ThinkPad T22 Flaky RAM Issues

[hax0rwax0r]

First, sorry for the lengthy write-up as I wanted to detail with as much as I could in case any of this seems indicative of an obvious issue aside from just "you got bad RAM stick(s)." The TL;DR of it all is that original 2 x 128 MB RAM passes Memtest86+ but any variation of 256 MB modules eventually fails.

I bought an IBM ThinkPad T22 from Shop Goodwill a couple of weeks ago and aside from the busted screen bezel that happened during shipping (a real shame), it's in great shape. I have since bought and replaced the CMOS battery so it would hold time and BIOS settings. I also bought a dock (I wanted the ThinkPad Dock II but they are nowhere to be found these days) as well as a MSATA to IDE drive converter and upgraded the SSD to a 32GB MSATA SSD. Then I bought RAM to upgrade from 256 MB to 512 MB because that was next on the list. That's where things got frustrating...

Unfortunately, the RAM upgrades have given me nothing but fits and so I am here to ask for advice.

The system originally arrived to me with 2 x 128 MB sticks for a total of 256 MB of RAM. The RAM is in the other room as I write this but I am pretty certain the RAM is technically different labels and possibly models (point being not exactly 1:1 matched) but both sticks do have an IBM label on them so maybe they are the same RAM with different labels. That aside, my issue is not with those 2 x 128 MB RAM. I can Memtest86+ those for several hours and multiple passes, seemingly without issue.

First, I bought a 2 x 256 MB PC100 low density (512 MB total) upgrade kit on eBay and PXE network booted up the latest Memtest86+ v6.20 on it to start testing. Within a few minutes the test started throwing errors. I powered off, swapped RAM in the slots (ensuring they were reseated fully and securely) and started to retest again. Same result. I individually tested the RAM sticks and one seems to test out fine while the other does not. Cool, at least that makes sense - I got a single bad stick of RAM and the eBay seller is replacing the faulty stick (arrives Monday).

Second, I bought a second ThinkPad T20 that was listed as "for parts" on eBay (I wanted it for a replacement screen bezel that busted on my T22 during shipping due to improper packing) and it just so happened to have 2 x 256 MB of RAM in it. The sticks were mismatched brands but both were 256 MB each and both were PC100. I popped those in and began my PXE booted Memtest86+ testing again. It ran for over an hour and got a clean pass but then errored out in a second or third pass.

Third, I theorized that maybe the second RAM slot was causing issue (yes, in spite of both original 128 MB modules working with both populated) so I tested one of the sticks from the T20 in slot 1 only for several passes then tested one of the sticks from the T20 in slot 2 only for several passes. I thought I had a confirmed good RAM module from the T20 so I did the same procedure with the other unmatched stick of RAM from the donor T20, first testing slot 1 only then slot 2 only for several passes. Both passed the individual tests (either in slot 1 OR slot 2) without issue. So it seems that individually, my RAM from the donor T20 was at least a known good entity. I put both in together, PXE booted up the Memtest86+ v6.20 and started testing. The test seemed to increment the clock/timer showing how long the test had been running for very erratically, almost as if it was going unresponsive at times and not doing anything. Then it just hard hung and did nothing for a while, no fields updating on screen, no additional tests, no errors. Just a seemingly frozen screen.

So starts my confusion and frustration...

Fourth, I found an OLD version of Memtest86 (non-plus) v2.7 (from July 12, 2001), downloaded the disk image, dd wrote it with my MacBook to the external 3.5" floppy drive and booted that up on the T22. I started to test the 2 x 256 MB modules from the T20 and it ran fine without issue for several passes and no errors.

Fifth, I swapped in the 2 x 256 MB eBay RAM (the one with a seemingly 100% faulty module) and sure enough, it failed with the old Memtest86 v2.7. I swapped the eBay sticks out to be individually tested. Same as the new Memtest86+, the old Memtest86 almost immediately showed the one stick as bad. OK, consistency at least - a known bad stick of RAM. I set that bad stick off to the side to ensure it's not in the testing mix going forward.

I then had the theory that the old BIOS v1.0 I was running might not be the best option for RAM compatibility. I didn't see any change notes from v1.0 to v1.12 (latest) on my T22 BIOS revision history that would suggest that an upgrade would fix anything but whatever... I downloaded the disk image, wrote it to a floppy, swapped the original 2 x 128 MB RAM sticks back in, booted the BIOS flashing floppy and upgraded the BIOS from v1.0 to v1.12.

