PDA

View Full Version : 3 CPUs plugged into a 386 Motherboard


DelTron
October 21st, 2007, 08:46 AM
Hello I'm new to this forum.

I'd like to start by saying:

I have a 386-40 that has 3 cpu's plugged in. One is an am386-40 embedded. The next is a Cyrix Fasmath-40mhz 387 CPU. The next is a second am386-40. I installed the two non-embedded ones.

Both am386-40's are running hot. The cyrix 387 is cold. I Did not run DooM or DooM II to test the cyrix 387 temperature theory, but I can. BIOS generated startup screens and windows 95c both detect the 387, but this is a side issue.

Do you think this PC supports multiprocessing?
Windows displays "80386, 80387" under my computer->properties

My system info is as follows:

• am386-40
• am386-40
• 387-40

• 20mb RAM
• baby (?) AT board (very tiny appearance)
• 1mb trident video
• creative awe64 sound
• 400mb hdd
• 8x cd-rom
• 3.5" floppy drive
• basic peripherals - mouse/keyboard/speakers/monitor
• Turbo Button feature
• AMI BIOS 1993

thanks for your comments on this interesting topic

mbbrutman
October 21st, 2007, 09:00 AM
It is highly unlikely that your machine is using both of the primary processors (The AMD 386-40) together. That would have been a very very high end and specialized motherboard back in 1993.

I would remove the extra 386-40 immediately - it probably does not belong there, and it is causing extra current draw through the motherboard.

Vlad
October 21st, 2007, 09:49 AM
I doubt its capable of SMP. The earliest dual socket I've seen was a 486, and even then it was pretty hit or miss. I really doubt its multiprocessing, none of the OS's of that era would have supported it anyway.

evildragon
October 21st, 2007, 10:03 AM
I doubt its capable of SMP. The earliest dual socket I've seen was a 486, and even then it was pretty hit or miss. I really doubt its multiprocessing, none of the OS's of that era would have supported it anyway.
NT 3.1 supports SMP 386's ;) ..

NT 3.5 however does not.

Terry Yager
October 21st, 2007, 11:08 AM
The 'extra' socket may be intended for an upgrade cpu, such as the DLC. When you pop a chip into the socket it should disable the onboard 386, but a standard 386 probably doesn't. There were also other processors available in the '386 era, that would work along with the main CPU, like I/O speedups, NPUs, etc. One type was from Weitek, but I don't recall any of the others...anyone else? That would be my best guess. Look in your BIOS setup and see if there's a setting to en/dis-able the Weitek processor. The NPU should be just fine though, but as others have said, remove the socketed CPU most ricky-tick before it destroys something.

Here's a wiki to get ya started: http://en.wikipedia.org/wiki/Weitek

--T

DelTron
October 21st, 2007, 11:55 AM
Thanks guys, I will remove the extra CPU.

After I put it in, it has had a couple days of uptime, but I agree it cannot be great for its health, as there must be a great deal of current draw on the bus. This isnt a bad site afterall.

DelTron
October 21st, 2007, 12:30 PM
additionally I'd like to note that it was indeed for a weitek co-cpu. I, for some reason believed that only one type of CPU could fit into its respective socket. I hope the extra cpu isnt burnt out :blush:

However one has to be curious if both NPU's could be enabled.

JDT
October 21st, 2007, 03:44 PM
This isnt a bad site afterall.


... understatment of the millenia

IBMMuseum
October 21st, 2007, 05:15 PM
additionally I'd like to note that it was indeed for a weitek co-cpu. I, for some reason believed that only one type of CPU could fit into its respective socket. I hope the extra cpu isnt burnt out :blush:

However one has to be curious if both NPU's could be enabled.

