T20 Blue screens w/ memory upgrade *Large PIC*

[KyleDEA]

I recently inherited a T20.

The memory setup that it initially had was the stock 128mb pc100, and an "add-on computers" 256mb pc-100 144-pin dimm

altogether the machine recognizes 348mb

I had 3 other sticks of the add-on computers 256 laying around and tried to upgrade to 512, but it blue screens at random.

Windows will load, and the pc recognizes the 512mb but will blue screen on random tasks.

I ran memtest on each individual stick to see if there was one that was bad, none of them failed in either slot. BUT, when the 256X2 placed together they fail at about 16% through the memtest.

I read the sticky post about the memory crucial had for sale and read there were specific types to look for such as lo-density and CL2; the website for the addon memory doesn't really say much detail about the memory such as timing or anything like that.

http://www.acp-ep.com/sdram_memory.html

I was wondering if there's something I should know about the memory setup for getting this pc to 512 and why I am getting the BOD with the extras I have laying around. Thank you

[rkawakami]

There shouldn't be anything special you have to do to run 2 x 256MB in there, as long as the modules are low density (16 chips on each module). As you are able to boot memtest and Windows (ignoring the blue screens), then I'd say your modules are the correct type. If they were either high density or the wrong CAS latency, your system wouldn't even boot.

As you have run memtest on the modules and have only seen errors when two modules are installed, I would have to say that something in the modules or your T20 is marginal. From your description that the failures are occurring at about 16% into the testing, that would imply that you are getting fairly far along. About 10 minutes or so? Maybe some clarification is needed...

- what version of memtest are you using? memtest, memtest86 or memtest86+?
- if there is a test name/test number listed when the failures first start appearing, what is it?
- if there is a display of "good" and "bad" data, what is a representative sample?

For instance, this site documents a problem I had with a module a few months ago:

http://www.kawakami-ca.com/ebay/130209500441/

The first photo shows the output from memtest86+ v1.65. The program had run for a total of almost 70 minutes when the picture was taken (under "WallTime", 1 hour, 9 minutes, 58 seconds is listed) and failed at Test #7 (the line in red). The "Good" and "Bad" columns shows the data that was expected and was actually read. Next to that is "Err-Bits", which shows what output pin(s) failed. These data words are displayed in a hexadecimal number system (base 16). Suffice it to say that there's only one output pin that failed that test (as represented by the "00000040").

Why do the modules pass when only one is installed but fail when two are used? My best guess is that when you have two modules in the system, there's some additional electrical "noise" on the power supply lines to the memory and that results in some errors. You could try cleaning the contacts on both sides of both modules using a pencil eraser or isopropyl alcohol. That might or might not help. The problem could just be in the design of the modules or a flaky part. Or there could be an issue with the motherboard in the T20. If you continue to see failures to memtest using your existing modules, you could try sourcing another module (say, Micron or Samsung or another "major brand-name") and see if that helps.

[KyleDEA]

Thank you for the quick response. As an update, I actually got another t20 that I put the same modules into (2x256) and ran memtest86+ and passed just fine with 512.

I cleaned the pins with a pencil eraser and placed back in the initial t20 that was giving me the issues, and I received less errors, but errors none the less.



failing address- 00006426a28/00007426a08
good- ffffdfff
bad- ffffffff
err-bits- 00002000
count- 1

2 errors total

also, this time it took 53 minutes to receive these errors. So what I know now is that it seems to be pc specific due to not getting the errors on the new t20.

Is there anything I can try with the pc itself for this to be corrected?

[rkawakami]

The only things that comes to mind would be to make sure that all of the large bypass capacitors and inductors are firmly connected to the motherboard or use another set of memory modules. Let me explain and sorry for some technical details...

The screen shot you provided tells me that the module failed during one of the worst possible data conditions that the memory can be subjected to. First technical explanation:

Hexadecimal "ffffdfff" means 32 bits of data is being used. Each hex character (0-9,a,b,c,d,e,f) is 4 binary bits. Thus:

Code: Select all

Hex ffffdfff equals binary 1111 1111 1111 1111 1101 1111 1111 1111

This means that 31 data bits are supposed to be a "1" and only one is supposed to be a "0". The "Bad" data is "ffffffff" which means that the memory diagnostic "read" all ones; even for the bit that was supposed to be a zero. The "Err-Bits" field confirms the position of the failing data:

Code: Select all

ffffdfff: 1111 1111 1111 1111 1101 1111 1111 1111 (Good or expected data)
ffffffff: 1111 1111 1111 1111 1111 1111 1111 1111 (Bad or read data)
00002000: 0000 0000 0000 0000 0010 0000 0000 0000 (a "1" means a fail in that bit position)

Technical point #2:
From a circuit design point of view, the time when the most electrical current (power) is being consumed is when all of the data bits are going the in same "direction". That is, when they are all trying to go to zeros or ones. This stresses the power system into providing a large amount of current. That can cause a sharp, but very brief reduction in the voltage being generated. This is usually referred to as "power supply noise". These power fluctuations can be enough to either write the wrong data into the memory array or falsely read the wrong data.

Technical point #3:
In order to "smooth" out these power spikes, capacitors are connected to the power supply. This occurs on both the laptop's motherboard and the memory module. Since you seem to have one T20 that passes a set of modules that fail in another T20, I can only assume that the T20 that fails either has a weak power supply or some of the capacitors are not doing their jobs. The only way to check would be to tear down the system, remove the motherboard and closely inspect all of the large capacitors.

Here is a link to a thread which has information about the T20 motherboard:

http://forum.thinkpads.com/viewtopic.php?t=55300

If you click on the "top side" view, the large black cubes in location C5 are capacitors. On the "bottom side" view, the orange cube at I9 is also a capacitor. I mentioned inductors earlier... those can be found at E5/F5 and G6 on the top. They are the mostly black, squarish looking components. If any of those parts are loose, that can affect the system's power supply. Of course there could also be other components in the power supply circuits which are marginal.

In short, it may not be easy to find and fix any component(s) on the motherboard. Using a different set of memory modules might work since they may not be as sensitive to the power supply noise.

[nikki605]

FWIW, I found a similar problem with my T21 back at the beginning of the year. Ray and several of the other members also helped me at that time.

Bottom line - I never found the real source of my errors and I am still running the same 2 sticks of memory. I have not had any BSODs or other errors, but I keep my experience in the back of my mind, just in case it reappears.

Here is the link:

http://forum.thinkpads.com/viewtopic.php?t=56102

[KyleDEA]

Wow, thank you rkawakami that was very informative.

I'm just going to use 348 until I either purchase new modules or take the time to check the capacitors on the mobo.

thanks again.