This brings me to my next point worth mentioning - I think it's worth noting that during the RAM tests, I noticed both Memtest86 and Memtest86+ up to this point often (always?) showed the CPU speed as 665 MHz (it was 65x or 66x MHz but I do not recall with 100% certainty that it was 665 MHz) when the CPU itself was a 1.0 GHz CPU. It was displaying basically 2/3 the CPU rated speed. This seemed to be resolved after upgrading the BIOS to v1.12 - I am not sure if because BIOS fixed something or the upgrade cleared/changed some CPU setting but the CPU was definitely running at 1 GHz in Windows 2000 Professional prior to RAM upgrades and/or Memtest86 testing. I don't believe any CPU settings changed as SpeedStep was always enabled and the CPU was set to run max performance when plugged in - same as pre and post-BIOS upgrade. Edit: This might have been due to the versions of Memtest86/Memtest86+ I tried. I tried two versions of Memtest86 and one of Memtest86+ so I am removing this as factual to avoid inaccuracies.

Post-BIOS upgrade, all RAM tests from here on out, the Memtest86 (old) and Memtest86+ (current) now shows the proper CPU speed (I think 998 Mhz or something). Edit: Same as above, removing since this might be inaccurate in hindsight. I can confirm the last three Memtest86+ tests I have ran right now show 995 MHz.

From here on out, any combination of RAM in either slot position (slot 1 or slot 2) have I found a stable combination that totals 512 MB RAM. The one "good" stick of eBay RAM with one or the other stick of T20 donor RAM or both sticks of donor RAM fail no matter what. It might only be one test (test 8 I think?) that fails on the third or fourth iteration/pass of Memtest86+ but it always fails.

I also seem to have random single stick test failures on each of the three 256 MB modules on the new Memtest86+ but the old Memtest86 v2.7 from 2001 passes them just fine. The single stick module tests don't seem to follow a particular RAM slot that I have noticed but I cannot guarantee that.

For what it's worth, not that the eBay seller would admit to selling shoddy RAM, but they said they haven't had anyone complain who bought "over 200 sticks" of this RAM before. Again, I am sure a seller isn't going to openly admit their product is sketchy, but their reseller rating is quite good. The exact RAM I bought was this: https://www.ebay.com/itm/125856665866 (512MB 2 X 256MB 144 Pin PC100 3.3v SDRAM Laptop Notebook Low Density SODIMM RAM).

So, do I REALLY have four bad sticks of 256 MB RAM modules (I know one eBay stick is 100% bad - the only thing really consistent in all this)? Two of the sticks from the T20 were name brand, albeit of different brands and now surely 20+ years old and the eBay RAM was new but not from a brand name I recognize.

I thought maybe there was some other setting or underlying issue (bad capacitors?) that could cause this behavior as it seems flaky at best. What gets me is the T20 RAM tests out individually good in either slot but paired together, it fails. But it's not a simple failure because both RAM slots are populated as the original 2 x 128 MB IBM labeled sticks test fine.

Maybe I will hook an external monitor to the T20 parts machine as I can't see much on screen without an LCD backlight and run some Memtest86+ tests in that unit to see what I find in the meantime.

Thanks.

Edit: The replacement eBay RAM stick arrived today since I initially posted and I will be doing tests on it soon and posting an update. I am also re-testing everything and am documenting this time more thoroughly since after reading the above, it's clear that I went through a lot of iterations of tests, versions, combinations, etc. and my write-up was from my memory which, like my system RAM, might be faulty, too. I want to be as accurate as possible on my testing results and doing a better job of definitively testing each stick, documenting which slot it was in and capturing a screenshot and notes of the final result.

[hax0rwax0r]

An update on my troubleshooting measures -

Replacement stick of RAM the eBay seller sent me also fails in either slot tested in.

I have noted that the tests from any of the now three sticks of new PC100 low density eBay RAM all outright fail in some fashion. So far I have observed a failure to mean with memory errors but the tests continue to run, the system hard locks with the timer freezing with no errors on the screen or it hard locks with an "unexpected interrupt on CPU 0."

