• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

BFG 8800GTX locking up

Man of Honour
Joined
18 Oct 2002
Posts
12,310
Location
Vvardenfell
About a week ago, my graphics card started locking up during gfx-intensive work. It started with Oblivion, but has slowly got worse. When the problem started it would play an old game like Postal 2 at 1600 x 1200 without issue, now even that locks up. I can run Firefox for hours, but more than ten minutes of Media Player running movies will lock up the graphics.

First of all, please note that it is ONLY the gfx which is locking up: if I access the machine via another on the network, everything is fine, but the screen is frozen. It will respond to ctrl-alt-del, but only very slowly - usually after several minutes.

The card is a BFG 8800GTX OC. Clocks are default: 600/900, and the fan duty cycle is 100% for everything. The GPU does not appear to be overheating: temp monitoring programs say that the temp never goes above 58oC, and is usually lower than that. Right up until the temp monitor display freezes along with everything else, anyway.

It happens with 158.19 and 158.22 drivers, but never used to happen with either. Note that nothing (hardware or software) was added in about two weeks before this happened. The Windows logs shows no errors except to say that the app (Oblivion) was unexpectedly terminated (because I used ctrl-alt-del to kill it)

The less the gfx card has to work, the longer it takes to happen: Oblivion at 800x600 and lowest settings takes about two minutes, but 1152x864 and highest will often lock during the load screen, and certainly within seconds of finishing loading.

Turning the gfx clocks right down has no effect. Turning the system oc down has no effect. Turning the RAM settings down has no effect. I can run SuperPI to 32 million DP with no issue.


System is:

E6600 at 400 x 9.0
2GB OCZ 1066 RAM
Asus P5N32-E SLI
etc
Antec TruePower 650W.

Windows XP Pro.


Edit: yes, the fan is turning! RivaTuner says 2800rpm IIRC.


My immediate thought is that the card is borked. My next plan is to move it to another machine to see if the problem moves or stays, but unfortunately that means driver playing as the other machine has a 7900GT in it. And changing the drivers means that I can't be sure it's not a driver issue - although I'm 99% certain it isn't because it's over a month since I loaded the 158.22s and the problem only started a week ago or so.

All the googling I've done throws up cards which are overheating, and this does not seem to be.



Anyone have any ideas?

Cheers

M
 
Last edited:
Permabanned
Joined
19 Jun 2007
Posts
10,717
Location
InURmama
Unless card has suddenly went bad, but ram would show up artifacts and core normally wont die, try your PSU, get multimeter reading for the 12v and 5v on load using a spare molex, for 3.3V your going to need dig in the 24pin connection to find the 3.3v line which can be found on guides with google.
 
Last edited:
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
helmutcheese said:
get multimeter reading for the 12v and 5v on laod using a sapre molex, for 3.3V your going to need dig in the 24pin connection to find the 3.3v line which can be founf on guides with google.
Simplest place to check is down the back of the 24 pin connector on the board header.

DMM black to black

DMM + red to

orange 3.3v
red 5.0v
yellow 12.0v

Tolerance is +-5% from nominal.
 
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Safer than a floating molex and no different to the modular outlet. Its always a good habit when taking readings not to have to actually touch anything, I've enough HT burns to prove it. You simply go down the 'back' of the fixed connector attached to the header it can't move. Just push the probe down black pin 24 then go orange>red>yellow.
 
Last edited:
Man of Honour
OP
Joined
18 Oct 2002
Posts
12,310
Location
Vvardenfell
But if the voltages were bad, why would a 3.6GHz overclock be completely stable? Just out of interest... Although I could understand an issue with the supply down the PCI-E cables.


M
 
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Meridian said:
But if the voltages were bad, why would a 3.6GHz overclock be completely stable? Just out of interest... Although I could understand an issue with the supply down the PCI-E cables.


M
Couple of things. Could be different rails. 12v1 is 'supposed', according to ATX spec, to feed the cpu. 12v2 et al, subsequent rails are for graphics etc. Could be there is some form of regulation holding the cpu rail at nominal, while the other rail is overloaded and drooping. I doubt that one tbh. The other possibility is the cpu load even overclocked doesnt get near the load from the graphics card (with cpu) at full tilt. Check the voltages you'll soon see.
 
