Indigo2 IMPACT TRAM repair

SGI hardware problems, solutions, tips, hacks, etc.
Forum rules
Any posts concerning pirated software or offering to buy/sell/trade commercial software are subject to removal.
User avatar
jan-jaap
Donor
Donor
Posts: 4939
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Indigo2 IMPACT TRAM repair

Unread postby jan-jaap » Wed Aug 27, 2014 2:39 am

I have an Indigo2 MaxIMPACT with the TRAM option card(s). Unfortunately it fails the diagnostics test, but the MaxIMPACT board set sans TRAM option passes.

First of all: the IRIX 6.5 'ide' test for an IMPACT Indigo2 runs for several hours before it reports 'pass' or 'fail'. If you CTRL-C the diagnostics after they print a 'hinv' like output and start running, you get an 'ide>>' prompt. Type 'help' to get available commands. There's at least two dozen 'GR2' tests (Extreme graphics and co) but nothing related to IMPACT. An inspection of the 'ide' binary showed that the IMPACT tests are 'mg_*' (for MardiGrass), and the full IMPACT test is a script called impact. This cuts down the test time to about 15 minutes. :)

Good thing I have a spare HighIMPACT as well, so I could test the TRAM cards individually. So now I know which one of them is faulty.

I took the heat sink off, and found that on one TRAM chip several pins had lifted from the PCB. To be honest I may have inflicted that damage myself when removing the heat sink, because the heat sink is stuck to the chips with double sided metalized tape which sticks pretty well. I'll fix that but have to find a replacement for the thermal tape because I cannot reuse it.

I also found this. Unfortunately there's no camera on the microscope so I was forced to use my Nikon:
TRAM.jpg

Two pins next to the left 'gap' are soldered together. First I thought this was my problem, but it's like this on all three TRAM chips. It must be intentional.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. (IRIS 4D/70 Superworkstation Technical Report)

User avatar
mila
Posts: 310
Joined: Thu Dec 31, 2009 4:20 am
Location: Bålsta, Sweden

Re: Indigo2 IMPACT TRAM repair

Unread postby mila » Wed Aug 27, 2014 12:39 pm

jan-jaap wrote:I have an Indigo2 MaxIMPACT with the TRAM option card(s). Unfortunately it fails the diagnostics test, but the MaxIMPACT board set sans TRAM option passes.

First of all: the IRIX 6.5 'ide' test for an IMPACT Indigo2 runs for several hours before it reports 'pass' or 'fail'. If you CTRL-C the diagnostics after they print a 'hinv' like output and start running, you get an 'ide>>' prompt. Type 'help' to get available commands. There's at least two dozen 'GR2' tests (Extreme graphics and co) but nothing related to IMPACT. An inspection of the 'ide' binary showed that the IMPACT tests are 'mg_*' (for MardiGrass), and the full IMPACT test is a script called impact. This cuts down the test time to about 15 minutes. :)

Good thing I have a spare HighIMPACT as well, so I could test the TRAM cards individually. So now I know which one of them is faulty.

I took the heat sink off, and found that on one TRAM chip several pins had lifted from the PCB. To be honest I may have inflicted that damage myself when removing the heat sink, because the heat sink is stuck to the chips with double sided metalized tape which sticks pretty well. I'll fix that but have to find a replacement for the thermal tape because I cannot reuse it.

I also found this. Unfortunately there's no camera on the microscope so I was forced to use my Nikon:
TRAM.jpg

Two pins next to the left 'gap' are soldered together. First I thought this was my problem, but it's like this on all three TRAM chips. It must be intentional.

It is very odd to have missing pins on a IC that I can tell since I am a HW designer :) But it could be pins broken that was not connected inside the chip.
--
No Microsoft product was used in any way to write or send this text.
If you use a Microsoft product to read it, you're doing so at your own
risk.

robespierre
Posts: 1578
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: Indigo2 IMPACT TRAM repair

Unread postby robespierre » Wed Aug 27, 2014 1:27 pm

If I have time I'll look at the TRAMs on my Max Impact, which worked last time I put it away. The solder bridge and missing pins are indeed very unusual, but I can imagine possible reasons for them.
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
pentium
Posts: 4747
Joined: Mon Aug 28, 2006 6:29 pm
Location: Kamloops, BC

Re: Indigo2 IMPACT TRAM repair

Unread postby pentium » Wed Aug 27, 2014 1:50 pm

That is so bizarre. All that effort to trim the individual legs when you just had to delete the traces for the pads.

