Under Construction, watch your head!
The P70 Project
|This page details my attempts to upgrade the IBM 8573 P70 to get the most out of this neat little PS/2 portable computer. I will also include information on the P75, which is really pretty cool just the way it is.....|
|P70 Upgrade Paths|
|P75 CPU Upgrade|
|P75 CMOS Battery Replacement|
|Return to My Home Page|
|The IBM P70 is a PS/2 Microchannel portable, often called "luggable",
386 computer. It had no batteries, which required you to lug a power
cord along with it wherever you went. It was a fairly sophisticated
machine, having more in common with a high-end server than with other luggables
of it's day. It came in 386DX-16 and -20 MHz speeds, with 30 MB,
60 MB, or 120 MB DBA ESDI hard disk options. The display is a gas-plasma
type, which means basically that it is an incredibly complex neon lamp.
This makes it one of the few modern computers that can be said to actually
contain a vacuum tube! The display/controller combination supports
standard VGA resolution at 640x480 pixels (16 gray levels), and includes
a 15-pin connector on the rear panel for an external VGA color monitor
(16 colors supported). CGA and EGA resolutions are also supported.
A maximum of 8 MB of RAM is supported on the system planar, with an additional
8 MB on a memory expansion card in one of the two microchannel slots.
One 16-bit and one 32-bit Microchannel expansion slots are provided, along
with a socket for a 387DX math coprocessor. Other features include
a PS/2 mouse port, serial port, parallel port, internal 1.44 MB floppy
disk drive, and an external floppy disk drive port. All of this snaps
together in a neat little package about the size of your average briefcase
- a lot of technology in a small area in it's day!
The P70 came in two planar versions: the older 38F4688/65X1564, used primarily in the 20 MHz -061 machines (but also seen in some -121 20 MHz boxen), and the 38F6973/56F9085 used in 16 MHz and 20 MHz -121 models. The 38F4688/65X1564 planars have the 386DX socket immediatly to the right of the 387DX coprocesssor socket, and two BIOS ROMs near the upper edge of the board. The 38F6973/56F9085 has the 386DX socket above and to the left of the 387DX socket. Another distinguishing feature of the newer models is the inclusion of video output filtering on the video card instead of on a 'daughter card' fitted in series with the video output cable. The newer planar seems slightly faster on some benchmarks, but the difference is hardly noteworthy.
P70 Model Number Breakdown
|8573-031||30 MB Hard Disk, 16 MHz planar|
|8573-061||60 MB Hard Disk, 20 MHz planar|
|8573-121||120 MB Hard Disk, 20 MHz planar|
Return to Table of Contents
|I've been getting a few questions on this lately, so here it is.
There are two mounting methods, referred to as H1 and H2 in IBM documentation.
H1 was mainly used in the P75, but I have also seen it in the P70.
Louis Ohland has H1 removal instructions here.
Detailed H2 removal can be found here.
While you are there, look at the rest of the P70 and P75 info on his web
For the H2 version, remove the plastic housing, starting at the left side. Then remove the large screw that will be visible as you peek in below the bottom edge of the drive where the cable plugs in (use a flashlight). The frame will then slide up and out.
Blow the dust out of the drive and run a cleaning diskette through it before giving up on it completely; sometimes this is all it takes to get them operational. Check Bob Eager's P70 page for replacement drive part numbers and sources.
NOTE: if the machine is unconfigured (dead CMOS battery), it will ONLY boot from a reference disk or a DOS boot disk that has been tricked up with Bob Eager's REFSTAMP utility. Boot from the internal hard drive will not be possible until the machine has been configured.
|If you have just replaced the CMOS battery, or installed a new card,
you may get this error. There is nothing wrong with the computer
or the option card. This is easy to fix; you will need the reference
disk and the option disks for any option cards you have in the machine
(usually available for download from the card manufacturer). If the
option disk is not available, you can probably find the required ADF files
Wendt's web site. Follow his downloading instructions carefully,
and copy the file(s) to a blank diskette.
Next, power the machine up and boot on the reference disk. If the machine wants to run 'Automatic Configuration', say NO at this point. Select 'Copy an Option Diskette' from the first menu, and follow the directions on the screen. This will copy the option files to your reference disk, and the error message will not appear the next time you run the configuration.
Return to Table of Contents
Most of the good stuff can still be found HERE. Send a thank-you note to Louis - really.
