[jan-jaap]

I would leave the PCB alone, but machine a new heat sink instead that's not kapton-taped to the TRAM, putting stress on it, but serves as a structural element instead. This is what SGI did for the Octane IMPACT boards and you rarely hear about broken Octane TRAM.

[Irinikus]

I still have to acquire TRAM’s for my MAX IMPACT card. When I do I’ll look into this. (Not a priority at this point, but on the to do list)

[sotlampr]

Thanks for the warm reception 😀

Sorry for not making my self more clear,
@Irinikus
got what the situation is. I have a serial console to the machine, that’s how I got the terminal output in the photos.  Everything is performed on a wiped disk (and problem persisted).

Diagnostics do not complete with the graphics board (Stuck at “Testing Indigo graphics” and amber light), while they complete fine without.

While messing around I got this screen a few times      . Could be that the card becomes loose (no screw at ) but I think something’s not right… it’s funny how only the lower byte is zeroed - is this a proof of a hardware issue?

The TRAM operation sounds interesting - I don’t have a lot of experience in chip soldering but I would be willing to sacrifice the card in the case it is bricked😄

I also noticed the heat sinks on the upper board get uncomfortably warm really quick…

UPDATE

Ok I can reproduce this error whenever executing 

Code:

/usr/gfx/stopgfx

   

[robespierre]

VC3 is the video controller chip on the High Impact board. The test failure shows that 8 signal lines to or from it are just dead: Did you check that the flex cable between the two cards is properly snapped in place?
If that is not the cause it can be (perhaps) a bad ram chip or bad solder joints on the VC3. Sometimes these Impact boards are damaged by burning tantalum capacitors.

[weblacky]

In regards to those thin film brown ribbon cables that connect the boards, there was a posted case once where another user was having intermittent issues, the cables were fully seated and locked, so appeared fine. They ended up removing each connector in turn carefully anyway and found that there was in fact a totally bent pin in the board connector. After carefully bending it back straight up and down and reconnecting the cable, their problem went away.


I cannot find the posting right now, but I distinctly remember it. It showed it was entirely possible to have a seated connector while one of the male pins was folded sideways.

So carefully disconnecting the ribbons and checking the connection may actually yield results. It did in at least one case. However in that case the system was newly acquired and the previous owner was the culprit.

Also, it may be possible to use a multimeter on the boards’ test points around the connectors to check for continuity between boards with cables in place, as well the shorts to neighboring pins. Thought both activities maybe called for here.

[sotlampr]

Hello everyone I’m back with an update.

I took the boards apart and cleaned them and the situation didn’t change much. The connectors looked good as far as I could tell. At some point I got this error running the diagnostics     
but I mostly get freezes.

I’m also getting this now instead of the vc3:     



@weblacky could you help me finding the test points you mentioned?


   

Any other ideas? It seems very difficult to find that stuff online, don’t even know where the TRAM is 

[weblacky]

By test points I really meant the exposed legs or solder points (based on the connector).

Before you rip into this card, get a cheap solid IMPACT graphics board from eBay or elsewhere (they do come up for like $100 occasionally) and see if all your problems really do go away with your current graphics card removed and another graphics in it’s place.

Because if that’s not the case, you’d have a power/mainboard issue and not totally a graphics one.

Assuming you’re certain the issue is totally on the graphics card, then your only recourse is to actually carefully take the graphics board sandwich apart. I’ve never done this personally so the only help I can give would be this showing you the inside of the board sandwich: http://www.sgistuff.net/hardware/graphics/impact.html

So once you’ve carefully exposed the ribbon cables, I’d recommend trying to perform a test with a multimeter where you place it in continuity mode (beeps when both probes touch) and carefully try to touch each end of the seated ribbon cable’s connector legs (one board to another board, same position of leg, counting from a single side) and check for continuity for each connected line (while seated) between the two boards for both ribbons.

It may be tricky to find your entry point. If you want to just test the ribbon you can carefully remove them (one side is flip-up door, the other is ejection by using the two handles around the connector). And you could try to test the ribbon itself, but this won’t prove the connectors are also okay and you’d not want to try to flatten the cables as they may fracture (so keep them bent as they are).

That might take an hour or so. But those cables are likely points of failure. After you’ve ruled those out. The only other two issues:

A local short (which I doubt, but you could check for a short using diode mode and check for chip data sheets to find vcc and ground and test between them. Or it’s more likely to be a broken/cold soldering joint. This is the more likely scenario if your ribbons with their connectors test okay. Then things get hard. You’d have to find someone (not me) with a professional reflow oven and then disassemble the board and remove the ribbons and other plastics and do a controlled reflow in a reflow oven as a last resort.

That’s very risky, I could work, it could fail, it could do nothing. At this point it would be a Hail Mary attempt.

Also if you have removable TRAM modules, you’d try to remove them and do a board test before the reflow oven.

But given the error, I don’t think it’s TRAM related.


I’ve only seen like 4 postings about graphics boards going bad in Indigo2, it’s very rare, but you’re not the first. I’ve only ever heard of one successful repair and that was the bent pin posting I mentioned earlier.

Normally the rest of these don’t yield fruit. These cards have so many components (dense) that the odds are against you when you consider the cost of the used replacement graphics card.

[sotlampr]

Thanks so much for all that info. I’m going to keep posting updates.

[mc68k]

Any update? I also have problem with 3 Impact gfx cards.