It's almost like there's more to that ASIC than we think and SGI had to go out of their way to prevent people from snooping.
:Crimson: :Onyx: :O2000: :O200: :O200: :PI: :PI: :Indigo: :Indigo: :Indigo: :Octane: :O2: :1600SW: :Indigo2: :Indigo2: :Indigo2IMP: :Indigo2IMP: :Indy: :Indy: :Indy: :Cube:

Image <-------- A very happy forum member.

User avatar
jan-jaap
Donor
Donor
Posts: 4939
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Re: Indigo2 IMPACT TRAM repair

Unread postby jan-jaap » Wed Aug 27, 2014 2:26 pm

robespierre wrote:If I have time I'll look at the TRAMs on my Max Impact, which worked last time I put it away. The solder bridge and missing pins are indeed very unusual, but I can imagine possible reasons for them.

I wouldn't take the heatsink off a working TRAM -- the thermal tape really sticks to the TRAM memory chips and you risk pulling the chips off the PCB (or at least lifting some pins).

I tested the TRAMs of my spare MaxImpact boardset this evening (one at a time in the HighImpact). One of them worked, one failed. So I have another TRAM to inspect and two good ones to rebuild one MaxImpact. I'll report back with my findings. Oh, and the TRAM memory chips are definitely SGI specific.

I'll have to order some of those split pins to attach the TRAMs to the Impact boards, and ask one of our HW designers about a good replacement for that thermal tape. The original is a metalized tape (unlike what you find in e.g. an Octane CPU), and these things run *hot*. A replacement tape with a too little W/sqm rating would end up acting as an insulator, cooking the chips underneath. I'm not sure that e.g. arctic silver is a good idea, that kind of stuff turns liquid at high temps plus the chips may not be even enough for proper contact on all three of them.

This is not the first time I see oddball chips in SGI hardware. Recently my Octane EVO board committed harakiri. At first I thought the 'piggy' board (really, it's labeled that!) was to blame. There's a QFP FPGA on it (a Spartan, I think). Missing legs, and the thing wasn't flush on the PCB, it looked like the EVO had been wedged under something else and the chip had been 'crushed' on one side. Except that cannot really happen inside an Octane, plus not a single pin had lifted. I browsed eBay for a replacement, and found one -- which had *exactly* the same 'damage'. The seller had apparently noticed it too and taken a picture of it. In the end I found a spare EVO in my collection of spares (had forgotten I had it) which did *not* have this. Oh, and the piece that blew up was the XIO bridge chip. After I took the heat sink off, the smell off magic smoke was undeniable.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. (IRIS 4D/70 Superworkstation Technical Report)

User avatar
guardian452
Donor
Donor
Posts: 3464
Joined: Tue Aug 21, 2007 10:12 pm
Location: United States
Contact:

Re: Indigo2 IMPACT TRAM repair

Unread postby guardian452 » Wed Aug 27, 2014 6:48 pm

If you DO replace the conductive pad/tape with e.g. arctic silver or similar, you may have to grind down the mounting bosses of the heatsink (if it's anything like an Octane tRAM) so the thickness of the pad is accounted for.

User avatar
jan-jaap
Donor
Donor
Posts: 4939
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Re: Indigo2 IMPACT TRAM repair

Unread postby jan-jaap » Thu Aug 28, 2014 5:47 am

guardian452 wrote:If you DO replace the conductive pad/tape with e.g. arctic silver or similar, you may have to grind down the mounting bosses of the heatsink (if it's anything like an Octane tRAM) so the thickness of the pad is accounted for.

Most thermal tape I know is rubbery, and 1 - 2mm thick. This stuff is thin and metalized. And there are no bosses on the heat sink but silicone 'feet' on the PCB, the stuff you might expect under an appliance. I think I would get away with thermal grease, but I'll still look for thermal tape.

