I was just coming to this thread to complain about the same thing. Last night I let Prime95 run overnight - was about 8 hours of stability. Then today I restarted it and it died within 11 mins. Ran it again and now I cannot break 2 hours. Mind you I made no changes at all to the system. Primes for hours, then can't prime for mins? I'm so frustrated with trying to overclock AMD64, I can't get a good overclocker and when I manage a decent overclock I dunno whether it's stable or not

Sounds like ambient temps changing your OC stability. At night, it is the coldest ambient temp, correct? In the afternoon, the ambient temp goes up and so does CPU temp, correct?

[trans am]

What about other A64 CPUs?

no, just the 90nm. I'm thinking it is a heat issue more than anything, I think running prime for any extended period of time is just making the cpu too hot because it's 90nm and it errors out due to heat. Does anyone have phase on a winnie that can prime for long periods of time with no errors?

[OmegaSys]

Sounds like ambient temps changing your OC stability. At night, it is the coldest ambient temp, correct? In the afternoon, the ambient temp goes up and so does CPU temp, correct?

I don't think that's it (at least in my case) because it will fail most often in the middle of the night.

-OmegaSys

[OmegaSys]

I had read a reference somewhere to a problem with the fan on the northbridge of the Neo2. I can't remember where but somoene mentioned that the fan sometimes stops running. I was wondering if that could be the cause of these seemingly random failures. Has anyone else heard anything about it? I personally haven't witnessed it but I suppose it's possible that it could happen without me noticing.

-OmegaSys

[IvanAndreevich]

trans am
Uh how is prime different from any other CPU-intensive task heating up the CPU? Winchesters run cooler than NC/CH, if anything.

[trans am]

trans am
Uh how is prime different from any other CPU-intensive task heating up the CPU? Winchesters run cooler than NC/CH, if anything.

I don't know about them actually running cooler. It seems at stock speeds they are definitely cooler, but overclocked it seems hotter than 130nm. I may be thinking crazy. I'm just trying to put something else on the table as to why they fail prime.

[Joe Camel]

ive linked to this thread in another thread over on OC...many over there are reporting P95 problems with 64 chips...
"Winnie" cant P95 @ stock clocks

as for the temps, i had this chip/mobo under a MachII and although i could get 3+GHz to boot/bench, even 2600's wernt P95 stable...i dont think its temps...

i just ran Super Pi 32M and a FULL (all tests) 05 @ 2750MHz with water cooling...NP. but P95 fails in <5min

[WiCKeD]

Well when i first got my a64 system ( 3500 gig of geil msi k8n neo2 ) i ran prime for 12 hours on blend.. then 34 hours on max heat/power.

both with no errors and i consider that "prime" stable pc never crashes or errors ever.

when my new ram arrives gunna overclock and do the same.

More people should follow your example.

There's no software bug that afflicts Winchesters only. That's ridiculous. Prime runs the same code the same way on every system. How is it you can always run stable if you lower your overclock? If it fails, other things may be having problems too.

Every consider that it might be a hardware problem with the Winchesters? They come standard at lower volts and are prone to severe power leakage at higher ones. So it's possible they are more sensitive to voltage fluctuations and are inherently less stable. What seems stable with the clocks we are used to getting with 754 - maybe they're not.

... and trans, blend stresses your memory, it doesn't stress the CPU at all. Look at your CPU usage. I'm not sure 3DMark01SE is going to get your CPU usage maxed on a new system either.

[trans am]

More people should follow your example.

There's no software bug that afflicts Winchesters only. That's ridiculous. Prime runs the same code the same way on every system. How is it you can always run stable if you lower your overclock? If it fails, other things may be having problems too.

Every consider that it might be a hardware problem with the Winchesters? It's possible they are more sensitive to voltage fluctuations and are inherently less stable. What seems stable with the clocks we are used to getting with 754 - maybe they're not.

... and trans, blend stresses your memory, it doesn't stress the CPU at all. Look at your CPU usage.

I thought blend stressed everything the hardest? Isn't that why it's called Blend? which do you recommend?
Ok, so it's not the temps, And it's not a software issue, then that only leaves one thing.... 'DO' Winnies are FVCKED!

[WiCKeD]

Ya that would make sense. But it's 99% for your RAM. The first one tests "maximum FPU stress," which would be your CPU. Your usage should immediately go to 100% using it. The second one will jack up your temps and test how well your CPU will perform at near-max temps.

[trans am]

Ya that would make sense. But it's 99% for your RAM. The first one tests "maximum FPU stress," which would be your CPU. Your usage should immediately go to 100% using it. The second one will jack up your temps and test how well your CPU will perform at near-max temps.

Ok, man thanks. I'll give it a whirl.

[paulopais]

I think this is a voltage problem in Prime95.. because when i have a big variation of vcore or +12v the prime95 fail.. if i up de voltage of my cpu at 2900 to 1,6 prime95 fail because de variation.. but if i put 2900mhz ate 1,5 its stay stable.

[WiCKeD]

Well that sounds like a voltage problem, but it's not due to some binary program - it would be a problem with the processor. Prime is good at catching the smallest instabilities and that's why we don't like it!

[saaya]

has any of you run memtest? does memtest run alright?

[conrad.maranan]

I think a good number of us will always run Memtest first prior to even booting Windows with new memory settings.

[saaya]

so memtest works but prime fails? then it sounds like a program related error, i never heard that memory passes metest but fails in prime before... weird...

[conrad.maranan]

Scroll up about 8 posts and read WiCKeD 's post.

[saaya]

Scroll up about 8 posts and read WiCKeD 's post.

why? i read that post before, what are you trying to tell me?

it makes sence to me that only one program has a problem with a cpu, its probably a hardware problem but it should be patchable, and even if its not, e0 will be out soon and will most likely (prays ) fix this problem

[saaya]

hmmm the problemcould also be fixable with a bios update wich lowers a chipset or memory timing slightly

[conrad.maranan]

why? i read that post before, what are you trying to tell me?

it makes sence to me that only one program has a problem with a cpu, its probably a hardware problem but it should be patchable, and even if its not, e0 will be out soon and will most likely (prays ) fix this problem

I'm saying it's a hardware problem. You also say that it's a hardware problem in the above post, but in your post prior to that, you originally state that it's a software problem.

so memtest works but prime fails? then it sounds like a program related error, i never heard that memory passes metest but fails in prime before... weird...

Catch my drift?

[jlccarv]

I noticed that failure is more random on the Winchester as well. I can run Torture Test Priority 10 all day long without errors. The next day, I can run it again and error out within minutes.

I found exactly the same thing.

I once primed for ~10hrs overnight, then got on the net while prime was still running and 5 mins later FATAL error.

Also, tried again last night (same clocks), failed after 45mins...