Lucking out with a "defective" T22 (Images!)

[Robbyrobot]

Bought a T22 base with mainboard and 900MHz CPU this week for parts. It was described as defective, with a graphics problem.

I received it today and immediately hooked it up with a port replicator, an external keyboard and an external monitor. And sure enough, the display was corrupted: yellow-green and purple lines running vertically around 3-4 mm apart pver the entire screen. No amount of tapping, twisting of components or pressing on chips changed this. OK, that's it, I thought, and started to dismantle everything. Took off the fan, the CPU, the Ultrabay cage, the loudspeakers, the HDD cage and the PCMCIA housing. Then I removed the mainboard from the base - and saw that the AGP chip and the thermal rubber underneath it were wet with some oily-looking material. No idea what it was, and it didn't look like a spill, as the wetness was very local and restricted to the chip and the rubber underneath.

At any rate, I decided to clean things up, got a syringe and isopropyl alcohol and ran around 15 mL of isopropyl alcohol through the BGA underneath the chip and the graphics chip next to it, passing the solvent as much as possible through the BGA and not just over the chips.

Then I used a hot air gun to warm up the area and get rid of the isopropyl alcohol. After around half an hour of sitting, I put the board back into the base, added a CPU and fan, and reattached the base to the port replicator.

Turned it on, and the vertical lines were still there. This time a pushed a little against the lower part of where the PCMCIA cage is mounted, and suddenly the lines disappeared and the display was perfect. I played with the board some more, but the display stayed good. I started the assembly several times, and the display stayed good. Now I'm running PC-Doctor and it seems the display is OK.

Now, I'm not sure what exactly was the problem - some conducting liquid in the BGA, maybe some poor connections. At any rate, the problem this board had is - for the moment - apparently solved. Now I'll just leave the setup for several days, starting it occasionally to see if the display is still OK. If so, I guess I have a functional T22 base and can comlete it with parts I have on hand here.

Nice experience... everyone likes success, and something like this makes a weekend more pleasant.

[jamerslong]

Good job robby, I bet the lines were just a loose connection that you somehow put back into place. You learn something new everyday!

Now hopefully other people can fix their video problems with the same fix.

[joester]

Excellent!

My only question would be with the oily substance...but I think I know the answer...

Unless I'm completely off my rocker, the rubber pad is actually a closed cell foam made of rubber. Contained in the cells of the foam is silicone, so you basically have a hermetically sealed sponge soaked with silicon. The silicon acts as a thermal transfer medium and increases the ability of the pad to transfer heat away from the chip it has contact with.
Just like thermal grease (silicon and zinc oxide), the foam will break down with time. When it does, the silicon will escape and spread.

Then someone like you has to clean it up.

I would consider applying a very sparing amount of thermal grease (Arctic 7 or equivalent) to the surface of the pad if you intend to re-use the pad. Replacing it would be best, but shouldn't be neccessary. Keep in mind that it's ability to keep the chip cool is decreased with the loss of the silicon. Applying the grease as a topical agent won't restore the ability, but can only help. MORE IS NOT BETTER. You won't get the pad off in my experience without destroying it.
Also, if you remove the pad with the intent to use grease only, you need to verify that you have physical contact between the chip and it's heatsink after assembly. This is easily done by removing the pad, applying a small dot of grease in the middle of the chip and re-assembling the unit. Then take it apart again and inspect the dot. If the dot has been spread out over the top of the chip, you have good contact. If it hasn't really changed much, you don't. Unfortunately, this method would require a new pad if contact is not made. Keep that in mind too.

Joe

[phool@round]

At work about every 6 months everyone from the surrounding area brings their old equipment in to our office. HP recycles alot of old stuff. Anyways, I picked up an NC1000 and an Omnibook 900 and took them home. Both would not post, took the keyboards off and found they had been victims of spills, juice and soda. Nice and sticky. Took both boards out, liberally applied Windex (all I had at the moment) and let it soak for a couple of minutes then rinsed under hot water (wild man aint I?) and placed them over the heat register to dry. I figured if it worked that's cool, if not, back to the recycle pile. Reinstalled both back in their cases and both posted.