I found reference on these forums for someone mentioning a part number for the T22 that was known to work: https://forum.thinkpads.com/viewtopic.p ... 23#p506823

This is the 256MB module listed for the IBM ThinkPad T22: https://web.archive.org/web/20090609122 ... 94A5CA7304

The specs from that page are as follows: SDRAM, PC133 • CL=2 • Non-parity • 133MHz • 3.3V • 32Meg x 64

You cannot buy this RAM from Crucial directly anymore. You can't buy it from ANYONE not eBay these days anymore.

That said, I found new Crucial CT32M64S4W7E-16FGY RAM on eBay matching this spec. The condition was new. If the inventory is fake, then they went through a lot of effort to print labels with models and serial numbers on them to seal the outer cardboard packaging as well as printing up a Crucial information insert on nice paper as well as sealed antistatic bags with a Crucial label on them.

I popped both sticks in and PXE booted up Memtest86+ v6.20:



FAIL.

I have SEVEN sticks of 256MB RAM. Two brand new Crucial sticks from eBay, three brand new generic sticks from eBay and two unmatched sticks from a "for parts only" used T20. Every single one of them fails in some way, shape or form as I've never had stable 256 MB x 2 configuration.

I have yet to test the Crucial modules individually in each of the two slots but to be honest, if it cannot run 256MB x 2 then I am no better off than the 128MB x 2 it came with which, again, runs 100% stable. And to confirm, again, the original 128MB x 2 modules are absolutely stable as I ran Memtest86+ v6.20 on them for just over 14 hours with zero errors.

I have absolutely no idea where to go next with this. Thoughts?

[RealBlackStuff]

From your tests above, maybe you could try another CPU?
But more likely your motherboard is going haywire.

In my stash I still have 6x 256MB PC-133 RAM.
I tested them all last year in my T23 (since sold) and they were all working fine.
Have to dig them out to check their brands.
If interested, you could have them all for $30 + shipping from Ireland.

[hax0rwax0r]

I tried to private message you to coordinate and determine what shipping charges might be and to see if you had any additional IBM ThinkPad T2x parts or components that you might be interested in selling. It appears that my account is too new to send private messages or post without moderator approval so I am not sure how to contact you unless an admin lifts those restrictions on my account.

I was considering that the flakiness could be related to faulty capacitors on the motherboard. I know the early 2000s were known for a slew of bad capacitors due to faulty/stolen electrolytic formula (https://en.wikipedia.org/wiki/Capacitor_plague) and my T22 is right from that time period. I don't know if IBM was one of the companies that fell victim to the plagued capacitors or not. I thought about disassembling it to see if there were any burned or bulging capacitors that were obviously bad. My concern in doing that is the plastics on a 22 year old laptop can't be the most flexible this day in age and I am worried about snapping off retention tabs and being unable to find replacement parts. I already learned the hard way that the hard drive door on mine was brittle and Shop Goodwill did a really poor job of packing it in a single layer of bubble wrap that the LCD trim bezel is broken and missing a piece.

I also question that if the issue is faulty capacitor related as I would think electronically speaking, the 2 x 128 MB sticks would put near equal load on them to cause the functional pair of 128 MB sticks to fail a test and yet they run for 12+ hours without a single issue.

For what it's worth, each of the Crucial 256 MB modules passed individually on a two to three hour Memtest86+ session each with multiple cycles/iterations completed. Certainly not a comprehensive stress test by any means but I would have seen errors on the 2 x 256 MB module test by that point. I also did not test them individually in the other socket on the motherboard. Like I said, not very comprehensive but at least it seems individually, the sticks can pass a few iterations of Memtest86+ tests.

I did notice that during the failing runs, the memory compartment door is quite warm. As such, I have the laptop propped up, memory compartment door open with a small fan blowing directly on the modules to see if it is heat related. I have that Memtest86+ session running now and will report back my findings.

[RealBlackStuff]

You can PM once you have made at least 5 posts.

You should do your testing in each of the RAM slots individually, as well as combined.
Also, when was the last time you refreshed thermal paste between CPU and fan?

As for shipping:
$13 without tracking
$18 with tracking
$21 registered, with tracking and insurance

Payment accepted via PayPal, using family/friends option.

[hax0rwax0r]

So just to add to the troubleshooting history --

First, to address your question regarding thermal paste - I just got the laptop a few weeks ago so I have never replaced the thermal paste. It is on my to-do list though even before you suggested it, though, as I mentioned I was hoping to find the core of the issue before having to tear into it as I was worried the 22 year old plastics were quite brittle at this point. At this point, I see no possible way forward without tearing into it.

