I suppose your motherboard doesn't come with the 12v 4p connector? I can't tell from the picture but many nforce boards had this connector.
I had to look up your PSU and it seems to provive just 18A on the 12v line, which isn't a lot for an Athlon XP. If it does have the 12v connector then the 5v line shouldn't be an issue (unless for some reason one of the 2 12v lines isn't being used. It's early so I can't think straight yet 🤣

Do you get a BSoD? In XP theres an option to not automatically reboot when a BSoD appears.

Tetrium wrote:

I suppose your motherboard doesn't come with the 12v 4p connector? I can't tell from the picture but many nforce boards had this connector.
I had to look up your PSU and it seems to provive just 18A on the 12v line, which isn't a lot for an Athlon XP. If it does have the 12v connector then the 5v line shouldn't be an issue (unless for some reason one of the 2 12v lines isn't being used. It's early so I can't think straight yet 🤣

Do you get a BSoD? In XP theres an option to not automatically reboot when a BSoD appears.

It does come with a 12V CPU power connector. 18A on 12V line is 216W max, way more than the system can chew. A 3200+ Barton draws up to 77W and 9800Pro under OC and heavy use draws up to 53W. That's only 130W if my data and math are correct...

No, no BSoD, just POST. How do I enable it? I use an official XP SP2 at deafult settings.

Do your voltages look good in the BIOS? The readings aren't always accurate, but still, check that. If you have any rails sagging, that could be the cause of the resets.
Gigabyte boards in that time period usually have questionable caps. I think OST are pretty common on those.

According to an old xbitlabs article, the 9800 Pro draws more from 5V than it does from 12V. I used the same card on an ABit AN7 nForce2 board with a 300W PSU, and all the voltage rails were solid. I doubt it had 18A on the 12V rail. It was primarily a 5V PSU.
Power supplies aren't very straightforward though. They're all a little different and every motherboard can also draw from them differently. I find it difficult to know for sure which ones are going to work well in a particular system, until I just try it.

RacoonRider wrote:

No, no BSoD, just POST. How do I enable it? I use an official XP SP2 at deafult settings.

Control Panel
System
Advanced tab
Startup and Recovery->Settings
Under "system failure", uncheck the box "automatically restart"

This way, if there's a BSOD, it will actually let you read it instead of immediately resetting.

Just hunt down an old Enermax unit with a strong +5 rail and you are set, shouldn't cost you more than $30-50 locally or off amazon for one.

Okay, the automatic restart is now off. Considering 9800 Pro, I've got 2 of them and both have light burn marks (yellow plastic) around 12V molex connector pin. That's the reason I thought there was much current, probably over spec.

nforce4max wrote:

Just hunt down an old Enermax unit with a strong +5 rail and you are set, shouldn't cost you more than $30-50 locally or off amazon for one.

Oh that would be difficult. Enermax is a pricy brand and I doubt anything like it was ever sold in Omsk back in the day. $50 is the retail price of the Zalman PSU and it's way over spec.

Btw, the board uses Sanyo caps. Again, not an expert, sounds fine to me. Not Rubycon, but a sound brand all the same.

I looked up that Zalman unit and its not good but keep an eye on the volts big time as you only got 120w between the 3.3v and 5v rails at the most optimistic specs. Just give a chance to search for a better unit but everything else is pretty good so keep up the good work.

Sanyo caps are among the top brands afaik. 😀 I also don't think it's an issue on the +5V. PSU output quality? Maybe. I'm afraid you'll find less than stellar components on that Zalman PSU (I've seen another Zalman model with bigger wattage and was... seriously not impressed). 😒

That burn marks on the 9800's connector can't be ignored definitely. Btw, if you run the system without the case fans (open bench like in the pic above), would the problem still occur?

Also just throwing out ideas, how's your AC mains stability? In my parents' house, that's usually the problem as we would sometimes get unstable power, combined with heavy loads from other home appliances running at the same time. This kind of conditions can seriously test the PSU's output quality, putting stress on the components (yeah, our crappy mains made me learn the hard way not to skimp on PSU and UPS 😁).

The temps aren't awful - could be better in terms of ambient (35* C is kind of high, but you have the heat on), but mid-40s is nothing to be worried about for the 9800 or AXP as long as the sinks are properly contacting them. I'd look at the PSU or the 9800 (if it has burn marks on it) - do you have another graphics card you can try? Or another PSU? Have you run memtest on the system?

On this cracking sound - is that from the hardware itself, or thru the speakers? If it's thru the speakers, look at your sound-drivers - it may be a compatibility issue there (I remember when I got my Audigy that was sometimes a problem, but driver updates almost always fixed it).

archsan wrote:

That burn marks on the 9800's connector can't be ignored definitely. Btw, if you run the system without the case fans (open bench like in the pic above), would the problem still occur?

Also just throwing out ideas, how's your AC mains stability? In my parents' house, that's usually the problem as we would sometimes get unstable power, combined with heavy loads from other home appliances running at the same time. This kind of conditions can seriously test the PSU's output quality, putting stress on the components (yeah, our crappy mains made me learn the hard way not to skimp on PSU and UPS 😁).

I'll try using UPS, I can steal one from my girlfriend 😀

Testing it on open bench would take a lot of time since the problem occurs oh so rarely 😀

obobskivich wrote:

The temps aren't awful - could be better in terms of ambient (35* C is kind of high, but you have the heat on), but mid-40s is nothing to be worried about for the 9800 or AXP as long as the sinks are properly contacting them. I'd look at the PSU or the 9800 (if it has burn marks on it) - do you have another graphics card you can try? Or another PSU? Have you run memtest on the system?

