IBM EGA card  -  Faulty RAM chip


under_construction.gif


If at power on of your IBM 51xx, you hear '1 long beep followed by 3 short beeps', and your IBM 51xx is fitted with an IBM EGA card, then it is highly likely that a video RAM chip on the IBM EGA card is faulty.

If your IBM EGA card has its optional 'Graphics Memory Expansion Card' attached, then do not expect to hear '1 long beep followed by 3 short beeps' if a video RAM chip has failed on that optional card.

If a video RAM chip is indeed faulty, there may be other symptoms, symptoms such as:
* Missing column/s in characters
* Pressing some keyboard keys results in the wrong character appearing on-screen  (see 'Side effects' below)
* Wrong colours  (see 'Side effects' below)


Using CheckIt to identify faulty RAM chip on IBM EGA card

I do not consider CheckIt to be useful for this purpose, for the following reasons:

The false error reporting problem described at here.
A failure to report certain video RAM chips as faulty.  For example (repeat: example), I have a fully functional IBM EGA card, complete with its (fully populated) optional 'Graphics Memory Expansion Card'.  I ground the DQ4 pin on RAM chip U1 on the EGA card.  At power on of the computer, I hear the expected error code of '1 long beep followed by 3 short beeps'.  I then get Checkit 2.1 to execute its video RAM test.  CheckIt's video RAM test unexpectantly passes.  Checkit 3.0 behaves the same.
The IBM EGA card maps its RAM ('display buffer') to different motherboard address spaces, depending on video mode (see second paragraph of 'Display Buffer' section of here).  So when Checkit reports an error at a particular address, how do you work out which RAM chips are being used for that address.


Using the 'IBM Advanced Diagnostics' floppy to identify faulty RAM chip on IBM EGA card

This involves going through the 'MAP 2400: Enhanced Graphics Adapter' process of the IBM Hardware Maintenance Service manual.

RAM chip on Graphics Memory Expansion Card

If the faulty RAM chip is on the optional 'Graphics Memory Expansion Card', MAP 2400 will get you to execute the '24 - ENHANCED GRAPHICS ADAPTER' test on the IBM Advanced Diagnostics floppy.  The '0 - DISPLAY ADAPTER AND MEMORY TEST' subtest of that will report the error of, "FAILING GRAPHICS MEMORY", followed on the next line by a six digit number.  The six digit number indicates which RAM chip is faulty.  MAP 2400 contains a diagram showing how to identify the chip from the code.

IMPORTANT:  It is critical that the '0 - DISPLAY ADAPTER AND MEMORY TEST' subtest knows how much video RAM is fitted to the combination of IBM EGA card and Graphics Memory Expansion Card, otherwise a wrong error code is reported.  So when you see the "256KB GRAPHICS MEMORY INSTALLED.  IS THIS CORRECT (Y/N)?" question, ensure that you answer as appropriate.  If you answer in the negative, you will be asked to enter the amount of video RAM fitted.

RAM chip on actual IBM EGA card

If the reported six digit number starts with 00, then the faulty RAM chip is on the actual IBM EGA card.  Unfortunately, the MAP 2400 process does not have a code-to-RAM-chip diagram for that - informing you instead to simply, "Replace the Enhanced Graphics Adapter." But through experimentation, I have created the code-to-RAM-chip list:

000001 = Chip U50
000010 = Chip U40
000101 = Chip U51
000110 = Chip U41
000201 = Chip U1
000210 = Chip U10
000301 = Chip U2
000310 = Chip U11

A photo highlighting those RAM chips is at here.

It is critical that the '0 - DISPLAY ADAPTER AND MEMORY TEST' subtest knows that the IBM EGA card has only 64K of RAM fitted, otherwise a wrong error code is reported.  So when you see the "256KB GRAPHICS MEMORY INSTALLED.  IS THIS CORRECT (Y/N)?" question, answer in the negative, then enter "64" at the next question.


Side effects

Sometimes (repeat: sometimes) you will see the consequence of a failed video RAM chip on-screen.  For example, when I simulated failure (always low) of the DQ3 pin on RAM chip U41, light grey text appeared as cyan.  For example, sometimes one of more columns of a character were missing.

For example, when I simulated failure (always low) of the DQ3 pin on RAM chip U40, '4' appeared as '0', '5' appeared as '1', '6' appeared as '2', and '7' appeared as '3'.  That created a problem, because when I typed in "24" to run the '24 - ENHANCED GRAPHICS ADAPTER' test, "20" appeared on screen instead, and the diagnostics thought that I had typed "20".