As for my test results, with the RAM compartment cover off and a small fan blowing directly on the RAM modules, the tests still failed. I felt the RAM modules and they were cool-ish to the touch this time so the fan was obviously working to cool at least one side of the RAM modules down. So now I've ruled the RAM heat out as being the cause.

I forgot I had ordered a second set of 256 MB x 2 RAM from eBay the same day I ordered the Crucial RAM. It finally showed up yesterday so I decided to test it as well. I don't know what the actual brand or CAS latency on it is as the memory was simply listed as " 256MB PC100 100Mhz 144-pin 16 Chip Sodimm for PowerBook G3 Bronze KB/Lombard" but it also failed 256 MB x 2 testing. Absolute shocker. /s

As mentioned in a previous post, I have a spare "for parts only" ThinkPad T20 with a non-functional LCD backlight. I hooked it up to an external display and let the 550MHz Pentium III T20 run all night with the Crucial modules in it. It was up to almost four hours when I went to bed and it had completed two passes without errors up to that point with the Crucial RAM. Unfortunately, when I checked it this morning, after running for 14 hours and 20 minutes, it had encountered four total errors at some point through the night and was still testing. I would have felt better if there were zero faults after 14h 20m of runtime but I also feel like given the age of the T20, and the "parts only" tag on eBay listing, it could also be something similar to my T22 with flaky hardware or dried thermal paste needing replacement.

At this point, I think I will be disassembling a T22 over the weekend, at the very least to reapply thermal paste.

[astral]

i do know that the a22 can be picky about ram. besides memtest what happens when you boot to an OS? because with the a22 i know that they can have random crashing, freezing, blue screens, and fail memtest with certain working ram modules just because it only works with certain types. dont know much else beyond that though.

with the a22 that i've seen it was the same sort of thing that you've seen - ran memtest for a while fine and then the errors show up.

perhaps the t2x series has the same issues.

i'd experiment with my t21 if only it worked

[kfzhu1229]

You cannot buy this RAM from Crucial directly anymore. You can't buy it from ANYONE not eBay these days anymore.

That said, I found new Crucial CT32M64S4W7E-16FGY RAM on eBay matching this spec. The condition was new. If the inventory is fake, then they went through a lot of effort to print labels with models and serial numbers on them to seal the outer cardboard packaging as well as printing up a Crucial information insert on nice paper as well as sealed antistatic bags with a Crucial label on them.

Can you take a picture of the labels of all of the RAM sticks involved here? I am very sceptical about "brand new" crucial sticks as most of the time I see them it's fake. Sometimes the fakes are harmless and are simply some mismatched sticks repurposed to make them look matching, but then again the max capacity sticks have bad reliability, like 256MB PC100, 512MB PC133, 1GB DDR, 4GB DDR2 and 8GB DDR3 SODIMMs.
For me if I see a manufacturing date over a decade newer than the RAM technology, it's a fake; if I see any non-micron RAM sticks on the crucial sticks, it's a fake.

[SMA]

Check for a loose memory socket.

To clarify what is ment by loose memory socket.
There is one socket into which 2 memory modules can be installed. This 'dual' socket is soldered to the motherboard.
It is the solder joints between the socket and the MB that comes loose.
Visual inspection using a times 10 magnifying glass might reveal if this is the case.

And as a side note.
When T20/T21 has no backlight it is usually a broken lcd cable.
It is only one wire breaking inside the cable, so it is repairable by running a soft wire along side the lcd cable.
Read about it here https://forum.thinkpads.com/viewtopic.php?t=45602

[hax0rwax0r]

besides memtest what happens when you boot to an OS?

With the original 2 x 128MB RAM it seems fine for the most part. I did maybe notice a bit of instability once or twice but chalked it up to an old operating system or whatever I was trying to do. After all, it was the original installation of Windows 2000 Professional that was from the original owner. The machine has files dated back to 2005 or so that were modified so I think this thing sat in storage for a really long time and eventually donated to Goodwill.

With any variation of 2 x 256MB RAM it random BSODs, assuming it doesn't BSOD on boot. Without a doubt not stable. And that's true of even a reinstall of Windows, too.

Can you take a picture of the labels of all of the RAM sticks involved here? I am very sceptical about "brand new" crucial sticks as most of the time I see them it's fake.

