Bill Beech and recent work on Intel 80/10 & ISIS system

Last edit Nov 7 2011 updates Dec 24 2016 (c) Herb Johnson


[Bill Beech NJ7P contacted me in June 2009 to exchange manuals and resources about our Intel System 310 Multibus systems. We've talked about a number of Multibus 8080 and 8088 subjects, an the Multibus systems he's rebuilding. See the linked Web for links to all those activities, and about Bill and his other work on the Web. IN addition, Bill also produced many Web pages on his own Web site, look there about his 80/10 and related systems and the software tools he's built.

Bill Beech acquired an Intel 320 system case from me; a several-slot Multibus backplane, power supply, and simple run/reset/power front panel. There's room for small hard drives or 5.25" floppy drives. He also bought one of my 386/25 CPU boards which needs work. Bill also has a 386/38 with slightly different ROMS and RAM setup. Bill has a stack of Multibus cards, that he's now working through to see their condition and what can be done with them.

Bill and I talked through this from the fall of 2011. - My half of the conversation is in italics, some embedded comments of mine are in []'s. - Herb Johnson


386 boards

A quick compare of your 386/25 board with my 386/38 board. I have different ROMS - 459494 (c) 1986. Yours are 452122 (2) 1987. A few PALs have different numbers. I guess I have a 386DX-20, you have 386DX-16's. And I have 16 MB of RAM vice your 4 MB. Interesting your ROMs are a year newer....the 30 and 22 differ in the existence of two jumpers, which I surmised had to do with the size of the on-board RAM.

386/38 CPU

Here is a run of the 320 chassis with only my 386/38 CPU board:

When my board comes up, it sends a character that looks like a slashed 0 on its side - maybe 30 or more of them. This may be it prompting for a "u" to set the baud rate. I reread the XENIX install on the 286 and it talks about a "u" to get the SDT to run. Then it clears the screen and sends the listing [below]:

Copyright 1986, Intel Corporation

  Processor Subsystem Tests ......................................... PASS
  Mass Storage Subsystem Tests ...................................... FAIL
  Memory Subsystem Tests ............................................ PASS

SCT FAILED ... entering System Debug Monitor

For Help Type:  "wh"

Interrupt 3 at F016:06BB
.[end of output]

The board does not seem to receive characters from the terminal. The RS-232 LED box shows data when I transmit, [but] nothing is being received by the 8251 on the RxD line. System is still running - green led indicates ALE transitions continue. Suspect maybe a misconnected device popped the RS-232 receiver on the board. That is easily located - but I must do a little work with an ohm meter to find the IC section in question.

Later on, Bill said: I was messing with my iSBC 386/38 and put it back in the 310. It is working correctly, now. I can run the monitor. The SDT-386 looks for a lot more memory than the 286. - Bill


[Meanwhile your] 386/22 board did not produce output, so there is a bit more to do. It did have a flashing green LED, which I remember means an error during the initial boot sequence. According to the iSBC-386-20P HW manual (close but no cigar on the 22 or 30 - 22 & 30 use same board layout except for larger RAM on 30), the flashing of LED 4 means a parity error. That might stop the ROM program from ever getting to where it sends data.

I removed the [memory] daughter card but that did not clear error. That means a problem on the soldered in RAM chips or support circuitry... That will be more tricky.

These [INtel] machines all came up at 9600 8 N 1 to start with.

[At this point] I think it may be time to pull all the [other 286 and RAM memory] boards out and test each CPU board to see if it works in the monitor mode. Then I can play with the memory boards and controllers. It would be nice to boot up one of the systems. I have needed to do this for a long time. The 320 chassis, with the board extender, makes a good platform for this test.

286 and RAM boards testing

Except for the cable from the back cover to the cpu board, I used the same setup to test all of my iSBC-286/10(*) boards. Three of them work, 5 fail. [A parity failure] is the same problem I have with several of the 286/10 boards.

[The count of five CPU failures includes] a pair of APEX boards, which was Intel's first step into multiprocessors in a complete system. That machine had 2 286 processors and spread tasks between them. I don't have the version of XENIX that supported this.

