XT Clone: POST Memory Error 201

[RobertK]

Last weekend I replaced my broken Taiwan clone Turbo-XT motherboard with a working DTK PIM-TB10-Z board. I also replaced the 256K RAM that came with new board with the 640K from my old board, and the memory check ran ok and everything was working fine with a Hercules screen.

Yesterday I reconfigured the machine to CGA graphics, and suddenly I'm getting a POST error "4000 201" with two short beeps, and it reports that only 256 K RAM are ok (the machine continues to boot after pressing F1).



Questions:

1. What does the "4000" mean that comes with the 201 error message? According to here and here, the result would "16 KB", but is that the first chip of the first bank or the second bank (considering that 256K RAM are ok)? Would the RAM test continue with a broken RAM chip in bank 0?

2. PC Tools report the total memory to be 272K. Does this mean that the first chip of the second bank is good, and the second tested chip of the second bank is bad?



3. Can I rearrange the chips and so find out which one is the bad one?

4. On the IBM 5150 the switches 3 and 4 of SW1 need to be set according to the amount of memory installed.
However, for my DTK board on stason.org it says for switches 3 and 4 of SW1: "Factory configured - do not alter".
My board came with 256K RAM and switches 3 and 4 both ON - I don't know whether these were the factory settings.
My question is: do I have to adjust these switches or not?

Thanks in advance for your help.

 

[modem7]

Yesterday I reconfigured the machine to CGA graphics, and suddenly I'm getting a POST error "4000 201" with two short beeps, and it reports that only 256 K RAM are ok (the machine continues to boot after pressing F1).

So, a coincidence, or not.
What happens if you revert to the earlier ('Hercules') configuration?

What does the "4000" mean that comes with the 201 error message?

It is up the the BIOS author as to how errors are presented. Maybe for your BIOS, the '4000' is a segment address, in which case, that (40000 hex) points to address 256K.

Can I rearrange the chips and so find out which one is the bad one?

You may not have bad RAM. Perhaps your CGA card is faulty, interfering with motherboard RAM. If you revert the machine to its earlier configuration and you are back to seeing 640 KB, a faulty CGA card is a possibility.

4. On the IBM 5150 the switches 3 and 4 of SW1 need to be set according to the amount of memory installed.
However, for my DTK board on stason.org it says for switches 3 and 4 of SW1: "Factory configured - do not alter".

That differs to the information in the PIM-TB10 user's manual at [here].

 

[RobertK]

Thanks, that helps a lot for a start! I will test it with the Hercules configuration again this weekend - the problem is that my Hercules monitor and the computer are currently 100 kms apart.

That differs to the information in the PIM-TB10 user's manual at [here].

Ah, the manual of exactly this board, great! I didn't expect it be on the web so I didn't even search for it!

But this raises more questions (there seem to be a few errors in the manual): SW1 switches 3 and 4 - probably the setting "3 = OFF 4 = OFF 256K MEMORY INSTALLED" stands for "All three banks populated", which in my case would be 640K. And switches 5 and 6: "5=ON 6=ON NO DISPLAY ADAPTER" will hopefully work with an EGA or VGA card.

 

[Chuck(G)]

I've got the same motherboard and run it with a VGA adapter card, if that helps.

 

[modem7]

But this raises more questions (there seem to be a few errors in the manual): SW1 switches 3 and 4 - probably the setting "3 = OFF 4 = OFF 256K MEMORY INSTALLED" stands for "All three banks populated", which in my case would be 640K.

Errors and misleadings is not uncommon.
As an example, see [here] for some of the errors/misleadings in the manuals that IBM created for the IBM PC family of computers.

 

[modem7]

Thanks, that helps a lot for a start! I will test it with the Hercules configuration again this weekend - the problem is that my Hercules monitor and the computer are currently 100 kms apart.

The results of that test may be awaiting moderation.

In anticipation of a possible result of, "No, 4000 201 is still reported", I did some experimentation.

I have a PIM-TB10-Z motherboard (shown at [here]), although mine has a DTK BIOS, not the "XT BIOS V2.05 COPYRIGHT Award Software Inc. 1986" BIOS that yours has.
Banks 0 and 1 are labelled as such.

I removed the DTK BIOS, and substituted in the "XT BIOS V2.05 COPYRIGHT Award Software Inc. 1986" one.