Return to Table of Contents
|According to IBM, 16 MB is it. End of story. The planar
supports a maximum of 8 MB, using four 2 MB 72-pin SIMMs with presence
detection. You can have another 8 MB on a RAM card. But
as we all know, there are often workarounds. One excellent workaround
is to use one of the Kingston/AOX MCMaster
busmaster processor complexes. This is perhaps the best option for
many people, as it also gives you a processor upgrade with the associated
architecture and chipsets to get the most out of it (more on this later).
The MCMaster 486 models support up to 64 MB in a 32-bit MCA slot, and have
two 72-pin SIMM sockets. Later versions, flashed with the most recent
BIOS, support non-parity SIMMs. Although the installation
guide shows 16 MB as a legal combination, when I installed two 8 MB
SIMM modules in mine it would not recognize the full 16 MB. One 8
MB and one 4 MB module worked fine, as does the 8 MB module alongside an
IBM 92G7322 16 MB 60nS non-parity module. The computer runs quite
nicely with 24 MB of RAM! Louis Ohland had this to contibute on the subject
of 8 MB SIMMS:
It would seem that the MCMaster doesn't support clone-type dual-RAS (double-sided) 8 MB SIMMs in both sockets at the same time. Other combinations should be possible with single-RAS chips; anyone who has information, please send it to me for inclusion here and I will give appropriate credit.
It may be necessary, with faster than 33 MHz CPU's, to set the MCMaster to "Report MCMaster Only" memory, as the planar memory may not be able to keep up and lockups will result. In fact, if you are running an advanced operating system like Linux, Window 9x/NT, or OS/2, you will want to set the MCMaster up to use only it's own onboard memory. The reason for this is the MCMaster has to access the planar memory over the MCA bus; this is much slower than accessing the onboard memory, and will result in a signifigant performance penalty when using a "flat memory model" 32-bit operating system. I found this out one time when I put an ISA RAM card in a fast 486 clone; OS/2 adjusted itself to run at the speed of the slowest memory in the system. It wasn't a prettty sight. More RAM is not necessarily better.
Other memory cards should be able to break the 16 MB barrier with the "stock" 386 P70, as has been documented on Fred Spencer's pages for the PS/2 Model 70/80 (subject to the same limitations). The main disadvantage to this approach is that you need to use a 32-bit memory card, which takes up your one-and-only 32-bit MCA slot. This prohibits using an IBM SCSI adapter or long-card network adapter in the computer. One potential RAM card is the Microram 386 (941366, Rexon/Tecmar Inc.). Other possibilities may be the "enhanced" cards from Kingston, Acculogic, and IBM, among others. If anyone has successfully done this, I would appreciate hearing from you (remove the spamblock from my email address).
|IBM did not intend the P70 processor to be upgradable, as far as I
have been able to find out. It's possible that one of the IBM upgrades
for the Model 70/80 may work, but every one that I have seen has not been
usable due to the orientation of the CPU socket in the P70. An IBM
upgrade for the 5530 (if such exists) may fit the -031/-121 planar, however.
Anyone having more definite info, please let me know.
Various aftermarket plug-in 386-to-486 CPU-only upgrades should work. I had a Cyrix 386/486 CPU installed for a while, that is pin-compatible with the 386DX. It worked fine, although the P70 BIOS did not support the write-back L1 cache in the CPU. DON'T enable write-back if you try this, or you WILL experience hard drive corruption! The increase in performance is nothing to write home about, but if you just need to add a little speed, it's the most economical way to go. You do need software from Cyrix to enable the internal L1 cache, though. I will be publishing comparative benchmarks here when I get time to take some data.
Kingston offered two types of 486 Now! CPU upgrades (in SX and DX versions) for the P70:
486/33PD3-P70 (DX) and 486/33PS3-P70 (SX) for the 38F4688/65X1564 planar used in the 20 MHz -061 model.
486/33PD3-P70LP (DX) and 486/33PS3-P70LP (SX) for the 38F6973/56F9085 planar used in the 16 MHz -031 and 20 MHz -121 models.
(Photos are ©Copyright 1993, Kingston Technology Corp.)
Note that I have also seen the 20 MHz -061 planar used in the -121 model. It is important that you get the right model for your planar, as the orientation of the CPU socket is different and the upgrade will not physically fit if it is not the correct one.
