VOGONS


First post, by Xenphor

User metadata
Rank Member
Rank
Member

edit: Right after making this post I tried it again and now it boots normally? Wtf? Is it dying?

Here's a video of the problem.

I just received a 9200 from ebay that looks to be in excellent shape. I put it in my system and it worked straight away. I ran 3DMark 2001 1 time so that I could warm up the heatsink to replace the paste and it ran fine, so I powered down the system and took out the card. I had already removed the pushpins, so it was only a matter of wiggling the heatsink free and removing the dried up paste. I couldn't completely remove the dried paste around the very outer edges of the GPU where it dips down into the PCB because I didn't want to risk damaging it. This seemed okay since the top was wiped completely with Arctic Silver Arcticlean kit. I put on some Noctua paste, put the heatsink back on and everything looked fine.

I put the 9200 in my Win98 pc and immediately I got no signal from the OSSC. I tried many times but nothing happened. Finally I took the card out, cleaned the contacts and made sure it was firmly in place. After doing that I get a signal but it is full of garbage although I can at least make out the boot up sequence.

I really don't understand what could've happened?

1GDkX4Ul.jpegAp50wN1l.jpeg
lz41nWxl.jpeg

Reply 1 of 8, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

He's dead, Jim.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 2 of 8, by akimmet

User metadata
Rank Member
Rank
Member

The most likely scenario is that you damaged some BGA solder joints when you twisted off the heat sink.
Intermittent failures with BGA mounted chips are somewhat common on electronics from this time-frame. Lead free solder is much more brittle than old lead based solder. Modern flux and solder formulations are still improving, but still not as durable as the old lead based formulations.

You were probably much better off leaving well enough alone. Why change the thermal paste it if there was nothing wrong in the first place.
I have worked on power amplifiers from the 1970's that have never been re-pasted. In my experience 50+ year old thermal paste is still effective. Unless I need to disturb the heat sink I leave well enough alone.
On the other hand, I always replace thermal pads. Those do tend to go bad from age.

Reply 3 of 8, by Xenphor

User metadata
Rank Member
Rank
Member
akimmet wrote on 2024-07-07, 17:15:
The most likely scenario is that you damaged some BGA solder joints when you twisted off the heat sink. Intermittent failures w […]
Show full quote

The most likely scenario is that you damaged some BGA solder joints when you twisted off the heat sink.
Intermittent failures with BGA mounted chips are somewhat common on electronics from this time-frame. Lead free solder is much more brittle than old lead based solder. Modern flux and solder formulations are still improving, but still not as durable as the old lead based formulations.

You were probably much better off leaving well enough alone. Why change the thermal paste it if there was nothing wrong in the first place.
I have worked on power amplifiers from the 1970's that have never been re-pasted. In my experience 50+ year old thermal paste is still effective. Unless I need to disturb the heat sink I leave well enough alone.
On the other hand, I always replace thermal pads. Those do tend to go bad from age.

I wanted to replace the paste since I've had other cards where there was virtually no paste left after I removed the heatsink. In particular, I had a Geforce Quadro 4 XGL with a similar heatsink only design that started showing artifacts after I used it for awhile without touching it. After I removed the heatsink, I noticed there was pretty much no paste left. The other issue is that I'm not aware of a way to get an accurate temperature reading from older cards, so I have no way of knowing if it is still operating at a safe temperature when leaving it stock. If I knew that it was still getting enough cooling I wouldn't mess with it.

I did apply some amount of force on the heatsink before running the benchmark to see if I could get it off but it wouldn't budge at all. After I ran the benchmark it came off fairly easily so I don't know if that caused any damage. I'm sure it's possible I damaged it.

Reply 4 of 8, by Repo Man11

User metadata
Rank l33t
Rank
l33t

I routinely remove the heat sink on "New to me" video cards - if it seems to be stuck, putting it in an anti static bag and letting it sit in the freezer overnight has always worked well for me. The difference in thermal expansion/contraction makes removing stuck on heat sinks quite easy (at least so far).

"We do these things not because they are easy, but because we thought they would be easy."

Reply 5 of 8, by akimmet

User metadata
Rank Member
Rank
Member

BGA mounted chips can be very sensitive to mechanical stress, this includes thermal expansion. Even the act of benchmarking the card with the heat sink hold downs removed could have potentially cracked some joints.

Another potential issue is that modern CPU thermal pastes are electrically conductive, applying too much can cause electrical shorts.

While the freezer method dose work, be aware of condensation when doing this. You just need to make sure everything is totally dry.
Since I have a hot air rework station, I just SLOWLY heat things up to 100ºC. That usually liquefies any stuck paste. Heat will also soften thermal adhesive if that happened to be used instead.

When I come across a heat sink that is stuck, I consider that as a sign the paste is in good shape and should be left alone if possible.

Reply 6 of 8, by Rwolf

User metadata
Rank Member
Rank
Member

I have an old laptop with an ATI 9200 based mezzanine VGA card, a number of those were failing from thermal stress.
The temperature on the card was hot when active, but cool when the laptop went to sleep, so the BGA chips solder eventually cracked.
You could test if it was a solder issue by pressing on the VGA chip by inserting a small rubber shim on top to push it down (between chip heat sink and keyboard) which usually got the image back, but it was not stable of course. Re-soldering got it working for awhile, but the issue came back soon enough.

Reply 7 of 8, by auron

User metadata
Rank Oldbie
Rank
Oldbie

i'm also in the camp of leaving it alone if it works, because of possible issues like this. if you touch the heatsink during operation and it gets warm, thermal transfer is clearly still working. though i do have a rage 128 pro that during testing did not get warm at all...

i wonder though, did anybody try the dental floss method? in theory, this seems the way to remove a heatsink while placing the least possible stress on the BGA.

Reply 8 of 8, by Masterchief79

User metadata
Rank Newbie
Rank
Newbie

I like to properly heat them up before, to about 100-125°C, so you can twist off the heatsink with barely any resistance.
For removing the old brittle chewing gum, I use melamine sponges in combination with glue remover. Haven't killed anything yet with this method.