What manufacturers did sometimes was put the 387DX pins on the inner pins of a socket, Weitek coprocessor on the outer pins (due to the pin arrangements one would not use the pins of the other; It also effectively prevents the two NPUs to be used at the same time). A 386DX would however use the same pin areas in a socket, and I'm pretty sure it is a different pinout for those areas. There is nothing preventing a Weitek and 387DX NPU from functioning on the motherboard, since the means to address them, and quite a bit of the functionality, is different.

Terry Yager
October 21st, 2007, 05:22 PM
However one has to be curious if both NPU's could be enabled.

Can't help ya there. I've never even seen one in the wild, I've only heard/read of them. OTOH, a '386DX & Cyrix FastMath combo should be just about all you need for vintage computin'. I can't see where adding a second NPU would gain any noticeable performance.

--T

Anonymous Freak
October 21st, 2007, 05:48 PM
I don't recall which variation of UNIX it was, but back in high school, we had a "mini-supercomputer" that had 8 386-387 pair's in it. (And an incredible 24 MB of RAM, and over 1.5 GB of total hard drive space! :-p)

Anonymous Coward
October 21st, 2007, 07:13 PM
I highly doubt that you plugged your 386 into the weitek socket by accident. The Weitek sockets are slightly smaller, and it is not physically possible to insert a 386 chip in there. I have a 386 motherboard that is similar to yours. It has an embedded am386 40MHz, an FPU socket, and an extra 386/486DLC ceramic socket. So far, I have not been able to get my extra CPU socket to work. Unlike yours, my extra CPU does not get hot at all...I tried both 386 and 486DLC. My guess is that the CPU in the socket cannot function while the embedded one is present. I think that the manufacturer was just careless and soldered on the extra socket when it wasn't needed or usable. You have to keep in mind that these motherboards were made by el cheapo companies long after the glory days of the 386.

Concerning the Weitek chips, I have found that just because your BIOS has an option for it doesn't mean that your motherboard is able to accept one. I have several 386 and 486 boards with Weitek settings in the BIOS and no way of physically plugging it into the motherboard. There is no extra row of pins on the 387 socket, nor any additional solder pads on the motherboard. Perhaps there was some kind of adapter available. Also, it is very possible to have both Weitek and 387 FPUs installed at the same time. I found a nice italian website a few months back with a guy who did very thorough testing. See if you can find it on this page:

http://www.cpu-museo.it/

You should be able to find two setups on the page. One has a combination rapidCAD and 3167, and the other is part of a motherboard from an HP 386 machine that has 386, 385, 387 AND 3167.

...and finally, there is one other chip that can use the Weitek socket. It is called the Cyrix EMC.

DelTron
October 21st, 2007, 07:49 PM
Thanks ananymous Coward. You must type fast because login times out a bit too soon for me. The weitek 387 is indeed 121 pins, not 132 like the 386. http://www.cpu-world.com/CPUs/3167/index.html. Confirms how the 387 works along side the weitek, which enhances CPU performance even further.

Yes, I think the mobo manufacturer was careless when placing the extra socket on the board, as 386's were budget PC's with budget components in 1993.

additionally I'd like to comment on Ananymous freak's post, that I like this machine he's talking about. when was that?

im willing to research exactly what this extra socket is, and i'll keep you guys posted. Thanks for the information, the guru here are very resourceful.

mbbrutman
October 21st, 2007, 08:47 PM
Two math coprocessors is not faster than one.

The main CPU is going to talk to exactly one math coprocessor. The other one is surely being disabled. If you read the Intel specs for these things, it's not possible to have two active at the same time working with one processor.

DelTron
October 26th, 2007, 01:59 AM
http://i173.photobucket.com/albums/w60/deltron61/386d.jpg?t=1193392725

Anonymous Freak
October 26th, 2007, 09:20 AM
http://i173.photobucket.com/albums/w60/deltron61/386d.jpg?t=1193392725

That is interesting. It's definitely not a 486 socket, definitely a 386-style socket. Yet the board includes a 40 MHz 386 soldered, and no faster 386s were ever available, so it's not like it was meant to be an upgrade socket.