Sure:

For me if I see a manufacturing date over a decade newer than the RAM technology, it's a fake; if I see any non-micron RAM sticks on the crucial sticks, it's a fake.

The sealed RAM package label is dated 2006, so almost two decades ago. I am not saying it's not fake, but I sure would be surprised if it was. I also checked the chip number listed on one of the memory modules and found reference to Micron in some of the search results.

It seems more likely to be new-old-stock inventory but I am curious on what your thoughts are based on the photos. Unfortunately, it seems I already trashed or misplaced the glossy paper printed Crucial marketing material insert, but I do have the installation instructions, box, antistatic bags and RAM itself still as seen here.

There is one socket into which 2 memory modules can be installed. This 'dual' socket is soldered to the motherboard.

I'm not sure I follow. I do have a couple of older laptops that the RAM configuration appears to be one socket, sort of, as the RAM is stacked one on top of the other. If that's what you mean, my T22 has side-by-side (or back-to-back?) RAM that mirrors the opposite side as you can see in the photos. Unless that is what you're explaining... Either way, I am not sure my interpretation makes much difference to the advice you're providing here:

It is the solder joints between the socket and the MB that comes loose.
Visual inspection using a times 10 magnifying glass might reveal if this is the case.

This is a good idea. I will pull the sticks out and use my soldering microscope to zoom in and see if I can check the RAM socket solder joints. To that point, I would assume the 2 x 128MB would be flaky as well if a solder joint were in play here but who knows.

To add even more to my troubleshooting efforts - I bought a ThinkPad T23 machine on eBay (I think this is my 5th T2x machine?) and tested each of the Crucial RAM modules individually as well as together. I ran the tests on the individual modules for over 30 hours on each module and almost two straight days for both paired together and they passed all the tests in the ThinkPad T23. Yes, I know the T23 uses PC133 RAM and not exactly the same as a T22. That said, the RAM in question is PC133 and the actual Crucial website verified and recommended RAM listed (from my archive.org link above) for a ThinkPad T22. As such, I feel it was a valid test to prove the modules themselves could function somewhere in some system without issue.

So on my to-do list that I still haven't completed yet: disassemble and inspect capacitors on system board, inspect solder joints on RAM slots, remove, clean and reapply quality thermal paste to CPU.

Unfortunately, I feel like I am still no closer to solving this than I was two months ago other than confirming that the Crucial RAM seems to be known good. I suspect all the other non-Crucial RAM modules are likely good as well and the T22, and to a lesser degree the T20, are flaky and unstable. I just have to figure out what is up with them. It's just baffling that the 2 x 128MB seems to work fine. If that failed regularly, it would make me feel more sane in my general assessment.

And as a side note.
When T20/T21 has no backlight it is usually a broken lcd cable.
It is only one wire breaking inside the cable, so it is repairable by running a soft wire along side the lcd cable.
Read about it here https://forum.thinkpads.com/viewtopic.php?t=45602

Thanks for the info. I will check that out and see about repairing my broken T20 screen.

[RealBlackStuff]

IIRC the T20/T21/T22 need PC100 RAM, while the T23 needs PC133.

[hax0rwax0r]

IIRC the T20/T21/T22 need PC100 RAM, while the T23 needs PC133.

From my research, as far as I can tell, the T20, T21 and T22 do not require PC100 RAM specifically and that PC100 is the minimum spec. PC133 should, in theory, work fine, especially considering the upgrade kit listed from Crucial for the T22 is PC133.

I think the issue with PC133 incompatibility stems from some (most?) PC133 RAM being high density. This was something I definitely noticed when searching eBay for RAM from this generation - most PC133 SODIMM modules I came across listed were high density. The T20, T21 and T22 require low density modules. The Crucial sticks I have are one of the few I have seen that are PC133 and also low density. The T23 seems to operate on both low and high density which is one of the differences I've found relative to the other T2x series.

Aside from the density, I believe that CAS latency is also relevant. That said, don't quote me on that as fact. I saw people reporting CL2 worked but CL3 did not. I don't think I've read anything official from IBM regarding this, but there's at least one or two posts on these very forums where someone mentioned CL2 being required. That said, the Crucial sticks I have, in addition to being low density, are also CL2.