Very interesting about the silicone in the thermal pads. I just bought a 570E and took the keyboard off for a preliminary inspection and when I removed the heat shield both thermal pads (over the northbridge and NeoMagic GPU) I noticed they where "wet". I wiped them off but now I think a little more attention is in order. Thanks Joe and Robby for your insights.

[Robbyrobot]

Interesting, interesting... I started up the T22 this morning again, and the blue (not purple) lines were back in the splash screen (with the IBM logo). A rectangular space around the logo, however, had no lines and a plain black background as should be. Then the "no hard drive installed" message came on, and the lines turned to green. When I entered the BIOS, the green lines were only outside the BIOS page. The white page itself had no lines. However, when I moved the cursor, you could see faint red lines shimmering on the white background for a second, and then they disappeared.

When I booted memtest86+, the lines disappeared and the screen looked normal.

I ran memtest86+ for a few minutes, rebooted, and made the same observations again. Then I let memtest86+ run for maybe 10-15 minutes are rebooted. This time no lines, both the splash screen and the screen for the "hard drive missing" message had a plain black background, as they should be. Now I've booted PC-Doctor again, no lines and the screen is perfectly normal.

This is beginning to look like a heat-related situation, but what intrigues me is the fact that the background of the logo in the splash screen and the BIOS page were not affected even when the computer was cold. I'm not really sure what to make of this.

[RealBlackStuff]

This time a pushed a little against the lower part of where the PCMCIA cage is mounted, and suddenly the lines disappeared and the display was perfect

Maybe a bit more of the same? Check for loose connectors, inductors (resolder them), whatever in that area.

[Robbyrobot]

Couldn't find anything loose, and the strange stuff continues. This time I let the computer sit without power for 2-3 hours and then started up again. Got the "usual" striped splash screen:

But then, when the screen went to the "hard drive not found" message, the lines were gone. And stayed gone when the BIOS was called up or a program booted from diskette.
The interesting thing is that after the lines disappear, they don't reappear even when I manipulate the board and components quite vigorously.

[RealBlackStuff]

Check in the BIOS/Config/Display/HV expansion.
Set that to ON, and the blue lines probably disappear.

[Robbyrobot]

The HV Expansion was already on, and I really think this is a heat issue. Booted again and got the striped splash screen above (better picture this time since I didn't have to angle to avoid reflections). This went on to the "hard drive not found" message:

And then booted (memtest again) to a perfectly normal screen, no stripes.

After a minute or so, I rebooted, got the striped splash screen but this time the "hard drive not found" screen was normal, no stripes. And naturally everything afterwards.

After running the computer for 5-10 minutes, a reboot shows a normal splash screen, no stripes.

I again tried to elicit some response by twisting and pressing various places, mainly the mainboard under the PCMCIA cage (graphics chips on the underside), but there was no visible effect.

Possibly some component is borderline and only works properly when the computer is warm. At the moment, I'm out of ideas, and I don't really think there's going to be a fix, but since the problem disappears after a few minutes anyway, it is bearable.

Certainly an interesting situation, though.

[RealBlackStuff]

That second picture shows you need a new CMOS battery. (Date/time)
If there are no other problems, you are indeed lucky.
Were both screen shots from an external monitor?

[Robbyrobot]

Actually, there was no CMOS battery attached at the time the photos were made. Have to get one for this machine, as I wasn't counting on being able to repair it.

And yes, both screen shots were of an external monitor, a very old and decrepit CRT monitor I just have left over from the "good old days". That's why the original image of the splash screen was so angled: I took it around noon, and the sun was shining on the monitor so I had to angle the shot to avoid reflections. The second picture was shot in the evening, and I could take it directly facing the monitor.

Edit: By the way, another interesting point in all this is that I have never seen any stripes or otherwise garbled graphics except in the splash screen, the "hard drive not found" screen and outside the white field of the BIOS screens. As soon as the computer boots (even right after being "cold"), the screen is absolutely normal. That doesn't really sound like a hardware defect, does it?