Fellow PS/2'er Karsten Harder was kind enough to send me a 486/33PD3-P70LP fitted with an AMD 486DX-40 CPU from the factory (thanks, Karsten!). Installation was simple and straightforward, and nicely documented in the Kingston manual. If the planar is fitted with a 387DX math coprocessor, it must be removed (even if you only have the SX upgrade). After properly setting the four DIP switches, the machine booted and seemed to run fine (although it seemed to take much longer to initialize the POST routines on power-up). I see no reason why clock doubled or tripled CPU's could not be used with this upgrade, space inside the P70 permitting. I did test the upgrade with an Intel 486DX4ODPR-100; it came up fine, and benchmarks reported a CPU speed of 120 MHz. Be careful when removing the CPU from the circuit board; the Kingston board is quite thin, and is very susceptible to damage. The DX4ODPR is tall enough that the heatsink will prohibit the use of a long card adapter in the 32-bit MCA slot, however. The SX version should accept a DX CPU without trouble. The downside: you are still using the slow (by comparison) planar memory. The lack of L2 cache also hinders performance. Still, it is a worthwhile upgrade if you can't find a MCMaster card.
The best performing CPU upgrade for the P70, by far, is the Kingston/AOX MCMaster 486 cards. My machine is currently running with a 25 MHz MCMaster that I managed to overclock to 32 MHz (yes, 32 MHz, not 33). See Fred Spencer's page for the info on how I accomplished this. Interestingly, the only CPU that would run reliably with a 32 MHz base clock on this board is an IBM (Cyrix) Blue Lightning DX4-100. An Intel DX2-66ODPR and an AMD 5x86-P75ADZ (with interposer) would both lock up after a random period of time running Windows. The problem did not appear related to cooling, so my conclusion is that some of the components on the card are slightly speed sensitive. I would be interested in hearing from anyone who has successfully used clock-multiplier CPU's on a true 33 MHz MCMaster. As mentioned above, I had to disable the planar memory in the MCMaster setup, as the machine would generally refuse to boot once the MCMaster took over. Disadvantages of the MCMaster upgrade: actually finding one, although they do seem to show up on eBay occasionally (surplus auctions are also a good source, it's a good bet that any university or government agency that bought a lot of PS/2's will have a few floating around); it takes the only 32-bit full-length MCA slot, which limits your options for adding other cards; and you need a CPU heatsink without a fan, because there simply isn't enough room for a fan. This shouldn't be a problem, as cooling seems to be adequate for a CPU with heatsink only.
My little P70 (a.k.a. "Punkzilla , fastest P70 in the west")
is currently running the MCMaster with the IBM DX4-100 CPU and 24 MB of
RAM, and runs PC-DOS 2000 and Windows for Workgroups 3.11 very well.
I have a 3Com Etherlink III network card (with the blue handle removed)
in the 16-bit slot, and it barely fits. I have the computer set up
to track Ham Radio satellites with Logsat
Professional 5.1, and it has come in handy as a portable satellite
tracking computer. I've got an AUI transciever for the network card
so I can connect to either coax or twisted-pair networks, and have used
the machine with both NETBUI (Windows) and Novell networks. The transciever
and the reference disk fit neatly inside the I/O cover, and all I need
is to run the reference disk to tell the network card which port I want
to use. These upgrade attempts have successfully turned what some
would consider a slow, old dinosaur into a very usable machine for service
work, portable word-processing or database functions, or just about any
other use you can come up with. Add a decent-sized SCSI hard disk
with a Future Domain MCS-700 controller (at the expense of losing the network
card), and it would probably even run Windows 95 comfortably (not that
I would actually WANT to...).
The P70 uses a DBA (direct bus attachment) ESDI hard disk, that is not
compatible with anything else except the PS/2 Models 55sx and 70 desktop
computers. Thus, a P70 with a 30 or 60 MB hard disk may be upgraded
with a 120 MB disk from a Model 70. Brad Parker and Thomas J. Watercott
have sent me info about an 8570-161 that has a 160 MB ESDI drive in it.
I have verified this through the EPRM, and Thomas relates that he has successfully
used this drive in a P70. This is probably the best source of replacement/upgrade
hard disks for the P70. As an aside, these drives are referred to
by some (and reported by a lot of diagnostic software) as "IDE".
