• Please review our updated Terms and Rules here

Dell system 200 memory failure errors

roadrash

roadrash

Experienced Member
Joined
Aug 17, 2011
Messages
153
Location
Huntingdon, U.K
I have got memory errors on my system 200 that we once sorted before some many years ago. I read through the earlier thread and it was a error in the soldered in onboard 640k of ram in chips U80, U81, U70, U71. I ordered in 4 new ram chips to replace the ones that gave the error but its make no difference. So it could other chips this time. Ive got landmark supersoft roms in and its reporting the errors but i cannot work out what chips have failed or need resoldered. Can someone please look at the pictures ive uploaded and help me sort this out again.
 

Attachments

  • 20260324_152922.jpg
    20260324_152922.jpg
    1.9 MB · Views: 34
  • 20260324_152925.jpg
    20260324_152925.jpg
    2 MB · Views: 33
  • 20260324_153224.jpg
    20260324_153224.jpg
    3.2 MB · Views: 31
  • 20260324_153230.jpg
    20260324_153230.jpg
    3.2 MB · Views: 35
The Supersoft ROM is showing bit 6 at address 6F45 (which equates to address 27.8 K)
Referring to [this] diagram, that is bit 6 in bank 0, i.e. chip U76.
 
modem7 said:
The Supersoft ROM is showing bit 6 at address 6F45 (which equates to address 27.8 K)
Referring to [this] diagram, that is bit 6 in bank 0, i.e. chip U76.
Click to expand...
Hi Ray, Thanks a lot for that info its very helpful. Is it normal for these types of RAM chips to fail with age as the joints look perfect. Ive just ordered some more Ram chips and when I get it soldered in I will let you know what happens.
cheers!
 
modem7 said:
Referring to [this] diagram ...
Click to expand...
FYI. Based on the 2017 dated thread at [here]. And I think something was done offline as well.

modem7 said:
The Supersoft ROM is showing bit 6 at address 6F45 (which equates to address 27.8 K)
Referring to [this] diagram, that is bit 6 in bank 0, i.e. chip U76.
Click to expand...
Having a rethink, because the address is odd (out of odd/even), the failing bit was probably bit 14. I will explain.

You have a 16-bit computer, but the Supersoft diagnostic only shows 8 bits in the 'failing address' area. I will get out my IBM AT motherboard later, and prove/disprove it, but I expect things to be:

If the failing address is even, the bit mapping is:
displayed 0 = actual 0, displayed 1 = actual 1, displayed 2 = actual 2, displayed 3 = actual 3, displayed 4 = actual 4, displayed 5 = actual 5, displayed 6 = actual 6, displayed 7 = actual 7

If the failing address is odd, the bit mapping is:
displayed 0 = actual 8, displayed 1 = actual 9, displayed 2 = actual 10, displayed 3 = actual 11, displayed 4 = actual 12, displayed 5 = actual 13, displayed 6 = actual 14, displayed 7 = actual 15

roadrash said:
Ok still got errors after replacing U76
Click to expand...
Three bits at a different address to earlier.

I know that the Supersoft diagnostic ROM does not stop as soon as it encounters a RAM error, and so it is possible that the three bits are at different addresses.

I wonder why the Supersoft diagnostic ROM did not show those before. Is this consistent, i.e. every time you run Supersoft diagnostic ROM, you see the same failed bits at the same address ?
 
modem7 said:
FYI. Based on the 2017 dated thread at [here]. And I think something was done offline as well.


Having a rethink, because the address is odd (out of odd/even), the failing bit was probably bit 14. I will explain.

You have a 16-bit computer, but the Supersoft diagnostic only shows 8 bits in the 'failing address' area. I will get out my IBM AT motherboard later, and prove/disprove it, but I expect things to be:

If the failing address is even, the bit mapping is:
displayed 0 = actual 0, displayed 1 = actual 1, displayed 2 = actual 2, displayed 3 = actual 3, displayed 4 = actual 4, displayed 5 = actual 5, displayed 6 = actual 6, displayed 7 = actual 7

If the failing address is odd, the bit mapping is:
displayed 0 = actual 8, displayed 1 = actual 9, displayed 2 = actual 10, displayed 3 = actual 11, displayed 4 = actual 12, displayed 5 = actual 13, displayed 6 = actual 14, displayed 7 = actual 15


Three bits at a different address to earlier.