[Herb comments: "That's a lot of CPU boards! Nice to have duplicates. You are in good position to diagnose and repair them when you have several."]

Bill says: Worse gets to worse, I put a good one in one machine, a bad one in another, and use the scope to compare the signals. I need to find another multibus extender board. The one I have now also has a switch to kill all multibus power on the tested board. Comes in handy, sometimes.

Bill continues: The memory boards [I have] are all of similar design [iSBC-010CX, 010EX, 012CX, 020CX], and have a single red led to indicate a parity error. I even have some CX boards that detect and correct errors. Plenty of socketed RAM to make a bunch of good boards.

[Herb notes: A Web search says "The iSBC 012B is a 512K byte Multibus-compatible MOS dynamic RAM memory board". It's not particularly fast - 550ns total cycle, 350ns read cycle. Looks like there's four banks of 18 chips arranged low byte/high byte, with parity. (Yup, that is the previous generation RAM card. - Bill.) The 010 and 020 has 256K HM50256-20 or similar -12 or -15 chips, the 012 use 64K 4164-12 chips. Some of these may also have ECC (extra RAM for one-bit correction). - Herb]

SInce there is NO RAM on the 286/10(*) boards, I need to be sure I have reliable multibus RAM to test them. The two RAM boards out of my "working" 310 occasionally show a parity error. Not sure what the deal is there.

RAM testing

I have been testing RAM cards. All my cards are the same layout, some used 4164's and some use 41256's. So I have a mix of 512KB, 102KB and 2048KB. Jumper pins and discrete ICs are the same, so one manual would cover them all. Unfortunately, they are the next generation past the 012B that we have a manual for. [Later he found a manual, see below.]

"All the RAM boards are CX types, which means they are error detecting and correcting. That is why they have 24 extra RAM chips on a full board. And I was surprised at how cool the RAM boards ran. I have used a heat gun for shrink tubing to warm things up in the past." [We had talked about cooling or freezing the boards to make any "soft errors" go away. - Herb]

I had 2 each 010CX boards in the "working" 310. I have found the jumpers for the base address and the board size. I have tested the 4 each 010CX boards I have and all check fine. The ROM tests for RAM in each 1 megabyte block of the 16 MB address space. I get the iSDM monitor with each board, so some amount of testing was passed.

I have a stack of 012CX and 020CX boards, as well as some 010EX's. I have set the jumpers on the 020CX boards exactly like the 010CX, and it passed. It would only show in the first MB of memory. I will give an 020CX a try and see if it sees it in two places. I have compared the chips on the 012CX, 010CX and 020CX and they have the identical chips with the exception of the RAMs and one PAL that is used to set the board size. I may pull the PALs and see if I can read them. I have scanned the boards so I will attach the scans for you to see. The scanner scanned a little less than the full board in each case, but the pertinent data is there.

Later, Bill added: "I also found 145158-001 - the HW reference manual for the iSBC 012CX. The jumpers and such are the same for the 010CX and 020CX boards. Intel used a 256kbit RAM chip in place of the 4164 chips on the 028CX, 056CX and 012CX boards. SO we have the jumpers for all the RAM boards I have."


By late OCt Bill said: [After resolving the DRAM cards,] then I will revisit the iSBC-286/10(*) cards and see if I can get more of them to work. I suspect that the problems may just be configuration. I have been marking and dating the tested boards with results as I go. Finally, I will revisit the 386 boards.

Days later, Bill said: "I have a mix of 286/10 and 286/10A boards. I have 2 each 286/10 boards that work fine. 3 each 286/10's that all fail with memory errors. All the 286/10A's fail. That may provide information in itself. All success to this point is in the 310 chassis.

