Another POS 5970 off Ebay - alternatives to sending back?

[Nancarrow]

I paid the princely sum of £309.46 for this 5970 card. That includes a f king customs charge of £25.37. Delivered today. I have a ragamuffin rig with three 5970's and a well-to-do rig with so far only one working 5970, but room for three more.

When I put the new card into my fancy rig with the fancy mobo and fancy psu, the fans spin for a fraction of a second before everything goes dead. If I switch off the PSU for a while, then try again, same thing happens.

So I thought I'd swap out a card from my other rig, the one from the wrong sort of neighbourhood. I put this new 5970 in there, and switch on. Hey, it keeps going! Nothing's cutting out. That sounds promising, yes?

Well I instantly realised I'd better keep my nose on red alert, and sure enough, a few seconds later, the pungent smell of burning plastic. Yanked the power cord out sharpish and no mistake. Examining the 16x-16x riser cable shows that the pins supplying +3.3V, all of them and only them, look and smell burny. Better not use that cable again.

I don't mind about the cable, it was cheap and I have more... but what exactly is wrong with the card, and is it something I could fix with the loan of someone else's brain?

ETA: should add that with a mug of hot cocoa and some soothing trance music, both rigs have recovered fully from their abuse.

[rjk]

I paid the princely sum of £309.46 for this 5970 card. That includes a f king customs charge of £25.37. Delivered today. I have a ragamuffin rig with three 5970's and a well-to-do rig with so far only one working 5970, but room for three more.

When I put the new card into my fancy rig with the fancy mobo and fancy psu, the fans spin for a fraction of a second before everything goes dead. If I switch off the PSU for a while, then try again, same thing happens.

So I thought I'd swap out a card from my other rig, the one from the wrong sort of neighbourhood. I put this new 5970 in there, and switch on. Hey, it keeps going! Nothing's cutting out. That sounds promising, yes?

Well I instantly realised I'd better keep my nose on red alert, and sure enough, a few seconds later, the pungent smell of burning plastic. Yanked the power cord out sharpish and no mistake. Examining the 16x-16x riser cable shows that the pins supplying +3.3V, all of them and only them, look and smell burny. Better not use that cable again.

I don't mind about the cable, it was cheap and I have more... but what exactly is wrong with the card, and is it something I could fix with the loan of someone else's brain?

ETA: should add that with a mug of hot cocoa and some soothing trance music, both rigs have recovered fully from their abuse.

You should probably be using powered risers. The wires that burned are likely the 12v bus, not 3.3v. As to why that card is different from the rest, I can't be sure.

[Nancarrow]

Nuh uh, I counted the pins against Wikipedia's description of which pin is which, relative to the notch, and it most definitely was the 3.3v pins, not the 12v ones.

It could have been worse. I could have used no riser at all on that rig. Then whatever nastiness had happened to the riser, would probably have happened to the bus on the mobo. Though it was a dirt cheap mobo.

[rjk]

Nuh uh, I counted the pins against Wikipedia's description of which pin is which, relative to the notch, and it most definitely was the 3.3v pins, not the 12v ones.

It could have been worse. I could have used no riser at all on that rig. Then whatever nastiness had happened to the riser, would probably have happened to the bus on the mobo. Though it was a dirt cheap mobo.

How very interesting. I had heard that 3.3v wasn't even used on those cards, so I have no idea what may have shorted out to cause that problem.

[AzN1337c0d3r]

Nuh uh, I counted the pins against Wikipedia's description of which pin is which, relative to the notch, and it most definitely was the 3.3v pins, not the 12v ones.

It could have been worse. I could have used no riser at all on that rig. Then whatever nastiness had happened to the riser, would probably have happened to the bus on the mobo. Though it was a dirt cheap mobo.

How very interesting. I had heard that 3.3v wasn't even used on those cards, so I have no idea what may have shorted out to cause that problem.

AFAIK, on modern ASICs, 3.3V isn't used for anything. Maybe go over the PCB of the card with a fine-toothed comb for the pins that burned and see if there is a short somewhere?

(Although I would suspect the amount of current to burn one of those cables will usually vaporize traces on the card... hmm)

[Gomeler]

3.3v is the supply voltage and signaling voltage on the VRM controllers for these cards. Without 3.3v your VRM controller will not power on and switch the fets. Good luck running your GPUs then