No, I haven't run memtest, but I'll do that once I have time to fiddle with it 😀 I've got Ti4200, Radeon 9550GE (128-bit bus), G450, might try some of them soon.
__________
For now I set up normal BIOS settings, as it appears, the "fail-safe" includes 100MHz FSB. Changed to default 166MHz for this CPU. I was expecting that higher settings might enforce the system to crash earlier so I can see BSoD. So far so good. The CPU load during gameplay now sometimes drops below 100% and I did not encounter any random restarts. I'll keep you informed, it would be fun if 100MHz FSB was the reason for this odd behavior.

P.S. Found a 3200+ on a local auction site for a good price, decided to go for it. Now it's on its way to me from the other end of the country!

Hope the 3200+ works out, they're nice chips. Will keep an eye for updates as you have time to fiddle with things - if I had to pick a card from your collection for testing I'd try the 9550 first (since it should use the same drivers as the 9800), and if it still crashes, uninstall the ATi drivers and try the GF4. If that fixes it, it was the ATi drivers (if the 9550 by itself fixes it, it was the 9800). I'm not sure if the G450 would be up to playing games that the GF4/Radeon9 can handle so I don't know how well it would do for testing purposes.

You may also try the new CPU - maybe it's the XP 2600 itself that's the problem?

Oh, another quicker thing than memtest you could do - can you try with some of the memory removed? May be a bad module in there.

Tried running 166MHz FSB for two evenings for now. Still no reboots. I don't get it, my hardware never failed from running below spec.

The problem was that the PSU did not supply 9800Pro with enough power. With 9600Pro and Ti200 the system is rock stable. I swapped the PSU with an older Inwin with 35A 5V line and it is fine ever since.

Thank you guys, I'm sorry for being stubborn and not willing to acknowledge the PSU was not enough.

Hmm, whether it's of relevance or not, but my Athlon XP based system would BSOD Windows 7 and not Windows XP if I used the third DDR RAM slot. Maybe you could check the RAM as well?

ReeseRiverson wrote:

Hmm, whether it's of relevance or not, but my Athlon XP based system would BSOD Windows 7 and not Windows XP if I used the third DDR RAM slot. Maybe you could check the RAM as well?

I could, but it's rock stable already with the "new" PSU 😀

On both Nforce 2 boards I've had (Leadtek K7NCR18D and Asus A7N8X-E), memory has always been the biggest factor in system stability. Despite being rated for DDR 400, you couldn't count on the memory controller being stable running, for example, 4 sticks. It also depends heavily on memory type. G.Skill PC3200 RAM was usually not an issue, so I could run 4 X 512MB in Dual Channel; but Corsair Value RAM wouldn't run at 400 reliably, only 166/333.

Also, my Corsair RAM would consistently produce memory errors in test 5 of Memtest while in D/C @ 400mhz; but not the G.Skill. Yet it tested fine in other boards! WTF? I came to the conclusion that the NForce memory controller was at fault.

FWIW, the most stable CPU/MEM combo in any of my systems was anything other than 400fsb. Also look up your actual RAM timings and set them manually in the BIOS if you can. Sometimes SPD detects them wrong; or even has the wrong voltage.

duralisis wrote:

On both Nforce 2 boards I've had (Leadtek K7NCR18D and Asus A7N8X-E), memory has always been the biggest factor in system stability. Despite being rated for DDR 400, you couldn't count on the memory controller being stable running, for example, 4 sticks. It also depends heavily on memory type. G.Skill PC3200 RAM was usually not an issue, so I could run 4 X 512MB in Dual Channel; but Corsair Value RAM wouldn't run at 400 reliably, only 166/333.

Also, my Corsair RAM would consistently produce memory errors in test 5 of Memtest while in D/C @ 400mhz; but not the G.Skill. Yet it tested fine in other boards! WTF? I came to the conclusion that the NForce memory controller was at fault.

FWIW, the most stable CPU/MEM combo in any of my systems was anything other than 400fsb. Also look up your actual RAM timings and set them manually in the BIOS if you can. Sometimes SPD detects them wrong; or even has the wrong voltage.

I have a bunch of identical Micron / Apple branded DDR400 memory modules which always seem to fail memtest at DDR400 when tested in an ABit AN7 (nForce2 Ultra 400), yet pass in a DDR400 P4 motherboard (Intel 875P I think). The nForce2 motherboard is stable - I used it as my desktop for a long time, and the RAM works reliably also except in that scenario.
The problem was so consistent that I believe it might be a compatibility problem with the nForce2-400 chipset like you're describing.

I remember JEDEC was slow to officially adopt a DDR400 specification. DDR400 was starting to gain popularity as an informal speed grade while JEDEC was still resisting it. When JEDEC did finally adopt a spec, they even bumped up the target voltage compared to what it had been on DDR333 and below. This suggests that there were stability concerns with this type of memory.

This all leads to my suspicion that perhaps the nForce2 has an off-spec implementation that doesn't quite work right with all DDR400 modules. It was after all one of the first generation of DDR400 chipsets.

Concerning DDR400, there's a BIOS option for asynchronous DDR/FSB rate. Funny thing though, it works "down", but never "up". I tired usung DDR400 modules as DDR400 at FSB400, then as DDR400 at FSB333. The second variant hanged at startup 😀