Herb said: So now it's a matter of testing and refining the RAM cards with the "good" 286 or 386 CPU's; and testing and refining the "bad" 286 cards with the "good" RAM. One will help with the other. Then with a good stack of RAM and 286's, and at least one 386, you can set a known-good system with backup boards, to test the drive controller boards you have. - herb

Bill replied: With either of the good 286/10's any of the RAM boards work correctly. No parity errors and the monitor works correctly. 896KB of the RAM board passes the memory check (unless it is a 012CX, then 512KB passes)." Bill subsequently confirmed: "I have tested all the RAM boards with both good 286/10's. So I believe I have that stage complete. I retested the 386/38 this morning, and it still is working. Going to wait to declare it good."

Disk controllers

As I see it, the 310 was never proved operational beyond CPU and RAM. So I now have 2 each 286/10 CPUs and about 15 known good RAM cards all from the CX family. I have a valid RAM HW reference manual, so I can configure multiple RAM cards into one system. The next step is to get disk controllers to work.

So [for floppy controllers] have 3 each 214's, 1 each 221 (which is an upgraded 214), and 2 each 215 with 217/218s. Condition of all is unknown. The floppy connector on the 218s is 50-pin - for the 8-inch drives. The floppy connector on the 214/221s is 34-pin for 5-inch drives. I need to build an adapter for the 218's.

The only OS that will run from floppy, is MSDOS. It would require I build a BIOS ROM to convert BIOS interrupts to the chips on the 286/10 and FDC. This would be starting from scratch and involves making guesses about addressing on the 215G and 286/10. I have a HW reference manual for the 218 FDC and for the 208 FDC. Too many variables here without more manuals.

The next option may be to make a set of load floppies for Xenix-286 from the code on bitsavers. I can prove the floppy and drivers by being able to boot the floppy. Then I can attack the hard disk and controller.

Sound like a plan? This will give me a bunch of tested boards I can use to populate both chassis.

[Herb mentioned the monitor supports some kind of testing.] The tests for the [Winchester hard drive in the monitor] are good. It does NOT test the floppy. My wini fails on all the controllers I have tried. The message is "controller board failed" which may mean I have a jumper issue or may have a memory conflict. The original 310 had 2 each 012CX memory boards - total 1024KB. I have a 020CX in there, now, total of 2048KB. Will put a 010CX and retest the controllers [in that chassis].

[Herb suggested the 8080 cards may be too slow to support double density data rates.] Not an issue. The iSBC 208 is a bus master - it takes the bus and uses DMA to transfer data directly to RAM. Then it is supposed to turn control back to the iSBC 80/10. That is where the system is failing, now.

Nov 2011 "inventory":

Bill added in early Nov 2011: Currently, the 3-slot chassis contains the 80/10, 064 and the 208. Meanhwile, the 310 [system chassis] never used a 208 controller-just 214/215/221s. The failure is my trying to cobble the 208 and the 80/10 and make them work together. I had to make one wire change on the back of the multibus frame. There is probably another or a jumper change still missing.... I know for sure no actual DMA transfer takes place as I don't get any data in the buffer."

"Because the footprint of the 310 is smaller than the 320 and the cards are easier to install/pull in the 310 than the 320. Remember, the multibus cage is set about 2-3 inches in in the 320. The cage is even with the back in the 310. Its a matter of saving skin on my hands and not letting all the blood thinners out." - Bill

[So I'll build] one three-card for the CPM80 and the 310 for the CPM86. Once I get some controllers working - looks like I may have found a good 214 - we will move on to the 320.

Here is a run down of my boards by function, if you are keeping score.

iSBC 80/10 - 8080 2MHz cpu card.
iSBC 86/12A - 8086 5 MHz cpu card
iSBC 286/10 - 80286 8 MHz cpu card
iSBC 386/38 - 80386 20 MHz cpu card
iSBC 064 - 64KB memory card
iSBC 012CX - 512 KB ECC memory card
iSBC 010CX - 1024 KB ECC memory card
iSBC 020CX - 2048 KB ECC memory card
iSBC 208 - DSDD FDC card for 8/5 inch floppies - i8272 based
iSBC 214 - ST-506 HD, FDC for 5 inch floppies - WD2791 based, QIC-02 Tape
iSBC 215 - ST-506 HD controller
iSBX 217 - QIC-02 Tape controller
iSBC 218 - FDC for 5/8 inch floppies - i8272 based
iSBC 221 - ST-506 HD, FDC for 5 inch floppies - WD2791 based, QIC-02 Tape (advanced 214)

Currently, the 3-slot chassis contains the 80/10, 064 and the 208. The 310 chassis contains a 286/10, 010CX and 214. [The 320 is empty.]

