Home 

IBM 5160  -  IC Sockets U18 and U19


Introduction
Early 5160 motherboards
Late 5160 motherboards


Introduction

On the 5160 motherboard, there are two IC sockets that hold the BIOS ROM chips, sockets U18 and U19.

On early versions of the motherboard, the sockets are wired for Mostek MK37000 series and MK38000 series ROM's (or equivalent).  In later versions, the sockets are wired for 27256/27C256 EPROM's.



Early 5160 motherboards

Early versions of the IBM 5160 motherboard are wired for Mostek MK37000 series and MK38000 series ROM's (or equivalent).   Photos of example sets are in links at here.

For the U18 and U19 sockets, if you find that pins 1 and 27 have continuity (0 to say, 2 ohms), then you have the MK37000/MK38000 wiring.

27256 substitution

The problem with putting in a 27256/27C256 (or equivalent), is that the ROM socket (not designed for 27256/27C256) is taking pin 1 of the ROM to LOW for the first half of the 32K, and taking pin 1 to HIGH for the second half of the 32K.
That violates a requirment that is in the data sheet of many 27256/27C256, which is that pin 1 (Vpp) be at +5V for all reads of the 27256/27C256.
For such 27256/27C256, the early 5160 motherboard will only be able to successfully read the second half of the 27256/27C256.

But for some make-models of 27256/27C256, even though the data sheet of that make-model has the pin 1 requirement, it was discovered by at least one person, that the make-model appears to work on the early 5160 motherboard.

You should take the following table 'with a grain of salt', because makers may have changed the design of their ROM's at some point/s.

Make Model Type Works ? 'Works' info source     Comment Data sheet Pin 1 requirement
AMD AM27C256 EPROM Yes modem7 at the VCFF  Make-model verified by electronic signature (device ID). Download No (page 6)
Atmel AT27C256R PROM Yes Moogle! at the VCFF  OTP EPROM  (an EPROM without a window for erasing) Download Yes (page 3)
Atmel AT28C256 EEPROM Yes Malc at the VCFF  Not a 27256/27C256 equivalent - see here. Download No (page 6)
National Semiconductor NMC27C256Q EPROM Yes mrmanse at the VCFF   Download Yes (page 8)
NEC µPD27256 EPROM       Download Yes (page 1)
NEC µPD27C256 EPROM No mrmanse at the VCFF  More info at here.   Note that NEC normally do not print the "µP" prefix on their chips. Download Yes (page 3)
NEC µPD27C256A EPROM       Download Yes (page 4)
ST Microelectronics M27C256B EPROM Yes modem7 at the VCFF  Make-model verified by electronic signature (device ID). Download Yes (page 3)
Texas Instruments TMS27C256 EPROM Maybe See note 1 below  See note 1 below Download Yes (page 3)
Texas Instruments SMJ27C256 EPROM Yes pcdata76 at the VCFF  Do not confuse with the ASI version of the SMJ27C256 - see here.    
Winbond W27E257 EEPROM Yes modem7 at the VCFF  See here. Download Yes (page 3)


Note: For the IBM 5160, use ROM's that are rated at 250 ns, or faster.  (Reference.)

Note: Be aware of possible relabelling of 27256/27C256 EPROM's that you source from eBay (and other sites).



Later 5160 motherboards

At some point, IBM switched to using 27256 EPROM's, or equivalent.  (Examples.)

To accomodate that change, the wiring of sockets U18 and U19 was changed.  Pin 1 is no longer connected to pin 27; instead connected to the +5V line.

It is probable that the change was made in the move from the 64-256KB version of the motherboard to the 256-640KB one.  However, contradicting that, is that the circuit diagram of the 256-640KB motherboard in the MAR86 edition of the technical reference, still shows the earlier pin 1/27 wiring.  A documentation oversight ?

The photos at here show the rewiring of the U18/U19 sockets on a 256-640KB labelled motherboard that I have, this motherboard being one that I believe IBM upgraded from a 64-256KB one.  Another one is shown at here, excepting the mod is all done on the component side of the PCB.




Note 1 Malc at the VCFF indicated suitability of the TMS27C256 for early 5160 motherboards.
However, in 2026, ricsne2010 at the VCFF discovered that his Texas Instruments TMS27C256 (make-model verified by electronic signature) were unsuitable, behaving like the NEC µPD27C256 (first half cannot be read).
Perhaps Texas Instruments changed the design of its TMS27C256 at some point/s.