AthlonXP "Barton" + nForce2 mobo stability problems?

[deveraux]

I just bought an AMD AthlonXP Barton 3000+ along with an Abit NF7-S2G board. The problem is that the system is ridiculously unstable with it crashing every 20m (sometimes even 5m from booting), max 30 mins. Anyone has any ideas what is going on wrong?

I ran memtest86 for 5hrs+ and it didn't give me any problems, however, Prime95 failed at stock within 15mins. Tried prime again and failed within 15s. Could it really be the CPU? I find it hard to believe that a brand new CPU is giving me problems.

Any ideas?

 

[YOyoYOhowsDAjello]

Sometimes brand new parts are the worst. You could have gotten a bad part.

Did you run memtest with a different CPU in? (since the computer was on for 5hrs)

If you tested the computer with a different CPU and it was fine, and then changed the CPU and it was crap... it's probably the CPU.

Have you tried lowering the clock of the barton to see if it's stable at lower speeds?

Is this cpu more power hungry than your old one? Maybe your powersupply can't handle your new system?

 

[VirtualLarry]

Did you put the heatsink on backwards, so it is not making good contact with the CPU? That was a common oversight when assembling Socket-A systems. Sounds like the CPU might be overheating. Also, does the chipset on the board have decent cooling? The NF2 needed it, especially when running at a 200Mhz FSB, otherwise data-corruption could occur. IOW, load MBM5 or something and check your temps, especially your CPU core temp if that board supports CPU diode temp monitoring.

 

[deveraux]

Originally posted by: YOyoYOhowsDAjello
Sometimes brand new parts are the worst. You could have gotten a bad part.

Did you run memtest with a different CPU in? (since the computer was on for 5hrs)

If you tested the computer with a different CPU and it was fine, and then changed the CPU and it was crap... it's probably the CPU.

Have you tried lowering the clock of the barton to see if it's stable at lower speeds?

Is this cpu more power hungry than your old one? Maybe your powersupply can't handle your new system?

memtest86 was run with the new CPU hence why I don't think its the CPU. My powersupply is an Antec SmartPower 350W (came with SLK3700-BQE) and the voltage rails seem to be quite stable. My system specs:

AMD AthlonXP 3000+ (Barton core, no OCed)
Abit NF7-S2G (nforce2 400)
2 x 512MB PC2700 Apacer RAM (2.5-3-3-6)
ATI 9600XT 128MB
Creative Audigy DE
40GB Seagate Barracuda IV
80GB Seagate Barracuda IV
LG DVD-ROM
LiteON DVD-RW (SOHW-812S)
1 x PCI LAN card

I think 350W should be able to handle this quite well. It was only upgraded to an XP 3000+ and new mobo, that's it. Everything else is brought over from the last system.

Originally posted by: VirtualLarry
Did you put the heatsink on backwards, so it is not making good contact with the CPU? That was a common oversight when assembling Socket-A systems. Sounds like the CPU might be overheating. Also, does the chipset on the board have decent cooling? The NF2 needed it, especially when running at a 200Mhz FSB, otherwise data-corruption could occur. IOW, load MBM5 or something and check your temps, especially your CPU core temp if that board supports CPU diode temp monitoring.

I've never run this CPU under load yet since it usually reboots before then. The closest I can get it to load was the time I ran memtest86 for 5hrs and then the CPU temp was at 51 C which I think is acceptable for this core. I am pretty sure I installed the heatsink right, but I will check on it and post back if I did put it the wrong way round, but I seriously doubt it.

Any other ideas?

 

[maluckey]

That PSU is nominal for the load, but it should work for a while.

I would remove the HS/fan and reseat it, then check to see if disablbling CPU interface solves the issue. Also, just because you pass memtest does NOT mean that your PSU is OK. It only means that your RAM can operate without a hitch at it's settings while in DOS mode. Even that is no 100 percent guarantee that your RAM is good. P95 tells you that your setup is VERY unstable, and needs attention. It has a hard time telling you where though.

Does it crash with running large(larger than 500 K), in-place FFTs, or only in "blend" mode? If it crashes faster with FFT then you have a CPU issue of some sort. Could be heat, HS, or voltage (lack of).

 

[fell8]

What kind of temps are you running?

 

[imported_fatal]

Is that the 333 or 400 FSB 3000+ CPU?

 

[deveraux]

Seems to completely lockup whenever I run Prime95's small FFTs within 20s. It can pass the blend test though. Seems like its pointing towards the CPU.
When it locks up:
CPU Die: 46-48 C
System : 41-43 C

Idle:
CPU Die: ~41 C
System : ~37 C

System voltages (from nVidia utility):
CPU die: 1.680 - 1.692
AGP : 1.488 - 1.520
Mem : 2.624 - 2.656
NF : 1.616 - 1.632 (what is NF?)
AUX : 1.616 - 1.632

fatal: It is a 333 FSB CPU with a 13x multiplier.

maluckey: what do u mean by disabling the CPU interface? Never heard of that before.

 

[dunkster]

It sounds like CPU-cooling. If you suspect it's just a HS-mounting problem:
- Remove the board from the case.
- Dismount the HS and clean both physical interfaces.
- Re-apply compound.
- Carefully mount and secure the HS to the CPU (It's easier to check proper flush-mount installation outside the case).
- Remount the board.

I'd do that process - with the best-reviewed HS you can afford that fits your case.

In my experience with Memtest86, the most critical test is Test#5. You can save a lot of mem-testing time by repeated looping of Test#5.

Hope this helps!

 

[deveraux]

I am a bit worried about using a different thermal compound (the stock cooler came with a thermal interface pad) and I do not wish to void my warranty. I actually bought another HSF (ThermalTake's Silent Boost) but the fan on that unit is faulty and cannot spin, hence why I haven't used it on the system yet. Therefore, I am returning that HSF and hoping to get a replacement soon. Am currently talking to the hardware people at ThermalTake.

I could use the thermal compound from that HSF but like I said, I do not wish to void my warranty. Can anyone confirm that just applying a different thermal compound will not void the warranty? Thanks.

dunkster: Why do you think its a CPU-cooling problem, especially since this system even locks up with idle temps (41 C)? Any chance the heat is just concentrating on one part of the core?

 

[deveraux]

To those of you that speculated that the chip wasn't being cooled enough, my hat is off to you. I read through the warranty which stated that I am allowed to change the thermal compound and hence I used it from my other fan. I managed to finish prime95's small TFF test 1 (not a big stability statement -- I know, but still better than before).

I have reseated the hsf as properly as I can with a thin layer of thermal compound. I will try running prime95 again later. In the mean time, should I just get a different HSF and void the warranty? If cooling's the problem, I don't really see another way around this.

Suggestions?

 

[maluckey]

So long as the cooler is approved for AMD CPU of no less than the model you have, AMD should honor the warranty. You could contact hem to be sure.

BTW...Thermal pads are garbage for high-end CPU.

As far as P95, when you test, you should try in-place FFT of not less than 1024 k, or try S @ M. either gets you stressed to the max really quickly.

 

[Alex]

ive never had any problems with my nforce2 and barton

 

[deveraux]

Ok, I can't take it anymore. I managed to get the system more stable by re-applying the thermal material but it still errors in Prime95 using the in-place FFT within a minute. And all the random lock ups/reboots still occur so I am going to RMA the lot (CPU, mobo, hsf). This is just silly since I think I might have thrown part of the packaging for the barton CPU. Am emailing them now to check if I can still return it like that.

For everyone that gave suggestions/opinions, thanks a lot for your help/time.

EDIT: Spelling errors