[So Bill has a three-slot chassis, suitable for simple systems and testing like the 80/10 stuff; a System 310 chassis (from me) with many slots but it's too "deep" and hard to use physically; and a System 320 chassis which had a small system but which he's using for most of his testing. - Herb]

related notes

Side issue: Operating systems notes

Herb, Last night I did some thinking about OS's and testing. Here is a table of the available OSs for testing.

No    OS            Density   Size    CPU           Controller     HD Req?   EOC?
1    CPM80         SD/DD     8/5     80            208/214/218    No        Easy
2    ISIS-II       SD        8       80            2-BD FM        No        Hard
3    ISIS-IV       DD        8       80            2-BD M2FM      No        Hard
4    CPM86         SD/DD     8/5     86/286/386    208/214/218    No        Easy
5    MSDOS         DD        8/5     86/286/386    208/214/218    No        Hard
6    iRMX86        DD        5       86/286/386    214/218        Yes       Hard
7    Xenix286      DD        5       86/286/386    214/218        Yes       Hard

What will drive testing are availability, familiarity, and ease of configuration (EOC). If it requires a HD, then it is not a candidate for initial testing. If we don't have a hardware reference manual, then it is not a candidate for initial testing.

Based on these criteria, we are left with CPM80, CPM86, 80/10, [86/12A, 286/10, 386/38], 208 and 218. I say CPM80/86 is easy because I am very familiar with both, I have source code for both, and configuration only requires writing a BIOS to match the hardware. Both will run on a 5-inch disk. I place the 286 and 386 in brackets because I don't have a hardware reference manual for either but suspect I can get them to work with the 208/218 controllers. My 86/12A board need a 22.1184 MHz crystal replaced - broken wire.

5-inch drives seem to be the order of the day. I am waiting on the 200 DD 5-inch disks I [recently bought on eBay.]

The testing built into the boot ROMs on the 286 and 386 does a good job on the wini but do next to nothing on the floppy. I was able, at some point this last week, to have the floppy disk cleaning routine select the floppy drive and do its thing. I will attempt to get back to that point.

The three-board multibus cage I am using for the 80/10 adventure is also working well. I am able to read address marks off of 5-inch disks. I still have a problem with DMA grabbing the bus and not returning it. Thus, I have 3 workable chassis/power supply combinations.

I have an operational Kaypro-4 with the Legacy 8088 board on which I can develop CPM80 and CPM86 software and generate bootable floppies. It may be time to get it back out, and set it up as part of this operation.

So I will work on the iSBC 80/10, iSBC 064, and iSBC 208 in the 3-card chassis with CPM-80. I will work on the iSBC 286/10, iSBC 0XXCX, iSBC215 with iSBX 218 or iSBC 214 or iSBC 221 in the 310 chassis and CPM86. I will place the 386/38 in the 320 chassis and not do anything with it, for the moment.

So, now to reconfigure my room to get this done while I await your thoughts? - Bill

[I told him to proceed with CP/M-80 and CP/M-86. They are very simple, well-documented and support almost the same BIOS. Since all but his model 218 controller will support 5.25-inch drives, I suggested putting the 218 models aside until the other floppy controllers are resolved. Bill noted the floppy controllers are DMA so they are fast and support double-density. - Herb]

Herb: You mentioned using your Kaypro to develop CP/M bootables. That's OK, of course it's 5.25-inch. Those drives are supported on most of your controllers. For 8086 code, you'd have to assemble that elsewhere. But you can still create the disks and copy files onto them from the Kaypro.

