A22P Is this what a bad graphics chip does?

[Neil]

I recently bought this A22P for parts, at a price that will allow me to more than recover my cost if I sell off some of the parts. It was advertised as non-working, but I'm using it to write this post right now. It works, just not very well. Certainly needs something, I'm just not sure what. The seller described it, "seems to work fine, until it heats up and the the graphics start to fail". And that may be accurate, but I wanted to share my findings, and see if any of you Thinkpad experts have any advise.

At first I booted up using a Linux live CD. It worked OK, but the graphics were somewhat corrupt right from the start. Not a heat issue, must have been something else. Tried three different distros that I had on hand, and all did similarly. Except for a while, DSL using fluxbox desktop seemed to be working perfectly with no corruption, I just didn't have the full 1600X1200 resolution as an option using the VESA driver. I never tried to install ATI drivers under Linux.

So, installed Windows 2000 Pro. Works fine. Downloaded all the drivers from the Lenovo support site, installed 'em. Still works fine. Was running for about a hour, using the Windows default graphics driver (not the Thinkpad ATI driver), but could only get 640X480 resolution. Installed the ATI driver, rebooted, and the display shuts off when the ATI driver starts to load. That is, the Thinkpad splash displays fine, the Windows splash displays fine, but when ATI starts to load, the display (LCD and external monitor, when connected) shuts off. Doesn't just go blank, but shuts off. Reboot in safe mode, works fine. Disable ATI driver in safe mode, reboot, still works fine.

Now I have Winows XP installed. XP has 1600X1200 resolution available with the default display adapter driver, so that is what I'm using now as I write this. But, when I installed the XP ATI driver. same thing happens as with 2000. Monitor shuts off as soon as the driver starts to load on boot, right after the Windows XP splash. And the computer doesn't continue to boot, it requires a cold restart to get it going again. Earlier today, running the default adapter driver, I left the laptop running with a web browser open and a clock on the desktop, along with MobileMeter, and the temp stayed between 40C and 50C, running on A/C power. But after about an hour and a half there was a temp spike to about 65C and it locked up. No BSOD, just a frozen desktop like I had left it, only the clock had the second hand displayed in about a dozen different locations, and was not moving now.

I'm watching MobileMeter as I'm writing this, and it has been running steady at around 41C for almost an hour now, on battery power (looks like the battery that I got with it will last 1:30 or more). So doesn't really seems like a heat problem. But there must be something wrong with the graphics chip, don't you think. What else can I look for?

[rkawakami]

I'd have to say that your problem is more likely with an intermittent contact of one (or more) of the connections between the graphics chip and the motherboard. It could seem to be a thermal issue in the fact that as the system heats up, it expands very slightly. That might be enough to cause the contact to go bad. From your description of various screen modes not appearing to be affected, you could have a bad connection to the video memory. As you go "higher" in resolution, the graphics chip needs to address more memory for the display. If you can, try loading just an old DOS-only program (memtest86+, PC Doctor, hard drive diagnostic, etc.) and see if you encounter any video issues.

If it indeed is a loose connection, there's not much you can do besides trying to re-flow the solder for a "permanent" fix. As outlined in this T42 thread:

http://forum.thinkpads.com/viewtopic.php?t=33952

the suspected cause of a broken connection has been "repaired" using pressure on top of the ATI chip (wedging something non-conductive between the top of the chip and the keyboard), to actually getting a heat gun and doing a home re-flow.

[Neil]

I've been using the A22P tonight for about three hours at 1600X1200 32bit, with the Default Monitor and Default Display Adapter drivers, with no graphics problems. I did have one system lock-up, right in the middle of installing a Thinkpad utility. But, a cold reboot got me going again. Doesn't feel like a loose connector to me...(but what do I know?). How could the ATI graphics driver activate a cable fault, and the default driver not do so?

Haven't tried to go back to the ATI display driver yet. I could actually use this machine, as-is, for my normal daily routine, of checking email, reading news and weather, etc. while I have breakfast in the mornings.