- Removing a RAM chip from bank 0 (0 to 256K) resulted in what appeared to be a dead motherboard. The exception was the U57 chip (at the end near the resistors). With U57 removed, there was no error and the POST indicated '640 KB MEMORY GOOD'. Presumably, U57 is the one used for parity in bank 0.

- The symptom seen if a RAM chip from bank 1 (256K to 512K) was removed, varied:
Chip U56 = '640 KB MEMORY GOOD'. (U56 must be for the parity bit.)
Chip U60 = '256 KB MEMORY GOOD'. No 201 error.
Chip U64 = '4000 201' error.
Chip U68 = '256 KB MEMORY GOOD'. No 201 error.
Chip U73 = '4000 201' error.
Chip U77 = '256 KB MEMORY GOOD'. No 201 error.
Chip U79 = '256 KB MEMORY GOOD'. No 201 error.
Chip U84 = '4000 201' error.
Chip U89 = '256 KB MEMORY GOOD'. No 201 error.

To be noted is that the chip labels ('U56', 'U60', etc.) printed on the motherboard do not properly line up with their respective IC socket.

 

[RobertK]

modem7, thanks a lot for your research!!

So, if I swap U56 with U64, U73 and U84, I should know which one of the latter three is bad.

I only fear that either the graphics card or the mainboard itself could be a RAM killer, and a replaced RAM chip could get bad again.

But before that, I will try the Hercules configuration again: if the memory test should run fine again, then the graphics card (a combined CGA/Hercules one) must be definitely causing the problem.

I didn't mention before that in Hercules mode the card behaved quite strangely, despite being configured correctly according to the manual that came with my PC. Text mode was always fine, but I tried a few Hercules games and just couldn't get anyone running. For example, Spacewar (swh.exe) reported that no Hercules card was present, and the CGA emulators all crashed immediately. My own game H-Tron didn't run properly either, but at least it eventually displayed the game screen correctly. In CGA mode however, the card runs without any problems.

 

[RobertK]

Ok, I just did some testing. First, I configured the machine to Hercules graphics again, and the 4000 201 error was still there. So, there must be damage to one of the RAM chips in bank 1.

I labelled each chip and began swapping each potentially bad chip with the "parity chip" U56.

I started with U64, U73 and U84 (the ones that produced the 4000 201 error in modem7's test), but that had no effect.

Then I tried it with the remaining chips, and when I had swapped U73 with U56...



...I suddenly got a successful 640K RAM test result!



I swapped them again to make sure that it wasn't just a contact problem, and the 4000 201 error was there again, and after swapping them once more the 640K RAM test ran fine again. PCTools also report 640K to be present.



So that problem seems to be solved, thanks! But should I bother about the bad chip that is now parked at the parity position? Will it have any negative effect? Should I better replace it with a working one? These Siemens chips are currently easily available on eBay.

I'm afraid of switching to CGA again, so I will first try it with an EGA card tomorrow.

 

[modem7]

So, there must be damage to one of the RAM chips in bank 1.

Of course, RAM chips can just fail. Relatively common (see [here]).

Via experimentation, I established that (viewing the motherboard with the expansion slots at the rear) the motherboard RAM chips run [7][6][5][4][3][2][1][0][P] from left to right.

So that problem seems to be solved, thanks! But should I bother about the bad chip that is now parked at the parity position? Will it have any negative effect? Should I better replace it with a working one?

It is normal for the POST in PC family compatibles to disable reaction to RAM parity errors, then at the end of the POST, enable reaction. But with the "XT BIOS V2.05 COPYRIGHT Award Software Inc. 1986" BIOS, and the parity chip removed from both banks 0 and 1, my motherboard is not showing any of the symptoms that are expected for a bad/missing parity chip. The RAM check in the POST is not pointing out a bad parity chip. I can boot to DOS and successfully execute a range of DOS programs. And using DEBUG, I can successfully write and read back many different test bytes to/from banks 0 an 1, all without a RAM parity error being generated. It is is though the POST in the "XT BIOS V2.05 COPYRIGHT Award Software Inc. 1986" BIOS disables reaction to RAM parity errors very early, and never subsequently enables reaction.

So, if your "XT BIOS V2.05 COPYRIGHT Award Software Inc. 1986" BIOS is like mine, then in effect, the parity chips are not being used.