I know that the Supersoft diagnostic ROM does not stop as soon as it encounters a RAM error, and so it is possible that the three bits are at different addresses.

I wonder why the Supersoft diagnostic ROM did not show those before. Is this consistent, i.e. every time you run Supersoft diagnostic ROM, you see the same failed bits at the same address ?
Click to expand...
This time round i replaced the 4 ram chips i replaced in 2017 again because it was visible that there was residue of battery acid under and around some of the chips and i wanted to make a better job of cleaning this off the motherboard to it didnt come back again. I obviously didnt do a thorough enough job of cleaning it in 2017. As you said supersoft roms dont stop checking when they encounter a error so maybe I didnt wait long enough for all the errors to be shown. I wonder how long I need to leave it running before i take a screen shot or do i take a screen shot after each pass to show all accumulated errors. Out of interest which chips do these new errors relate to as i still have more new chips i can fit.
 
I brought out an IBM 5170 motherboard of type 3.
* With only the bit 6 chip removed, the Supersoft diagnostic ROM failed the '16K CRITICAL MEMORY REGION' test, indicating a bit 6 failure at address 6E40. Multiple passes did not change that.
* With only the bit 14 chip removed, the Supersoft diagnostic ROM failed the '16K CRITICAL MEMORY REGION' test, indicating a bit 6 failure at address 6E41. Multiple passes did not change that.

And so what I mentioned in post #6 about looking at whether the failing address is odd or even (to determine high byte or low byte) applies.

In a different experiment:
- I removed three chips, bit 0, bit 4, and bit 14. The Supersoft diagnostic ROM only showed bit 0 as faulty.
- When the three removed chips were, 2, 3, and 4, the Supersoft diagnostic ROM usually only showed bits 2 and 3 as faulty. But on some passes, only bit 3 was shown as faulty.

roadrash said:
As you said supersoft roms dont stop checking when they encounter a error so maybe I didnt wait long enough for all the errors to be shown. I wonder how long I need to leave it running before i take a screen shot or do i take a screen shot after each pass to show all accumulated errors. Out of interest which chips do these new errors relate to as i still have more new chips i can fit.
Click to expand...
I see in your post #5, that the error address shown is 6E40, which is the same as what I saw above. If we assume that to be your bank 0, then the bits to replace are 2, 3, and 4, in bank 0. Based on [this], they correspond to chips U72, U73, and U74.
 
Well i replaced those 3 memory chips but still have 2 errors. One is in the same memory address and the other one is in a different one. I dont understand. Also what chip does this new address relate to?
 

Attachments

  • 20260404_154623.jpg
    20260404_154623.jpg
    1.7 MB · Views: 7
  • 20260404_154634.jpg
    20260404_154634.jpg
    1.6 MB · Views: 7
  • 20260404_154644.jpg
    20260404_154644.jpg
    1.6 MB · Views: 7
roadrash said:
I dont understand.
Click to expand...
First, be aware that when it comes to the reporting of RAM related errors, I do not like the SuperSoft Landmark Diagnostic ROM (SLDR).

For example, in post #8, where I removed the bit 6 chip from bank 0 of my IBM AT, I would have liked it had the SLDR stopped at the first erring RAM address. Instead, by SLDR reporting address 6E40 hex, it falsely gives the impression to the user that SLDR tested good addresses 0000 through 6E3F.
Rhetorical: And why address 6E40 ? That corresponds to address 27.5625 KB. If the SLDR is testing all conventional memory, why is not the address shown 512-1 KB, or 640-1 KB ?

BTW. The XT version of the SLDR is known to do something similar, except that the address seen is 5040.

And so, when using the SLDR, one has to 'enterpret' the failing address shown 'with a grain of salt'.

When Ruud created 'Ruuds Diagnostic ROM' (RDR), it was created based on the XT version of the SLDR, differing in that it stops at the first erring address, plus the bits shown for the address are a result of testing all bits at the address. And unlike the XT version of the SLDR, it does not display incorrect and misleading information when fitted to the 16KB-64KB version of IBM 5150 motherboard.