[ajkula66]

Neil wrote:

I could actually use this machine, as-is, for my normal daily routine, of checking email, reading news and weather, etc. while I have breakfast in the mornings

.

You could, the question being for how long...if the GPU is coming loose as Ray suggested, and I would tend to agree, the lockups are going to start at one point whether you load the ATi driver or not, machine will become unreliable and work intemittantly. At that point you'll have to part it out, buy a new motherboard ot try to re-flow the solder.

I've had a T40 which had the exact same issue that you're describing...once the ATi driver was downloaded the machine started going haywire. After a while it ended up as a parts unit.

If you like your new ThinkPad enough to consider keeping it for a while, you may want to stick a post-it or two on the top of your GPU...

Good luck.

[rkawakami]

How could the ATI graphics driver activate a cable fault, and the default driver not do so?

I was not talking about a cable per se being a problem when I wrote earlier about a "loose connection". The connection(s) I'm referring to are the soldered joints between the (assumed) surface mounted graphics chip and the motherboard and/or the same type of joints which affix any separate video memory chips to the surface that it is supposed to be mounted on.

Can't claim to be too knowledgeable about the differences in video drivers (default [Windows, I assume] vs. ATI) but at 1600x1200x32 bit, my calculations say that that requires just under 8MB of memory. The ATI Rage Mobility 128 in the A22p is spec'ed with 16MB of memory. It may be possible that the default driver is not using any of the upper video memory locations, whereas the ATI drivers do. That's just speculation on my part. If you can download a copy of the PC Doctor for DOS for your A22p from here:

http://www-307.ibm.com/pc/support/site. ... IGR-4JRMNE

and run the Video tests, that might help to localize the problem area. The diagnostics are supposed to check out the entire video subsystem and information about which tests are passing and failing could be helpful.

I have two A31 systems that exhibit display problems which are minor compared to what you are experiencing. The display shows random blocks and lines but not all of the time. Flexing the chassis makes the problem worse. It's been so long since I've played around with them that I can't remember if the they ever locked up.

[Neil]

Thanks, rkawakami, for the reminder about PC Doctor. Ran it, and no surpises really. Video Memory, Video Pages, VGA Controller Registers, VGA Color-DAC Registers and RageMobility MM tests all Passed. ATI RageMobility 3D test, Failed. And ATI RageMobility 2D and MM tests just would not run, the LCD would shut down just like when switching to the ATI driver.

Next step will be to open her up and see what the GPU connections look like, and maybe put a post-it or two on it (thanks ajkula66 for that tip). Not sure when I'll find the time to do that, however.

[Neil]

you may want to stick a post-it or two on the top of your GPU...

Good luck.

OK, guys, I have started disassembly following the HMM directions, and have the keyboard and bezel off. Before I dig any deeper I need to know where, exactly is the GPU located on the mainboard? I don't think I can see it yet, but I don't want to pull any more parts than necessary to get to it and try a post-it on top of it.

Thanks for the help.

[sjthinkpader]

You can see most everything by removing the FDD and the drive cages. The FDD FPC connector is a little tough to release without removing the CPU heatsink but it can be done.

[Neil]

Well this is just great. I have removed the FDD, PCMIA slots and the drive cages, and you were right sjthinkpader, I can see most everything on the mainboard now. But, silly me, I thought the GPU would be marked with an ATI part number or something that would identify it. I can't tell what is what still.

I can identify the CPU (easy), the PCIbus chip (it's marked), there is a chip marked Crystal (BIOS?), a PC97338VJG (could be sound),a Samsung 103 (don't know), and various other ICs with part numbers that aren't found by a google search.

So without knowing exactly where to look for the GPU, I'm still at a loss. Maybe I'll just put sticky notes on top of everything!

[sjthinkpader]

Sorry, the GPU is actually on the underside of the board just to the left of the CPU and FDD ZIF connector. You can see the thru holes. Below that slightly larger is the Northbridge.

[Neil]

Oh! See, that's what I wanted to know going in. Could have saved a lot of time looking around had I known what I was looking for. Thanks for the help, and apology accepted.