Fixing a non-booting graphics card at the component level?

They make noise, too.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Fixing a non-booting graphics card at the component level?

Post by edh » Mon Jul 22, 2013 11:31 am

I have an old 6600GT AGP. It's 8 and a half years now and it's just 'decided not to work anymore'. It was running Furmark and only at 71C on the core given my 5 volted Socket A heatsink mod 8) . Then it froze. Now it doesn't boot at all. Cross-testing with other hardware shows that it is the card that has failed.

It's definitely not the GPU itself overheating due to the heatsink and memory generally doesn't just go pop so I'm looking at the ancillaries on the card and as this is an AGP model with the HSI bridge, there's quite a few! Maybe something else got to hot even though the core was running cool.

I don't like throwing things away and as it's already out of service I thought I'd have a crack at trying to diagnose and fix the board itself. I've already removed the cooler, totally cleaned up the card and reassembled but to no avail. The next stage would be trying to measure voltages on different parts of the card to see if it is getting voltage through to the core and memory. What I would like to try is the solder reflowing technique popularised by people putting cards in their kitchen oven and just melting the solder enough to fix any bad connections. The card is already not working so I can't make it worse so no harm in trying.

Does anyone have any experience in fixing graphics card problems? Cooling issues excluded of course, we can all do that. :P Any experience with solder reflowing or similar techniques? Any thing that worked or didn't work?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Fixing a non-booting graphics card at the component leve

Post by CA_Steve » Mon Jul 22, 2013 3:49 pm

Check for odd looking caps in the VRM circuitry. Always a weak point.

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Fixing a non-booting graphics card at the component leve

Post by edh » Tue Jul 23, 2013 1:20 pm

CA_Steve wrote:Check for odd looking caps in the VRM circuitry.
The caps all look OK, there's only one of the type that tends to bulge and blow. Caps is a good suggestion though so I have cracked out the multimetre anyway to investigate. Cpacitors don't have a steady resistance as such and so sensible measurements of impedance have to be done with an oscilloscope which I don't have. Sticking a multimetre across them gives massive fluctuations between the hundreds and thousands of ohms. I did find something interesting though in that 2 of the SMD capacitors, right next to the heatsink area actually have a steady resistance of 4.8 ohms! In this case they're maybe shorted slightly so therefore not working as a capacitor. They're each 1500uF 2.5V and I have tracked down some possible repalcements but I might have a look at Badcaps first to get some more advice before attempting. Also I need to find a suitable source, closest replacment I've found is about £1 each.

For reference here is the card with the stock cooler, the 2 suspect caps are the big ones just to the right of the heatsink:
http://www.guru3d.com/articles_pages/galaxy_6600 gt_agp_128mb_review,6.html

james
Posts: 22
Joined: Tue Jun 11, 2013 9:46 am

Re: Fixing a non-booting graphics card at the component leve

Post by james » Wed Jul 24, 2013 6:13 am

With the caps still in circuit, you are probably measuring the resistance of the rest of the circuit.

Also, measuring the resistance of a capacitor is a pretty fruitless exercise, it'll only show if they've failed short. Cheap capacitors just tend to loose capacitance, so you'd need a capacitance meter.

In my opinion, it's not worth the effort. If it is a cap or something, that makes it easy, but it tends to be the chip that fails, which is pretty much impossible to replace, especially at a home level.

I would be very suprised if it was a solder joint gone, the die would have to reach 217+ approx to melt the solder! Common on Xboxs, when there's no temperature sensing, but not a problem here!

edh
Posts: 1621
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Fixing a non-booting graphics card at the component leve

Post by edh » Wed Jul 24, 2013 8:46 am

The GPU was only running in the low seventies at the time of failure so it is unlikely to be acute GPU damage. Similarly my worry about solder is more about the joints of all the ancillary components which may have got hotter, even if 217C seems unlikely. I can't say it's definitely the caps but it seems worth trying as they're the only replaceable component. Testing to show they're failed is an important part of that and I'll try badcaps for further help on that.

Post Reply