Bill replied: "Not true. My Kaypro IV has a Legacy 8088 card, as I mentioned before, and runs CPM80 and MSDOS 2.11. It contains a Z80 and an 8088 processor with 256KB of memory. It will probably run CPM86 native, as well." - Bill

Herb: Given all that fuss to accommodate 5.25 inch drives on the 218, you might put those aside and work on all the others. When they are mostly resolved, you'll have working disks you can test with the 218's.

Bill: "My thoughts, exactly.So we are back to the 8080 machine booting CPM80 and getting the 310 to boot CPM86. At least we have a plan..."

[I'll continue my discussion of Bill's 80/10 and CP/M, on my Web page where I have his earlier notes on his 80/10 system. - Herb]

side issue: INtel's iSDM monitor program

I am playing with the iSDM monitor to display memory so I can figure out the jumper settings on these boards. I have forgotten all I knew about it from many years ago when we discussed this. [Later, Bill noted "every one of my complete RAM cards tests out in the 310 with a good [386/38] processor board." That explained why he was looking at the jumper settings as a likely cause. - Herb]

I have found the description of each [monitor] command letter, but not the format of specific commands like D and M. Got more work to do on this. If I need to, I may write a short memory test program and use iSDM to load it, run it, and test out the other memory boards. I have a good algorithm in the 8080 monitor. The iSDM [monitor] is almost identical between systems. It was developed for the 8086 and will work on almost all of the Intel multibus boards, including the 88/45, 188/48 and all the CPU boards. It was distributed as ISIS-II 8-inch disks, iRMX 5.25-inch disks, and as 27128 EPROMS for a lot of the multibus boards. Wish I could find a complete distribution of that!

[But I] have your ROM and my ROM images here. I am still looking for my disassembly of the 286/10 ROMs. I had it almost to the point where it could be assembled and put into EPROMs. But then, these ROMs are all PLM86 code to start with. I need to continue with the decompiler.

[Later] I found the SDK86 serial monitor listing last night. I OCR'ed it, have the source, which compiles with 2 warnings. It was originally done on PLM86 V1.1, and the DOS version I have appears to be 3.4 (X304). I will document all the commands as I figure out the syntax. We need that for anyone dealing with an Intel CPU card with a proper Intel boot ROM. I believe the SDK86 monitor came first. iSDM is a further development of SDK86. Unfortunately, they removed the hex load and store commands in iSDM. - Bill Beech

Herb adds in 2016:I checked Bill's Web pages in response to an inquiry by Richard Main. Here's Bill's results as of 2016. Some documents were obtained by Bill from - Herb

sdk-86 mcs-86 system design kit monitor listings

SDK-86 user's manual

On Bill's monitor Web page: "I [Beech] have OCRed and edited the PLM-86 file for the SDK-86 monitor. This monitor will build with the PLM-86 tools for DOS." Here's the ZIP file of that PL/M-86 monitor and compilation.

If those links fail, check my "where's Bill?" Web page for some suggestions.

Herb in 2016: Richard Main found a version of the iSDM ISIS-ii distribution and built a iSDM Monitor for an iSBC 88/25 CPU board. He asked for more information about about "the iSDM Monitor", when he found the above references on this Web page. I added the links to Bill's results and I'll update this page if Richard reports results and details. Also check Richard Main's Intel MDS-content Web site, as he and others did enormous work in 2015-16 on recovering Intel development tools. - Herb

side issue: on line file archives

I need to get all the Intel disk images I can find converted to straight ISIS or xenix images so I can pull all the code out of them. I have to believe we will find some more goodies buried in those images. The stuff off of Al's [] site gives us a complete iRMX86 and basic and extended Xenix3. Also, the full networking package - iNA960. I just need to get the hardware working. I see mention of an MSDOS 2.15 for the 310, as well. With a proper BIOS ROM, I bet a PC MSDOS disk would boot. Been looking into that, as well. Been looking at too many things to get anything really done! - Bill

Herb in 2016: Progress was made by Bill Beech and others in 2015-16. Check the notes above on iSDM monitor for some Web links. - Herb

Contact information:

Herb Johnson
New Jersey, USA
To email @ me, see see my ordering Web page.

Copyright © 2016 Herb Johnson