In fact, they are and early form of Integrated
(Intelligent, Imbedded) Drive
(Device) Electronics (people
can't even agree on what the acronym actually stands for!) drives, but
don't bear much resemblance to IDE as we know it today. Peter Wendt
has this to say on the subject:
If you need more than 160 MB, you will have to go SCSI with an add-in card. I will not go into detail on this, since it is described nicely on Bob Eager's P70 page, using an IBM SCSI adapter. This is a 32-bit SCSI card, so if you use one of these, forget about putting in a RAM card or an MCMaster processor card. If you need your 32-bit MCA slot for other goodies, another SCSI adapter ideally suited to the P70 is the Future Domain MCS-700 (short-card version). The MCS-700 is a 16-bit SCSI-2 card capable of 10 MB/s transfers, has a standard SCSI-2 external connector, and also has a power connector on board to supply power to the hard drive. It is not, however, a busmaster; it has no onboard processor and uses PIO mode for data transfers. It is still a very nice controller; I used one in my Model 95 as a secondary controller, and it ran a CD recorder and two hard disks without a hiccup. Louis Ohland has a nice page on the IBM OEM version of the MCS-700. The IBM version (found mainly in the Model 76s/77s "Lacuna" machines) should be equally usable in the P70; however, it lacks the onboard power connector. The solder points for the connector are still there, so it is easy to add in if you are comfortable with a soldering iron. A ROM BIOS upgrade (version 3.61) was available from Adaptec at no charge if your MCS-700 exhibits the "hard disk LED not functioning" problem, or you can download the code from my IBM ROMS page and burn your own.
Return to Table of Contents
|The P75 can be described as a P70 on steroids. It boasts a redesigned
planar with a i486DX-33 MHz CPU on a plug-in processor complex, as
well as an imbedded SCSI subsystem to replace the P70's direct ESDI attatchment
bus. Internal SCSI hard disk options were 160 MB and 400 MB.
The SCSI subsystem may leave something to be desired, however; one user
that I know of has installed a Future
Domain MCS-700 SCSI-2 controller in his P75 and reports better performance.
UPDATE, 1/21/2001; Casolai
sent along this P75 SCSI info:
You can see Casolai's web site at http://www.angelfire.com/hi/casolai/index.html. Beware of multimedia content! :-) 6/13/2001 - link is broken. Cas, update me!
Wendt provides the following info on the video subsystem:
System memory was 8 MB standard, upgradable to 16 MB on the planar. In addition to the I/O connections found on the P70, the P75 I/O panel has an RS/6000 style external SCSI connector. The machine also has one more of each (16- and 32-bit) MCA expansion slots, for a total of four MCA slots. One 32-bit slot is an AVE slot. The extra slots make the case slightly deeper than the P70. I've been told that the P75 was marketed as a "portable server"; based on my experiences, it could certianly fulfill that function quite well.
P75 Model Number Breakdown
|8573-161||160 MB SCSI hard disk|
|8573-401||400 MB SCSI hard disk|
Return to Table of Contents
|Unfortunatly, there isn't a lot the average person can do to upgrade
the CPU in the P75. The fact that the processor complex is on a plugin
card suggests that IBM may have intended it to be upgradable, but no upgrade
is even rumored to exist. The CPU itself is soldered to the circuit
board, so popping in an Overdrive isn't an option. However, there
are two possibilities for those willing to go to the time and trouble.
The first, and most viable option for most people, would be to find one of the Kingston MCMaster 33 MHz cards. There were versions of this card with clock-multiplied CPU's, but they seem to be extremely rare. More common are the 486DX-33 and 486DX/SX-25 versions of the card, which should be able to support faster CPU's. A 33 MHz base card with a DX2-66 or faster CPU fitted should perform nicely. The advantage of this approach is that you can also easily upgrade memory beyond the 16 MB limit of the planar, with better performance than adding a memory card to the MCA bus. The disadvantages are the loss of one 32-bit expansion slot to the upgrade card, and the height limitation on the CPU/heatsink combination imposed by the limited space inside the P75. The hardest part would be locating one of these cards, but one place I have seen them for sale fairly often is eBay.
The second approach is the more difficult, and should only be attempted by experienced hardware hackers. I have personally upgraded four P75's by desoldering the CPU from the processor board and installing a 168-pin PGA socket. This SHOULD NOT be attempted unless you have a lot of soldering experience AND access to professional desoldering equipment; it is very easy to damage the copper traces on the board. That being said, the results were well worth it. The two 'local' machines ran very well with an AMD 5x86-133 CPU chip (with 3.45-volt interposer). Various other "turbochip" upgrades should work equally well; one of the two machines currently has a i486DX4-100 Overdrive CPU installed, running Windows NT 4.0 very nicely. I don't think the P75 BIOS supports write-back cache, so any "turbochip" should have the capability of running in write-through mode for best results. The disadvantage to this method, other than the risk of destroying your CPU board, is the height limitation imposed by the cramped quarters inside the P75. Some of the CPU/interposer/heatsink combinations are quite tall, and will limit the use of the two adjacent MCA slots to short cards only; even at that, some 32-bit "short" cards may still be too long.