I also think I know what's causing the problems here: it's the tape used to attach the heat sink. On one of the long sides there's the connector so the kapton tape is on the short sides of the TRAM. Kapton tape doesn't stretch. As a result (?) the TRAM PCB is visibly warped when you look down the long side. If you look carefully at the legs of the TRAM chips you'll see that they look mostly normal at the middle chip, but at the outer chips they look a little 'crushed'. Once you take the heat sink of (and remove the down force on the chips), the legs lift.

For both faulty TRAM modules I've inspected so far pins have lifted in the lower right corner (red mark).

TRAM2.jpg
Sorry for the crappy photo :(


The story was always that TRAMs get so hot they desolder themselves. I think it's simply mechanical stress due to poor engineering. Which also explains why it's always the TRAM option cards that fail, but never (?) the first 1MB TRAM on the graphics board itself.

There's also thermal glue. This would remove the need for the kapton tape. Thing is, that stuff is permanent so the next fault would be fatal. Not sure I'd be willing to risk that.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. (IRIS 4D/70 Superworkstation Technical Report)

User avatar
Kumba
Posts: 235
Joined: Mon May 24, 2004 12:14 am
Location: Byzantine Secundus

Re: Indigo2 IMPACT TRAM repair

Unread postby Kumba » Fri Aug 29, 2014 7:14 pm

jan-jaap wrote:...This is not the first time I see oddball chips in SGI hardware.

Funny that you mention oddball chips. This one that I pulled out of a dead O2 has always made me chuckle. I think it's the Graphics Backend chip, but I've never been able to confirm it.
Attachments
sgi-evil-chip.png
sgi-evil-chip.png (176.24 KiB) Viewed 864 times
:Onyx2: 4x R14000 :Tezro: 4x R16000 :Fuel: 1x R16000 :Octane: 2x R14000 :O2+: RM7000 :O2: R10000 :O2: RM5200 :Indigo: R4400 :Indigo2IMP: R10000 :Indigo2: R8000 :O3x0: 4x R14000 :Indy: R5000

"The past tempts us, the present confuses us, the future frightens us. And our lives slip away, moment by moment, lost in that vast, terrible in-between."
--Emperor Turhan, Centauri Republic

User avatar
jan-jaap
Donor
Donor
Posts: 4939
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Re: Indigo2 IMPACT TRAM repair

Unread postby jan-jaap » Mon Sep 01, 2014 12:43 am

Last weekend the heatsink dropped off the base 1MB TRAM of my High IMPACT board set which I had been using to test TRAM option cards. So now I know where it hides :) For the record, it's the small heatsink close to the GIO64 connector on the top board.

The TRAM chip underneath has the same part# (099-9030-002) as the three on the option card. But it doesn't have missing legs, nor legs soldered together.

Had another look at both of my Max IMPACT board sets, and they *do* have the missing legs on the TRAM chip (1MB on the board set). Revision thing?

Anyway, I re-installed the heat sink on the High Impact using a thermal tape we use at work (colleague warned me I would pull the chip off the PCB before the tape would let go so I don't want to use it for the option cards :oops: ). High Impact still works fine, and for what it's worth, the TRAM gets much less hot than the large heat sink next to it. I think those stories about TRAMs de-soldering themselves are BS, but seeing the PCB of the TRAM option flex when you install it on the board set gives me a different vibe now that I know what's under neat that heat sink.
highimpact.jpg
This image borrowed from Gerhard Lenerz site

Octane TRAMs really are a much more sane construction. Ah, well...