Here's a couple search results I found referencing PC133 working in the older non-T23 units:

- The memory modules are usually not compatible between T23 (PC133) and T20-T21 (normally PC100 - must be "low density"); it's possible that your T21 has PC133 modules and they could be used on a T23 motherboard, but you would be limiting the memory to 512MB (low density modules topped out at 256MB)

Then go to http://www.crucial.com/store/mpartspecs ... 94A5CA7304
Costs more, but guaranteed to work.
OR, check eBay for these Crucial part numbers:
T20 CT484548 \
T21 CT484656 > either one will do fine
T22 CT484771 /

All of the above are the same RAM, described as: 256MB SDRAM, PC133 • CL=2 • Non-parity • 133MHz • 3.3V • 32Meg x 64
The above is the one exception where PC133 will work in a PC100 slot!
Advantage: if you upgrade to a T23, the RAM will work there also!

Hah, that last quote is the post I linked to in my second post of this thread and it's actually from you! Way back 14 years ago when you were advising someone on what RAM to buy for their T22. That is the same post I went off of when ordering this specific Crucial RAM. That's fantastic.

[RealBlackStuff]

Well, I can't possibly remember all those (over 23k) posts that I made over the years!
But I was right with the PC100, have you tried them yet?

[hax0rwax0r]

Well, I can't possibly remember all those (over 23k) posts that I made over the years!

Yes, I was very amused when I went back this morning while writing my reply and saw the original post I linked for reference mentioning which Crucial RAM to buy was actually you. Man, 2009 both feels not that long ago and also an eternity.

But I was right with the PC100, have you tried them yet?

If by "them" you mean the new-old-stock Crucial 2 x 256MB PC133 CL2 SODIMM RAM modules I purchased recently on recommendation from your 2009 forum post, yes, I tried them in the T22. They also failed Memtest86+ tests. But as I mentioned, that Crucial RAM tests out fine in a T23 so I know the RAM is physically OK.

By using the T23 to test the sticks (and passing might I add), that spawned this whole PC100 vs. PC133 conversation we're having now as the T23 requires and will only work with PC133. Fact is, I am unconvinced I have ANY stable T20 or T22 laptop to test with. If true, I think that is throwing a red herring into the troubleshooting efforts in making me believe the 256MB modules are all bad. Yes, the modules still could be bad; however, I am unconvinced I have any system that can reliably test all of them that I can trust the results on. The T23 requires PC133 so that would disqualify testing any PC100 modules and the T23 tests out fine with the PC133 modules but then it calls into question PC100 vs. PC133. Really, it's just a numbers game - is every stick of the 8-10 or so sticks of 256MB RAM I have now acquired all defective or are these old laptops, some of which were listed as "for parts only," actually just really flaky?

I have a Gateway Solo laptop I just got that I think takes PC100 RAM. If it does, I will test my collection of SODIMMs in that and hope the hardware isn't flaky and I can get some consistent results to start ruling things out.

I also have some assorted RAM I bought on these very forums from PiZzA EnGiNeEr that I can try. Assuming there are some 256MB PC100 modules in the box, I suspect those will surely also fail but time will tell.

[RealBlackStuff]

I meant trying any PC100 sticks in your T22...

[hax0rwax0r]

I meant trying any PC100 sticks in your T22...

Yes, I have tried lots of PC100 sticks in this T22.

Yesterday I finally pulled the heat sink off, thoroughly cleaned the old thermal paste off, reapplied new thermal paste and reassembled the laptop. The laptop had no visible dust inside of it. The CPU fan had the slightest amount of dust on the blades which I took the time to clean off before reinstalling. I also verified the CPU fan freely spun and was not grinding. You can also feel hot air movement from the side vent when the laptop is running Memtest86+. It's now verifiably not improper CPU HSF contact, dirty or broken CPU fan or dried out thermal paste.

I also used my soldering microscope to inspect the traces on the RAM sockets and they are all clean and attached. It's not a broken solder joint on the sockets as far as I can see.

I also went through all the stuff that I got from Thinkpads forum user PiZzA EnGiNeEr last month and found two matched sticks of 256MB PC100 RAM that I tested. Failed.

I went back and tried every single stick of PC100 or PC133 256MB RAM I had (I'd guess 10 or 11 sticks total). EVERY. SINGLE. STICK. Result? FAILED.



Unfortunately, I think I've wasted too much time and money on polishing a turd. I'm up for trying more things if anyone has ideas but I feel like I've used up every trick I had in my own bag.

[RealBlackStuff]

In my first post I suggested to try another CPU (from your T20 in this case).
Also try the one from your T22 in the T20 and see how that goes.