Out of curiousity, I put back in the DTK BIOS, and the parity chip for bank 0, leaving the parity chip for bank 1 missing. When I then used DEBUG to write the test byte of 01 to bank 1, a parity error appeared. The DTK BIOS presented that as "P-CK 1(M)".

 

[RobertK]

I did some more testing over the weekend, and things are getting even more complicated. My conclusion is that there is definitely some problem with the CGA/Hercules graphics card, and the RAM chips *could* still be ok.

After the successful 640K RAM test with Hercules graphics, I tried the following configurations. I don't have an EGA display so I'm using a CGA-to-SCART adaptor with my TV set. In EGA mode I can't see the POST messages because this runs in 640x350 text mode which a TV set of course can't display.

1. EGA - all ok



2. CGA - I got an "6000 201" POST error (with 384 KB memory good)! I immediately thought that the graphics card must have destroyed one more RAM chip on bank 1.



3. Hercules - same "6000 201", but here with 400 KB memory good.



4. EGA - the RAM test ran ok again!

5. VGA - all ok!



6. VGA with the initial RAM configuration (with U73 and U56 at their original positions): 640K RAM test successful! Then I swapped chips U73 and U56 again for the Hercules-safe configuration.

7. Hercules - 640 K RAM ok again!



Especially the last test result is weird (why it worked here and didn't work previously), and also why the RAM chips need to be swapped with the Hercules card while this is not required with VGA.

Here is the problematic CGA/Hercules graphics card, it came with my PC when I bought it in 1989, but I never used it and immediately replaced it with an EGA card. I couldn't find this model on stason.org, it would be interesting to know what J3 and JP1 are for, maybe they are misconfigured.



I will try to find a working CGA card, because this would be my preferred configuration for this machine (until I find an EGA monitor at a reasonable price one day).

Final question: can I take the Phoenix BIOS chip from my old, almost identical motherboard and put it into the new one without the risk of doing any damage?

 

[modem7]

Final question: can I take the Phoenix BIOS chip from my old, almost identical motherboard and put it into the new one without the risk of doing any damage?

On my PIM-TB10-Z motherboard, the BIOS ROM is an EPROM of type 2764. If the subject 'Phoenix BIOS chip' is also a 2764, then give it a go.

 

[RobertK]

It says 27C64A on the old motherboard's BIOS ROM, that should hopefully match.

I'll put it in this weekend to see what will happen during memory test, and for nostalgic reasons to see that BIOS startup message again.

 

[Timo W.]

I would recommend checking the stability of the +5V rail from the PSU. RAM errors like this can be a result of voltage dropping or ripple current. And the fact that you get mixed results with different graphics cards would fit as well, as different cards draw more or less current, hence loading the 5V rail differently. Apart from the PSU, bad caps can cause the same issues.

 

[RobertK]

After some more testing last weekend, my conclusion is now that some RAM chips from bank 1 *are* bad and that a successful RAM test does not mean that you have a fully working RAM.

I have installed an XT-CF-Lite V4 card which works perfectly, so that I could now try a lot more software.

Both with Hercules and VGA (but *not* with EGA), the system very often hangs with the message

PARITY CHECK 1
4000 (s)

which is probably caused by the bad chip at the "parity position" of bank 1.

This happens for example every time I try to change the directory in Volkov Commander (making it pretty much unusable), but also when I start certain games. Especially with VGA I was hardly able to run any game.

Interestingly, with an EGA card installed most games (but not all) run fine. A few however have the problem that they run ridiulously slow, for example Outrun, Test Drive 2 and Grand Prix Circuit. Especially the latter is weird, because I was already running this game successfully one week ago with the current problebmatic RAM set, and it was exactly this EGA card that I had played the game with in the late 1980s.
On the other hand, more demanding games like Indianapolis 500 run without any problem.

I would recommend checking the stability of the +5V rail from the PSU. RAM errors like this can be a result of voltage dropping or ripple current. And the fact that you get mixed results with different graphics cards would fit as well, as different cards draw more or less current, hence loading the 5V rail differently. Apart from the PSU, bad caps can cause the same issues.

That explanation makes sense. I don't have an oscilloscope, but I have measured the voltage at the keyboard connector with a multimeter, and the result was constantly 4.98 V - even when the car selection scene in Grand Prix Circuit was running at 1 FPS or even slower.
In addition to the XT-CF-Lite card, I installed an Adlib clone sound card and I was expecting a power shortage, but the problems didn't get worse. The Adlib card is actually working fine, Arkanoid 2 is always a good test for Adlib sound on an 8088 PC.

