The PSU exploded last monday & Enermax replaced it with a brand new one.
I plugged in the new one & everything seemed fine for a few days, that is until last night.

Now when I play Crysis or HL2 EP2 I can get between 5-9 mins of play before a BSOD & the PC restarts itself. This doesnt always return back into windows & sometimes returns a BSOD before getting into windows.

Its all new hardware except for the pair of 6800s & the HDDs.

AMD chips don't have the capability that Intel chips do for enhancing the error diagnosis.

So how do I diagnose which piece of hardware is faulty?

To answer that question when I have this problem myself, I strip my PC bare to the point where it has only one thing left to be capable of: starting up. So I take out all HDD's except for the one with the OS on it I take out the gfx card, the DVDplayer, etc. Everything.

Then, when the PC starts up nicely, I start to put back hardware one by one. The moment it doesn't start up anynore I have found the faulty hardware. It's an annoying and time consuming process, but it is failsafe.

That would work if it didnt boot into windows which it does.

I can encode & no BSOD.

I've tested the RAM with WMDT & thats fine.

Gaming seems to stress the rig far more than encoding & thats when it BSOD's after 5-10 mins.

I suspect its probably an overheating issue but thats only a guess.

How do they diagnose hardware faults in the shop? Is there some program that monitors hardware & can report after a BSOD what exactly happened?

Error report says: Hardware.. probably RAM, CPU, Board or PSU & thus your suggestion wouldnt help to diagnose any of those as theyre all integral of the basics to run the PC.

Mmm, yeah. With so many choices it's hard to tell. Only replacing the parts to test could help, I think. At least that's what they do in the shop as far as I know (always test with hardware parts that you don't mind if they burn up - sometimes that happens with faulty hardware).

I've isolated the problem for now by unplugging 1 of the gfx cards.
No more BSOD, but I need to test the card I took out to see if its working & if it is, then its gotta be the board or the SLi bridge. (I think).

What gfx card was it?

A pair of PX6800 TDH 256mb
pre upgrade pic:


Things now get complicated as I've swapped the gfx card over & tested the one I took out & its not faulty.

Guess I should now try 1 card in PCI-E slot 2 & see what happens

Maybe the slot is faulty.

Its not the slot, game plays fine with either card in slot 2.

Now to plug in both cards again & leave the Bridge off & see what happens.

At least I've got a good excuse to play Crysis

Well, thats brought back the BSOD, so basically either card plugged into either slot & everything works fine.

Plug in 2 cards with or without the sli bridge & BSOD.

Guess the next thing to try is swapping the 480w out of this PC with 600w that I've just replaced in the game PC & seeing what happens.

Tested this now on Vista & XP (dual boot) & it happens on both so it cant be drivers.



Maybe a bios update might fix it although the latest bios versions state the only mods is updates to add extra CPUs

As I mentioned when you ask me on msn; RAM + CPU = encoding, gaming = POWER > GPU > CPU > RAM. Things like heat are common, but easier to diagnose.

In my expeirence its either a shit power supply (you'll get reboots with no BSOD), bad drivers (you'll get a stop code but it will mention a file) or failing hardware (you'll get a stop code). I think you've excluded the bad drivers by using two os, you haven't listed a stop number and since its SLI its almost certainly your PSU.

BTW, minimum for SLI is 480w if its expensive. Anything cheap, and you'll need a 600w. I wouldn't trust anything less than a 600w for stable power.


Two cards makes me think its your power supply, especially since you said it worked fine before.


Bring up a voltage monitor and stick it on a second screen with your game on your main screen. in fact throw up as many diagnostics as you can.

The 480w is a Thermaltake Pure Power, its as old as the case but it was used to run these 2 cards (no extra power sockets on them, just a 4pin molex socket on the old board for SLI) for a while before I built the 2nd PC so I know it works with these cards at least.

The rocker switch on the back of the Enermax broke yesterday so I RMA'd it & the new one arrived today... yes TODAY!!
I called them up, told them the new one was faulty, they sent someone to collect mine & sent me a new one by courier the same day. Cant beat that for a quality RMA service

PCProbe is now reporting the +12v at 11.97-12.03v where the one I sent back was only 11.90v (maybe thats the problem?)

I've updates the BIOS to the latest which added support for 1066 RAM but this was bugged & kept locking up windows so I've gone back to an earlier one (no 1066 option but set as auto it runs at 800) & with 1 GPU fitted on the 480w ran stable for several hours with no issues at all. Only issue is that on POST its showing my RAM as 800mhz but CPU-Z is reporting it as PC-6400 (400mhz) which is confusing.

