Software T & L stops freezing with a VGA in one particular system, diagnosis ??

[daveqb]

its GF2mx400 (hey its not mine) PCI card (he's got a brand name PC

it freezes up in games and 3DMark 2001SE (not that he runs it, but i use it to test it)

the onboard never had a problem but only renders in low tests @ 4-5 fps (the gf2mx400 does 40fps at least)

i tested the card in other systems and it runs fine, only freezes in this dudes system (C1.3, SiS chipset(?), 512 DDR Ram)
NB, CPU and RAM dont get hot, GPU does.

ok, now the interesting thing i found was looping 3DMark using Software T & L seems to (so far) fix his freezing problem.

to me this sounds like a VGA problem, perhaps getting too hot.
But i ran it colder in his system, a few fans, then the others that didnt freeze.

what is software transformation and lighting doing that saves the freezing??

faulty card ?? coolbits allows 0-100mhz mem setting

any hints would be appreciated

 

[BenSkywalker]

what is software transformation and lighting doing that saves the freezing??

Reducing the power strain on the PCI slot? It sounds like it is an issue with not enough volts making it to the board, particularly if it working fine in other people's rigs.

 

[daveqb]

hmmmm

maybe bad PSU ???

rails looked ok in the BIOS, but i must check with my voltmeter.

well i managed to get it stable, but at a price.

1) needed PC100 ram as apposed to the PC133 RAM it had.
so now running sync with FSB

2) underclock the VGA servely. downed it from 200/333 to 110/110

and with this it looped 3DMark for ~1hr freeze free


makes it seems like the NB and memory subsystem cant handle high bandwidth

idea's ?

thanx

 

[daveqb]

PS

software T & L did indeed freeze today too.

 

[BenSkywalker]

rails looked ok in the BIOS, but i must check with my voltmeter.

The trick here is when the board is being strained if it is a power delivery issue. Do you have a motherboard monitoring tool handy? Install one if you do and enable logging and then fire up the apps that crash and exit out before it actually goes down and check the log for the 3.3 rail and see what is happening. If it is dipping too much, then that could be your problem.

If the condition is getting progressively worse, as it sounds is the case, then it could be the PSU. You could try to throw in a decent PSU just to test it out and see if the situation improves, although it may end up doing nothing and being a waste of time(just a suggestion, I know how it is trying to track something like this down ).

When it locked up running soft T&L, what was the clockrates on everything at and how long did it run before it went down? With it going down now running soft T&L, it sounds like it is either a PSU headed to the grave or it is possible that you have a gfx board on its way(although I find the PSU the more likely culprit).

 

[MercenaryForHire]

gF2 PCI cards have had a notoriously bad history of hardware T&L, but it's mostly related to horrid performance (saturated PCI bus in a split-second) and not hardlocks. Your GPU may be on the way out, especially if underclocking fixes it. Try hardware T&L at the lower clocks, and ask your friend if he ever OCed the card/memory/PCI bus.

- M4H

 

[daveqb]

thanx guys, i'll be back it for a short while tmw,

he's never O/C a thing in his life and the card is near new ( i installed it for him last week)

yes, the volt rails does seem a good plan of attack.

MBM supports logging hey ?? ( i use it, but i dont log)

it locked up in soft T & L at the 20's low 30's frame rate, hmmm bout 15mins i think.

and yeah, these are all hardlocks.

how does the PC100 help stop the freezing ??
makes me suspicious of the NB, but its not even near hot (the other system i tested in, a P4, the NB got quite warm)

thanx guys