NB: in this picture from Gerhard Lenerz's site you can see the TRAM chip is missing legs as well.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. (IRIS 4D/70 Superworkstation Technical Report)

robespierre
Posts: 1578
Joined: Mon Sep 12, 2011 2:28 pm
Location: Boston

Re: Indigo2 IMPACT TRAM repair

Unread postby robespierre » Mon Sep 01, 2014 3:24 pm

the best theory I could come up with is that the TRAMs are compensated for inherent speed differences by a delay line made of fusable elements. During testing the speed is measured and then the fuses are blown to compensate the chip. the pins that are used for this purpose would then be clipped off to minimize any parasitics from them.
:PI: :O2: :Indigo2IMP: :Indigo2IMP:

User avatar
vishnu
Donor
Donor
Posts: 3189
Joined: Sun Mar 18, 2007 3:25 pm
Location: Minneapolis, Minnesota USA

Re: Indigo2 IMPACT TRAM repair

Unread postby vishnu » Mon Sep 01, 2014 5:18 pm

robespierre wrote:the best theory I could come up with is that the TRAMs are compensated for inherent speed differences by a delay line made of fusable elements. During testing the speed is measured and then the fuses are blown to compensate the chip. the pins that are used for this purpose would then be clipped off to minimize any parasitics from them.

To paraphrase Niels Bohr; we are all agreed that your theory is crazy. The question that divides us is whether it is crazy enough to be correct. 8-)
Project:
Temporarily lost at sea...
Plan:
World domination! Or something...

:Tezro: :Octane2:

User avatar
hamei
Posts: 10435
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Indigo2 IMPACT TRAM repair

Unread postby hamei » Mon Sep 01, 2014 6:42 pm

vishnu wrote:To paraphrase Niels Bohr; we are all agreed that your theory is crazy. The question that divides us is whether it is crazy enough to be correct. 8-)

Since you had to drag Niels into this ...

Einstein : "God doesn't play dice with the Universe."
Niels B : "Albert, don't tell God what to do."

:P :P :P
I spent a fortune on booze, birds, and fast cars ... the rest I just squandered

User avatar
jan-jaap
Donor
Donor
Posts: 4939
Joined: Thu Jun 17, 2004 11:35 am
Location: Wijchen, The Netherlands
Contact:

Re: Indigo2 IMPACT TRAM repair

Unread postby jan-jaap » Sun Sep 07, 2014 1:27 pm

Victory! (well, one out of two, at least).

I performed the fix one one TRAM. Installed it, it flexed dangerously, failed the test and when I took it out it had a whole lot of new dangling legs on the chips. Tried again, must have resoldered 250 legs in the end but it never successfully ran the test. The second one needed maybe 50 legs redone, and worked the first time. :D The first one took me a *lot* of time, number two maybe 1.5 hours.

This soldering is a nerve wrecking business, btw. These chips are roughly 1x1" each, 52 legs per side. The pitch between these legs is 0.02" (ca 0.5mm). Of course this board is nearly 20 years old and the solder doesn't re-flow nicely *at all*. I ended up putting a little 'caterpillar' of (leaded) solder over the legs, then removing it again with litze. Then manually checking and fixing one leg at a time, under the microscope, using the thinnest solder and tip I could find in the lab.

Here's a corner of a TRAM chip after rework:
tramfix1.jpg


And here's a closeup:
tramfix2.jpg


In this picture I reworked maybe 5 or so legs on the left hand side of the corner and around 10 on the other side.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. (IRIS 4D/70 Superworkstation Technical Report)

User avatar
vishnu
Donor
Donor
Posts: 3189
Joined: Sun Mar 18, 2007 3:25 pm
Location: Minneapolis, Minnesota USA

Re: Indigo2 IMPACT TRAM repair

Unread postby vishnu » Sun Sep 07, 2014 6:26 pm

Wow, nice (painstaking) work! 8-)

Were ya working without a net? ;)
Project:
Temporarily lost at sea...
Plan:
World domination! Or something...

:Tezro: :Octane2:

User avatar
hamei
Posts: 10435
Joined: Tue Feb 24, 2004 4:10 pm
Location: over the rainbow

Re: Indigo2 IMPACT TRAM repair

Unread postby hamei » Sun Sep 07, 2014 7:42 pm

jan-jaap wrote:This soldering is a nerve wrecking business, btw.

Damn, j-j. You get the Ship-Inna-Bottle Award for this one. With the captain on the quarterdeck and tiny crew in the rigging :D

Really nice work ...
I spent a fortune on booze, birds, and fast cars ... the rest I just squandered


Return to “SGI: Hardware”

Who is online

Users browsing this forum: No registered users and 3 guests