Barton failing on Prime95 even at stock speeds (???) (a.k.a. yet another Barton 2500+ overclock history)

[rayon]

Hi folks,

Yes, this is yet another Barton 2500+ history.

I recently received a Asus A7NX8-E Deluxe, Althon XP 2500+ Barton retail and 2x 512MB of OCZ PC3200 DDR RAM from Newegg.

Put everything in place of my old VIA based EPoX 8-KHA+ and Althon XP 1800+ and everything works like a charm. I didn't even have to reinstall Windows XP. It detected and installed drivers for everything.

Anyways, I'm happy with my new setup. Then I decided to go 400MHz FSB and try and overclock this puppy. How happy I was when I saw it working just fine at 3200+ ratings with stock voltage. After a couple of days of running it at 400MHz FSB and not having any problem I decided to test its stability by running Prime95.

Installed Prime95 23.7 and got a failure on the very first run about 1min after started. Bummer! Tried upping voltage in .025v steps but I get to 1.75v and I still get the problem (I got a BSOD when running it at 1.7v, go figure).

Anyways, reading this forum I realized that version 23.7 doesn't seem like a good one. Some one reported it might have a bug. Well, I tried running version 23.8 and got the same result. Then I tried 22.13 and got the same result.

I then went back to stock voltage and FSB speeds (2500+ rating at 333MHz FSB and 1.65v) and started from there. First I thought I'd check the stability at stock. I ran Prime95 version 22.13 and imagine my suprise when it failed after just a couple of minutes running it! :| :Q

The error message is always the same:

[...]
FATAL ERROR: Rounding was 0.4999990463, expected less than 0.4
Hardware failure detected, consult stress.txt
Torture Test ran 1 minutes - 1 errors, 0 warnings
[...]

After reading a few more messages in this forum I decided to test my memory. I ran memtest86 for 11 hours yesterday and all 11 tests passed.

What do you guys think? Is my CPU bad? Apparently everything else works -- I'm using it to post this message right now.

Thanks for your time,
Bruno.

 

[Killrose]

Typing on AT Forums is not much of a CPU stress test

Maybe your PSU is not upto snuff under load? though I would think something like that would lock-up your system.

 

[Shimmishim]

i'd say it was your ram but looks like you memtested your memory..

then again, i've had prime fail before even with memtest passing all 11 tests or whatever

you enabled cache on for all tests and ran all tests correct?

if all else fails, bump the voltage up for your vdimm and if that fails, try more cpu voltage...

i'd be willing to go up to 1.85 volts with a barton.. but that's just me

 

[rayon]

Thanks for the replies, guys.

Yeah, typing definitely is not much of a stress test, Killrose! But I've also been running a few games and benchmarks. So far so good. Re: the PSU, I don't know. I didn't have any problems with my old setup that would indicate PSU problems. I might try replacing it later, in case everything else fails.

Shimmishim, when you say you had prime95 fail even with all those memtest tests, do you mean that even with all those tests, you still had problems with your memory?

I ran all tests with caching set so that memtest decides when to enabled or disable it.

Do you think it's plausible to have memtest succeed and prime95 fail because of memory problems? I'll give your suggestions a try and will report back, just in case.

Thanks again!

 

[Duvie]

It is pausible but it has generally been the reverse for me.....Memtest has been a much more solid tester of the ram and ram timings...However that may have changed since the version 23's of the program but I find the multiple test rather confusing and just one more reason why prime95 is becoming less and less the program of choie..They took a simple program and now making it too complexed and now we can't tell what the heck is the culprit...


I run version 22.12 and it generally test the ram very little... I follow it up memtest overnight....

I do find it tough to believe ram is the problem with 11 successful passes of memtest....

In my experience rounding errors have generally been cpu errors....


Things to look at....

1) Power supply...maybe the 5v and 3.3v lines are dipping too much under the intense loading of the chip under prime....memtest loads but mainly isolate the memory and not so much stressing on the cpu...

2) Heat...Any heat issues??? Cpu or Northbridge chipset...

 

[rayon]

I tend to agree with you, Duvie.

Is Motherboard Monitor good enough to monitor the 5v and 3.3v lines?