I did the Crucial online tech support & the person I spoke to was reading from a flow chart by the looks of it & didnt have a clue except to say the RAM should run at 1066 in this board & if its reporting the above its PROBABLY faulty.

Could CPU-Z be reporting it wrong? or is it more likely faulty?

I've contacted ASUS with regards to being unable to plug 2 cards into the board with SLI disabled let alone enabled causing BSOD & lockups, but they take up to 48hrs to reply from Taiwan as theres no Tech Support in the UK that you can call for help.

I'll run it a couple more days with 1 card in & wait for ASUS to (hopefully) reply & see what they say as I'm afraid I'm gonna break something now & to RMA stuff back to the shop I bought from, thats not found faulty, is gonna cost me in the region of £30 as they charge you for testing & postage both ways.

As theres no case speaker I went down the local shop & asked if they had a scrap case I could have which he did & gave me for free.
Opening the case it was a complete PC...
anybody want an old AMD K62 450 or a 4gb quantum fireball HDD? its also got 64mb of sdram

Fixed

New PSU runs both gfx cards in SLi rock stable now after some minor tweaking.
I'm guessing that when the old PSU exploded it messed up the bios in the board which needed flasing & manually setting up for the RAM.

Thanks to all who offered advice with this

@spud, looks like you were right as usual, you clever bugger

I don't like getting things right, people start giving you more work to do

I've had a couple more BSODS. Now its crashing proper without shutting down, I can see the code 000blah00124. I fixed it by slackening off the memory timings from 5-5-5-15 to 5-5-5-18. Poxy incomplatible RAM.. mumble mumble

Here we go again, ugggh!!

As you know, everything was nice & smooth at last, that was until this morning when I decided it was time to replace the pair of 6800s with a brand new G92 BFG 8800 GTS 512Mb OC.
I fitted the card, plugged into 1 of the monitor sockets with a supplied adaptor & everything went pink . I plugged into the other socket & that fixed the colour problem.
Decided to try out Crysis & the gfx difference between the 2 setups is unbelievable, but... 5 mins later it rebooted & back to BSOD so I took the card back for exchange.
This time the 1st socket turned everything blue so again I used the 2nd socket. Again 5 mins into crysis.. crash central but this time it got worse & started locking up in windows.
I've taken the 8800 out & replaced it with a single 6800 just to see if I accidentally messed something up on installation & that runs smooth as before with crap gfx, so what is the problem this time?

Could the card be incompatible with ballistix memory or this board? or is it more likely to be such a juice sucking whore of a card that even this damn expensive PSU cant feed it?

God I hate PCs, Grrrrrrrrrrrrrr

I've had that problem before too with the screen going all pink. It's
usually the adapter coming loose or the crt connection coming loose.
But I guess that is not your main problem...




Well a couple of things:

What is the wattage on the new PSU?

Did you upgrade to the newest drivers for your 8800s?

What's the board? An asus?

Are you running Vista or XP?

I'm sure we can nail it down.

Thanks for the reply Oberon.

I think I've fixed it. Due to the fact I am using Vista on it & the control panel is different to what I'm used to, I unistalled the drivers for the 6800s then removed the cards. Fitted the 8800 & installed the drivers from the supplied CD followed by the latest 169.25 drivers from nvidia site.
After returning the 8800 yesterday I simply swapped the cards over which I guess is where I went wrong. After going into device manager & removing the hardware & drivers & rebooting then reinstalling the drivers everything has been running stable.
It would seem that this card is very hungry & needs 28A on its seperate 12v Rail & the Enermax Noisetaker 2 600w that I have has a combined 35A, so I wont be able to SLI without upgrading the PSU, although I have no intention of adding a 2nd card this time so its not a problem.

I'm now able to run Crysis on High gfx detail with very little freezing & I'm sure if I turn off the AA that will stop. What a card! (Faster than the original GTX due to being G92).

http://www.bfgtech.com/MHandler.ashx?proimageid=2216
http://www.bfgtech.com/bfgr88512gtse.aspx

They are available for under £200 here: http://www.scan.co.uk/Products/ProductI ... tID=713729

It's a nice card but it's not faster than the GTX, although there's not much in it to be honest. Value for money wise the GT pisses all over the GTX.

I'm glad to hear it's all fixed! I had a similar problem with my 8600 and
the drivers. When I installed mine, I installed the drivers that had come
with the card. Upon the first updated drivers, my compy booted to a BSOD.
I had never had that problem with nvidia drivers where I had to uninstall the
old drivers to update. But all is well for now...