Maybe one day I will start work on an AT version of RDR. For now though, I can crudely modify the XT version of RDR so that it will get to its RAM tests when run on an AT. That will give up a 'better picture'. I don't want that 'hacked' RDR to be public, and so I will PM it to you in the next few days (it's Easter).

roadrash said:
Well i replaced those 3 memory chips but still have 2 errors.
Click to expand...
Being:
- Bits 4, 3, and 2 at address 6E40 (the same as you reported at post #5)
- Bits 4, 3, and 2 at address 6E42

So, address 6E42 in addition to 6E40. This sort of behaviour is sometimes seen with the SLDR. Both addresses are even, and so we know from before, that the address is in bank 0.

Possibilities going through my mind:
- Bad soldering.
- Incorrect replacement chips.
- The diagram at [here] is incorrect.

After Easter, I will modify the XT version of RDR, as I mentioned earlier.
 
roadrash said:
Thats great thanks.
Click to expand...
The modified RDR is in progress.

roadrash said:
What size eeproms do i need to get for RDR diags.
Click to expand...
Whatever the EPROM requirements (type + speed) are in the Dell System 200.
Based on the Dell System 200 information at [here], fitted in your 200 are 27C256's that have an access rating of 150 ns.
So we know that what will work are:
- 27256/27C256, rated at 150 ns or faster
- Winbond W27E257

roadrash said:
What chip was the new error 6E42?
Click to expand...
It is only one word away from 6E40. 6E42 is sure to disappear when 6E40 is fixed.

BTW. Something similar is seen with the POST in the IBM AT. In my description of the 201 error at [here], I have written, "Sometimes, you may see an address that is one word beyond the actual failing address, e.g. 080002 instead of 080000"
 
Re the 'in-progress modified RDR for AT', I have something useable now, although for now, only for the RAM side of things.
Let me know when you are ready to use it.
 
modem7 said:
Re the 'in-progress modified RDR for AT', I have something useable now, although for now, only for the RAM side of things.
Let me know when you are ready to use it.
Click to expand...
Sorry about delay but been busy getting my motorcycles ready for summer here. I got the RDR rom thanks. It must be many years since i last burnt any roms. I have dug out 2 programers. One is a Willem with many adaptors and one called a Mini Pro. The software for the Willem looks like is only useable with 32 bit systems and it uses the parallel port. Again only on old PCs. The mini pro seems the most logical one to use and the software installs ok. I have a old windows 7 laptop here i can use or if i have to i can grab a older XP system computer from the loft.
I have 2 Eproms here and some EEproms coming.
Which programmer/system would be best to burn these roms and are these 2 Eproms ok to use?
 

Attachments

  • 20260421_124018.jpg
    20260421_124018.jpg
    3 MB · Views: 7
  • 20260421_123910.jpg
    20260421_123910.jpg
    1.7 MB · Views: 7
That looks like a TL866, not sure of your exact model, but that eprom seems to be on the list here: www.xgecu.com/MiniPro/TL866II_List.txt
That’s probably the easier programmer to use of the two, but maybe less customizable. I use one on windows 10 or 11 regularly.

Also there is open source Linux software for these programmers.

If those EPROMs aren’t blank, you’d need a UV eraser to make them be blank. Most programmers can do a blank check.
 
Thanks jafir that was my thoughts, The software for the Willem programmer is not recommended for use on later versions of windows and 64 bit architectures. I only use Ubuntu linux and have done for years now . I have a old laptop with windows 7 on it just in case I need it for something (extremely rare now), but even that id too new, so would have to grab a old windows 98 system from my loft. The TL866 software seems fine on windows 7. I will stick the eproms through UV eraser first of course.
 
modem7 said:
Yes. M27C256B-10 have an access time rating of 100ns, which meets the "27256/27C256, rated at 150 ns or faster" option in post #12.
Click to expand...
glad they are OK. Probably have some M27C256 eeproms arriving today too. Always handy to have some spare.
 
You must log in or register to reply here.
Back
Top