So my next step will be getting replacement RAM chips. I could first get two or three of the Siemens HYB41256-15 chips that I have in bank 1.
The other option would be getting nine Oki MSM51C256-80 chips (they are very inexpensive on Ebay) that along with those nine that came with my new board would give me 512K of faster (80ns) and hopefully reliable memory. However, the 4164 and 4464 chips for upgrading from 512K to 640K seem to be harder to find, and I assume that I shouldn't mix RAM chips with different speeds. I am also unsure whether I may mix CMOS and NMOS chips, as the Oki ones are CMOS (see here).

 

[modem7]

After some more testing last weekend, my conclusion is now that some RAM chips from bank 1 *are* bad and that a successful RAM test does not mean that you have a fully working RAM.

Yes, the RAM test component of a power-on self test is usually quick-and-dirty.

PARITY CHECK 1
4000 (s)
which is probably caused by the bad chip at the "parity position" of bank 1.

Agreed.
The "(s)" sigifies that the address displayed is a segment address. 4000:0 corresponds to 256K, pointing to bank 1.
A parity error because a read was done of an address somewhere in bank 1, an address where the parity of the stored byte did not correspond with the stored parity bit.

(This will not happen for all addresses in bank 1.)

... would give me 512K of faster (80ns) and hopefully reliable memory.

Because the motherboard's RAM timing remains the same, there will be no performance gain.

However, the 4164 and 4464 chips for upgrading from 512K to 640K seem to be harder to find, and I assume that I shouldn't mix RAM chips with different speeds.

Not a problem with these old motherboards. When the CPU reads a memory address, it waits a fixed period of time (i.e. based on motherboard maker's minimum spec for RAM, the RAM has been given adequate time to place its data onto the data bus), then the CPU reads the data bus. There is no consequence if some RAM chips of a bank output their data before other chips in the bank.

I am also unsure whether I may mix CMOS and NMOS chips, as the Oki ones are CMOS (see here).

Not a problem. The key is that the Oki chip's "I/O pins are TTL compatible."

 

[RobertK]

Ok, I have now replaced the faulty RAM chips, and I think that I finally have a fully working machine.

I have bought a set of five Siemens HYB41256-15 chips from an Ebay shop in Poland.

Before replacing any RAM chip, I got the idea to run a memory test with CheckIt. But I wasn't even able to run that program, it complained about a parity error every time I tried to run it:



So I first replaced the now rightmost chip at bank 1 (U56, previously at U73) which I knew was defnitely bad. After that, Volkov Commander was working correctly (from then on I could change directories), but the other games and programs (including CheckIt) were still not functional.

So I decided to try replacing always one chip of bank 1, starting from the left, and trying to run CheckIt after each attempt. And when I had replaced U77...



...I was able to run CheckIt, and its memory test was successful!



So I eagerly wanted to see how it works with EGA, and basically all games that I tried - and I tried many - worked perfectly. Grand Prix Circuit, Outrun and Test Drive didn't show any slowdowns anymore.

VGA was also fully functional, all games worked and also CheckIt's memory test.

And now the final question: is my CGA card a RAM killer? Since I had three spare RAM chips left, I decided that I had to find out. And the result was: the POST RAM test ran fine, and also the CheckIt memory test was a full success:



So it was probably coincidence that the RAM chips became faulty after I had switched to the CGA configuration for the first time.

There could be one final little problem: when I tried to run Ironman Stewart's Super Off Road in VGA mode, it crashed on a few attempts, and on one occasion I suddenly got the parity error again!
And every time I tried to start the game with the CGA configuration ("game /CGA"), it restarted my PC at the title screen.
So I thought that this game could be a good test to identify one more bad RAM chip. I tried to replace one more RAM chip and ran the game after each attempt, but eventually the problem was still there. Strangely enough, the game runs without any problem with the EGA configuration.
So I hope that this won't cause me any problems in the future.

 

[channelmaniac]

Burn this BIOS into an EPROM and run it on the motherboard. It'll sniff out bad memory.

http://www.minuszerodegrees.net/supersoft_landmark/5155_5160/256KB-640KB.htm