[hax0rwax0r]

In my first post I suggested to try another CPU (from your T20 in this case).
Also try the one from your T22 in the T20 and see how that goes.

Great point - you did say that, actually. In addition to that T20's CPU, I have another PIII that will work in it, too. I will try that hopefully tonight and report back. At least that's free to test! Thanks for the reminder.

[kfzhu1229]

I went back and tried every single stick of PC100 or PC133 256MB RAM I had (I'd guess 10 or 11 sticks total). EVERY. SINGLE. STICK. Result? FAILED.

So just to get it straight, right now the thing is you can't get 2x256MB working regardless of high or low density? But 128MB sticks work and probably single 256MB stick works as well right?
I have a more sinister theory for you and you might not like it if it is the case: that the RAM voltage supply is not stable. Generally it can be that either one of the components on the voltage rail is partially shorted, or that the filtering capacitors are on their way out if they are not Tantulums/Ceramics (which most likely gives a dead short when failed), or that the voltage IC is dying. (I just diagnosed a similar RAM problem on an Acer motherboard and then I found out the Northbridge is partially shorting the RAM power supply rail so that is an L immediately)
In your case, SDRAM runs on just regular 3.3V power rail, which is controlled by the MAX1632 (seen by page 60 of the schematics). So unfortunately it is very possible that your MAX1632 is dying (as that is the other IC that can fail on top of ADP3421 woes). Maybe remove your modem/ethernet card as well in case that is shorting out the 3.3V rail and perhaps stick in a known good wifi card and hook some makeshift antennas, use your known good 128MB sticks and stress the wifi and see if the 3.3V becomes unstable.
If you are really banging your head on this one and have nothing to lose, you can even try hooking up 3.3V external power supply directly to the power rail and see if your RAM crashes then on memtest86

[hax0rwax0r]

In my first post I suggested to try another CPU (from your T20 in this case).

Today I swapped the CPU out for a 1.0 GHz PIII that I got as part of a random parts package from PiZzA EnGiNeEr. It's a SL53S model CPU which is the same that I took out of the machine. If your goal was to have me test with a slower CPU then obviously this test does not accomplish that. If your goal was just to try another compatible CPU that wasn't the one in the system initially then I did that.

Same result. Memtest86+ fails.

So just to get it straight, right now the thing is you can't get 2x256MB working regardless of high or low density?

Yes.

But 128MB sticks work and probably single 256MB stick works as well right?

The original 128 MB sticks it came with, yes, they work as a pair (oddly enough) and they work individually, too. I have never gotten a single stick of 256 MB RAM to ever work - not by itself in slot A, not by itself in slot B or paired with another 128 MB or 256 MB module.

I have a more sinister theory for you and you might not like it if it is the case: that the RAM voltage supply is not stable. Generally it can be that either one of the components on the voltage rail is partially shorted, or that the filtering capacitors are on their way out if they are not Tantulums/Ceramics (which most likely gives a dead short when failed), or that the voltage IC is dying. (I just diagnosed a similar RAM problem on an Acer motherboard and then I found out the Northbridge is partially shorting the RAM power supply rail so that is an L immediately)

Yes, this is the type of diagnostic response I was hoping someone had some insight on. That's the type of thought I had when I mentioned way back in my first post:

I thought maybe there was some other setting or underlying issue (bad capacitors?) that could cause this behavior as it seems flaky at best.

I tried all the basic stuff I could think of prior to even posting on here. I've gotten some good feedback from others for testing things that I didn't think to do. I've remained skeptical this whole time that component swapping individual PC parts out would fix the issue, though, I am thankful for the help thus far from everyone and it's all certainly been worth the effort to test. That said, as time progresses and each part swap is the same net result, I'm more and more convinced it's failure of individual component(s) on the system board causing it.

In your case, SDRAM runs on just regular 3.3V power rail, which is controlled by the MAX1632 (seen by page 60 of the schematics). So unfortunately it is very possible that your MAX1632 is dying (as that is the other IC that can fail on top of ADP3421 woes).

I understand some of these words. Unfortunately, the ones I don't know are the important ones: MAX1632 and ADP3421

I can Google with the best of them, so I will be researching this tomorrow. I understand some of the basics of electronics but fact is, I've never gotten deep into it short of recapping a few motherboards or modding a few of my gaming consoles over the years. It's also been a long time since I've soldered SMD components. I've been eyeing a Hakko hot air rework station and an excuse to buy one and start dabbling in this stuff though.