According to NVidia's system utility, CPU is at 41C at normal load (not idle, not heavy). Don't know about the chipset. System temperature is reported at 34C though.

I think these are good and within normal range.

 

[Duvie]

I like speedfan...easier to install and it finds the sensors for you...Also compare it with the asusprobe that should have been bundled with your mobo.....

Temps as you reported them seem OK though I think the system temp which may or may not actually be the northbridge chipset temp may be a bit high....If not load thta thing may hit high 30's or low 40's....try taking off sid of case and blowing a large fan into the acse to try to rule out temp issues....

Definitely check the power issue...report idle and load (full load please) voltages vcore, +3.3v,+5v, and +12v...Also vdimm and vagp if it list them in the asusprobe...

 

[Kharlia]

I have a AMD 2500+ (1833mhz) oced to 2400mhz right now, ran Prime95; had similar error message:

FATAL ERROR: Rounding was 5, expected less than 0.4
Hardware failure detected, consult stress.txt
Torture Test ran 4 minutes - 1 errors, 0 warnings

Raised the core voltage to try it out, ran MemTest86 for stability, my memory passed all eleven tests as well. -.- No clue what it is, using an Antec 420W PSU, so I don't think its that.. relatively new as well but I"ll try it out on another PSU. -.-

 

[rayon]

I just realized that while the suggestions from everybody are valid, I think I didn't stressed it enough: I'm receiving errors in Prime95 running my Barton 2500+ at stock speed (1.8GHz) and voltage (1.65v).

I don't want to increase the voltage as there should be no need for it as I'm running it stock speed.

Have anyone seen anything like this before?

Thanks,

 

[Sid59]

Originally posted by: Kharlia
I have a AMD 2500+ (1833mhz) oced to 2400mhz right now, ran Prime95; had similar error message:

FATAL ERROR: Rounding was 5, expected less than 0.4
Hardware failure detected, consult stress.txt
Torture Test ran 4 minutes - 1 errors, 0 warnings

Raised the core voltage to try it out, ran MemTest86 for stability, my memory passed all eleven tests as well. -.- No clue what it is, using an Antec 420W PSU, so I don't think its that.. relatively new as well but I"ll try it out on another PSU. -.-

2400 is asking alot from that AMD chip. drop down the speed and contue for stability.

 

[MDE]

Originally posted by: rayon
I just realized that while the suggestions from everybody are valid, I think I didn't stressed it enough: I'm receiving errors in Prime95 running my Barton 2500+ at stock speed (1.8GHz) and voltage (1.65v).

I don't want to increase the voltage as there should be no need for it as I'm running it stock speed.

Have anyone seen anything like this before?

Thanks,

If it's failing P95 at stock speeds, and there is no other component that could be faulty, then RMA the chip.

 

[sugarkang]

reinstall windows from scratch before you do anything else.

WHY? BECAUSE YOU NEVER KNOW!

 

[Mark R]

I don't know whether it is something to do with the NForce2 chipset, but I've had the exact same problem with an entire batch of PCs I built using Abit NF7s.

Every single one in the batch would fail prime at stock speed, and no amount of tweaking would fix it. Running Windows memory diagnostic or memtest86 would fail on every single machine (but only on one test: Stride on WMD, and test5 on memtest). I tried 6 different sticks of RAM from 3 manufacturers and 2 different speed grades, 5 different CPUs of 3 different speed grades. Nothing made any difference.

I plan to dump the motherboards as soon as the A64 is more affordable.

 

[Grimner]

I saw that message only yesterday after installing 1 Gb of Corsair TwinX 3200LL.
One or two calculations on the torture test, and there it came.

Dug up some old memories and told the BIOS to let the SPD decide the memory speed.

My problem solved - any very likely yours, too

 

[rayon]

Thanks, MonkeyDriveExpress

I'll try and get another CPU to try on this MB. I'd like to make sure the chip is at fault before RMA'ing it.

Thanks, Grimner.

Unfortunately I'm already using the "By SPD" setting in BIOS.

 

[Sid59]

did you run the Small FFTs (maximum FPU stress, data fits in the L2 cache, RAM not tested much)

that isolates just the cpu and uses no ram.