Photos of one completed CPU upgrade can be found on Tam Pham's site here. While you're there, be sure and check out the rest of Tam's fine collection of PS/2 information!
On the subject of write-back cache, Brad Root has this to say:
Return to Table of Contents
|It's about that time, folks. It is the second week of June, 2001,
and the battery in my P75 (manufactured May, 1991) has failed. Unfortunatly,
this is not an off-the-shelf replacement. The P75 uses a custom battery
pack consisting of two Panasonic
CR2477 1000 mAh cells in series, soldered to a small circuit board with
a plastic cover. Bob Eager has a nice pic of the pack on his web
for Louis Ohland's pic of the circuit board with the cover snapped off.
The problem is, the CR2477 (with solder tabs) is not something you can just run down to your local battery shop and grab off the shelf. They are somewhat more expensive than a 'standard' CMOS battery, and must be ordered from a parts distributor. Assuming that there is nothing unusual about the CMOS memory/clock in the P75 (as compared to the Model 95, etc., that use the more standard and common CR2032 220 mAh cell for CMOS retention), there is no reason that I can see for using the CR2477 other than greatly extended CMOS battery life. Two CR2032 coin cell holders liberated from 'retired' clone boards will fit nicely inside the battery pack shell. They can be fixed in place with hot melt glue or epoxy and soldered with short jumper wires to the pads that the CR2477s connected to. Two CR2032s cost me USD$1.36 each at the gorcery store today. Much better than the $4-$9 each (plus shipping) that Bob found for the CR2477! Top it off with a small chunk of non-conductive foam or bubble wrap to keep things secure inside the battery case, and the result is an inexpensive replacement solution that looks stock from the outside. The CR2032 only has about 22% of the capacity of the CR2477, so expect to replace them about every 2.5 years instead of every 10 years.
Return to Table of Contents
|The P70 8573-121/MCMaster 486 and P75 8573-401 that I have tested both pass as being Y2K compliant under IBM PC-DOS 2000. PC-DOS 2000 (or PC-DOS 7.0 with Y2K patch) handles the rollover to 2000 and properly updates the century byte in CMOS. Both machines fail the year 2000 progression test when booted under MS-DOS 6.22, requiring manual intervention to update the century byte. The reboot test after manually setting the date to 1-1-2000 passes. With the proper operating system or software patches to ensure CMOS rollover, both machines are 100% Y2K compliant. Testing was done using YMARK2000, from NSTL. A very good technical discussion of Y2K compliance from a hardware standpoint, courtesy of Peter Wendt, can be found HERE. IBM has a very nice white paper on Y2K, available HERE.|
Return to Table of Contents
|I would like to thank the following people, without whom this page would not have been possible: Mark Stewart (Mr. Grench), for the loan of his beloved P75; Martin Adams, for the info on the Microram 386 RAM card; the "MCA Mafia" and the denizens of comp.sys.ibm.ps2.hardware, especially Peter Wendt for his boundless knowledge of everything PS/2 and Louis Ohland for kicking me in the rear and getting me working on this page; IBM Corporation for giving me the tools to understand these machines, years before I ever actually owned one; and finally, the surplus property divisions of the State of Nebraska and the University of Nebraska-Lincoln, for selling all this neat stuff for really cheap at auctions.|
Return to Table of Contents
(click on the banner)
All Articles and Pages are ©
Copyright 2003, all rights reserved, by Jim
Shorney (except where noted).
IBM, Microchannel, and PS/2 are registered Trademarks of IBM Corporation.
All other Trademarks and Copyrights are those of their respective owners.
All information contained herein is accurate to the best of my knowledge; however, the author assumes no liability for lost sanity, time, productivity, data, or property resulting from the use of information contained in these pages.
to Ebay (or other online auction) Sellers:
You may link to my site, or include my images in your auction description PROVIDED that you credit this web page as the source; BUT you may NOT print these pages for inclusion with an auction item, or 'Copy and Paste' any text or HTML code contained in this page or other pages hosted by or linked from this web site without prior written consent of the author(s). I (James R. Shorney) retain copyright to original text, images, and layout of my site; all non-original material retains Copyrights of the original owner(s).
Revised January 3, 2007