Maybe remove your modem/ethernet card as well in case that is shorting out the 3.3V rail and perhaps stick in a known good wifi card and hook some makeshift antennas, use your known good 128MB sticks and stress the wifi and see if the 3.3V becomes unstable.

I can tell you that after removing the modem/ethernet card, it did not change the outcome when testing with the Crucial 2 x 256 MB modules. I will see if I can eBay a cheap-ish Mini PCI WiFi network adapter and rig up some antennas to it and test with the 128 MB modules if you really think it would be helpful in proving, or disproving, something related to the 3.3v rail.

If you are really banging your head on this one and have nothing to lose, you can even try hooking up 3.3V external power supply directly to the power rail and see if your RAM crashes then on memtest86

Sounds like something I will have to read and understand the schematics for the T22 to determine where and how to do that. I do have a bench power supply that I can supply 3.3v with if I do decipher what you're advising I do and figure out what I'm actually doing.

Thanks again everyone for the continued help.

[kfzhu1229]

Today I swapped the CPU out for a 1.0 GHz PIII that I got as part of a random parts package from PiZzA EnGiNeEr. It's a SL53S model CPU which is the same that I took out of the machine. If your goal was to have me test with a slower CPU then obviously this test does not accomplish that. If your goal was just to try another compatible CPU that wasn't the one in the system initially then I did that.

I think the point is actually to try a slower CPU - way slower CPU. the 1Ghz coppermine is way too hot (hotter than most common Pentium 4-M models) and only the most reliable motherboards can tackle those. I would stick that only in something like a Dell Latitude C500/C600/C800.

I understand some of these words. Unfortunately, the ones I don't know are the important ones: MAX1632 and ADP3421

The MAX1632 and ADP3421 are 2 voltage regulating IC's that drives the MOSFETs to create the power rails (3.3 and 5V in the case of MAX1632 and CPU Vcore for ADP3421)
These 2 IC's are notorious for failure on these ThinkPads and the blink of death. If MAX1632 IC is dying it's possible your 3.3V power rail is unstable and may cause problems to your RAM stability.

Sounds like something I will have to read and understand the schematics for the T22 to determine where and how to do that. I do have a bench power supply that I can supply 3.3v with if I do decipher what you're advising I do and figure out what I'm actually doing.

I would do that only as an absolute last resort (and don't leave valuable components on the motherboard like the 1Ghz PIII or 256MB sticks if you do that until you see no vicious things happen). There are capacitors that directly filter the 3.3V power rail going to the RAM slots, you can try injecting there.

[hax0rwax0r]

I think the point is actually to try a slower CPU - way slower CPU. the 1Ghz coppermine is way too hot (hotter than most common Pentium 4-M models) and only the most reliable motherboards can tackle those.

OK, I swapped it out again:



Unfortunately, the result is the same:



I might try the Mini PCI WiFi card + 128 MB RAM modules test still. I will have to search eBay to find something compatible and buy it. Unfortunately, I feel like in doing so, I'm just going to buy a Mini PCI WiFi card that will never get used again and I'm just delaying the inevitable and I should try to replace the suspected failed individual components at this point. I need to see if they are even something I can buy still.

[RealBlackStuff]

In the end it might be cheaper to buy another motherboard, e.g. T20: https://www.ebay.ie/itm/304758433118

[hax0rwax0r]

In the end it might be cheaper to buy another motherboard, e.g. T20: https://www.ebay.ie/itm/304758433118

That's a good point and I had considered doing that. I actually saw that very eBay listing the other day when looking for a replacement T22 motherboard. I was actually super amped up to find someone listing TWO boxes of T2x parts on the for sale forum here. As soon as I saw the listing, I replied only to find they had sold it over the weekend to another TP forum member.

I just found out this morning my girlfriend's stepdad has a hot air rework station that he offered to let me borrow along with a dead board he said I could use to practice on. I have all the other soldering stuff (flux, kapton tape, solder, wick, iron, microscope, etc.) needed to pull this off, technically. It's just been a while since I've soldered things so I bought a few little practice kits on Amazon that were just delivered to brush up on my skills with.

I have yet to confirm the exact model/variation of chips the T22 uses, but in general, it seems that some variation of the MAX1632 and ADP3421 are still available to buy and are really cheap.