Last edited:
Man of Honour
OP
Joined
18 Oct 2002
Posts
12,310
Location
Vvardenfell
Believe it or not, I don't have a DMM at home, so if I'm going to play this game then I need to liberate one from work - which means tomorrow. In the meantime, the admittedly suspect MBM says the ranges for the voltages are:

+5V = 4.97-5.03V
+12V = 11.65-11.71V

BTW, isn't it 5% for 3.3 and 5V, but 10% for 12V? That's the figures I heard.

3.3V isn't showing up. As I said, not reliable, but an interesting indicator.


I might also point out that even when the graphics are run on settings that wouldn't strain a Voodoo5 much, it still happens - it just takes a bit longer. And as I said, Media Player causes the same issue - hardly a huge power-hog.


Anyway, I'll try to borrow a DMM, but the vid card swap comes first (both rigs have the same PSU).


M
 
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Meridian said:
+5V = 4.97-5.03V
+12V = 11.65-11.71V

BTW, isn't it 5% for 3.3 and 5V, but 10% for 12V? That's the figures I heard.

3.3V isn't showing up. As I said, not reliable, but an interesting indicator.
The spec is +-5% for nominal regulation and +-10% for short instantaneous peak loads on one of the rails (cpu iirc), continuous values are still +-5%. No one seems to follows the specs anyway.

Having said that I wouldn't be happy with 5% tbh, and your 12v is low, not that you can trust the sensor readings, but its not drooping so much. Better a little over than under, its regulated back to 12v.

Thing is, like you say underclocking the card and cpu would lower the load a lot. If the voltage holds, but it still does it after increasing lengths of time its not the psu. More likely something is getting warm and failing. Does sound like a borked card :(
 
Last edited:
Associate
Joined
6 Apr 2007
Posts
553
Location
Manchester
Hi,

I had this problem with my card locking up after 10 - 15 mins of play and gradually got worse, tried lots of things new drivers etc, however i got rivatuner and upped the core by 50 and upped the memory by 100 and not had a problem since, don't kno wif this will work for you but it did for me :)
 
Man of Honour
OP
Joined
18 Oct 2002
Posts
12,310
Location
Vvardenfell
Nope - raising the speeds does little except bring it on faster. It all points to overheating, except it has to be the RAM because the temp monitor shows no signs of a rising GPU temp. My only worry is: why now? The weather was much hotter a couple of months ago, and it worked just fine then. And it's getting worse. Only thing I can think is that the cooler is separating from the RAM.


M
 
Associate
Joined
17 Nov 2006
Posts
123
Location
Manchester
I had the exact same problem with an 8800GTX BFG and the P5N32-E SLI, turned out it was the motherboard causing the problems. Lock ups would occur in most games putting a large amount of load on the card, and in some cases i'd receive an nv4_disp BSOD. I tried various forceware driver versions, clean install of WinXP etc to no avail.

As you may have seen on these forums, the P5N32-E SLI is extremely picky with RAM settings/timings. Is the BIOS flashed to the latest release for the board? (albeit most are classed as beta revisions)

Also, post your memory timings, perhaps they need a little tweaking.

Regards
 
Permabanned
Joined
19 Jun 2007
Posts
10,717
Location
InURmama
Thats why I stated " I Think ", most reviews show the ATA 2.X are within 3%, (Thermaltake for one) 10% on a 12v is far too much IMO.

Again he still has not stated if FSB is overclocked or not.
 
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
Last edited:
Permabanned
Joined
19 Jun 2007
Posts
10,717
Location
InURmama
I have tested all my PSU's with multimeters on idle then full load and this Enermax does not move 1 bit and all readings are actually on the + side :)

I did think ATX 1.0 was 5% and ATX 2.X was 3%, I also would not want that drop as I OC.
 
Last edited:
Soldato
Joined
3 Nov 2004
Posts
9,871
Location
UK
The ATX spec is very loose to be honest. Any decent manufacturer should be miles ahead of it and I've yet to see a psu that sticks to the specs exactly. Especially when it comes to rails, seems its treated more of a guideline as it were. +ve voltage (not too much) is better, board/card regulation clips to the nominal voltage. Under voltage means more current if its available or less power if its not. Ripple voltage is just as important if not more so and needs an oscilloscope, a 3.5 digit DMM can't see it.
 
Last edited:
Back
Top Bottom