You mention that it runs power, but "doesn't seem to do anything". Have you tried running some basic benchmarks with the stock chip, and with, say, a SLOWER 386 in the socket to see if it will use the socket if you have a chip in there?

Unknown_K
October 26th, 2007, 10:24 AM
That motherboard was designed so that they could either solder on a CPU (permanent) or leave it off and use the processor of choice.

When it was first introduced they probably use whatever processor the customer wanted, at the end of its life the 40mhz part was probably the only one available and they just had it surface mounted by defaut.

Anonymous Freak
October 26th, 2007, 12:23 PM
That motherboard was designed so that they could either solder on a CPU (permanent) or leave it off and use the processor of choice.

When it was first introduced they probably use whatever processor the customer wanted, at the end of its life the 40mhz part was probably the only one available and they just had it surface mounted by defaut.

Since they had to use the effort of surface mounting the processor at time of motherboard manufacture, why wouldn't they just leave off the socket?

evildragon
October 26th, 2007, 12:36 PM
maybe incase you didn't want an AMD? just a guess.

nige the hippy
October 26th, 2007, 01:40 PM
Please also remember that at that time there was an awful lot of chip-supply-paranoia, it was only a few years earlier that IBM had bought up the whole world's supply of TTL chips, taking the price of e.g. a 7400 from about 11p to over a pound overnight. People were designing for multiple suppliers of key components just so the products could continue to be manufactured if supply of the preferred component dried up. it's true!!!!

barythrin
October 26th, 2007, 02:01 PM
I could be wrong with my ever-fading memory however IIRC it wasn't until the PII that we had different sockets for different manufacturers, so the 386 would be able to have an Intel or AMD taking the same pins so I don't think it's a manufacturer choice feature.. not that I know what it IS for.

Do you know the make and model of the motherboard? Maybe we could find the manual somewhere just for kicks and figure out why it's really there.

Anonymous Freak
October 26th, 2007, 03:21 PM
I could be wrong with my ever-fading memory however IIRC it wasn't until the PII that we had different sockets for different manufacturers, so the 386 would be able to have an Intel or AMD taking the same pins so I don't think it's a manufacturer choice feature.. not that I know what it IS for.

Do you know the make and model of the motherboard? Maybe we could find the manual somewhere just for kicks and figure out why it's really there.

Correct, Intel tried multiple times to lock their sockets from competition unsuccessfully. With Slot-1, they finally managed to patent the slot, thereby enforcing no competition. (Although AMD did do "Slot A"...)

The 386 was the last processor that was officially multi-sourced from Intel. AKA: AMD was an official Intel-approved manufacturer of the 386. Starting with the 486, all non-Intel 486's were "unofficial clones". And Intel could refuse warranty service on an Intel-manufactured motherboard if you had a non-Intel 486 in it.

In the early days, Intel was primarily a design house that happened to own a fab or two (ironically, AMD turned from a fab-only to a combo, to a similar structure to Intel pre-486, and has just in the past year or two become like Intel in being completely vertical.) Around the 486 timeframe, Intel turned into a complete vertical organization.

rmay635703
February 20th, 2008, 04:04 PM
I don't recall which variation of UNIX it was, but back in high school, we had a "mini-supercomputer" that had 8 386-387 pair's in it. (And an incredible 24 MB of RAM, and over 1.5 GB of total hard drive space! :-p)

Generally ZENIX would have been the OS, I know for certain compaq and HP had dually 386DX processor boxes. I also know of one 8 banger 486DX I missed due to distance. These oldies though didn't always work like a normal PC at all and many times required a terminal or the like to work with.

I have the remnants a 10 processor Motorolla 68k bank terminal/transaction server. (each board had its own purpose and processor/ i-o Sadly I don't have the planar that the boards would fit into and I also don't believe this ever hooked into a screen but would have to be terminal run.

Varmint
February 20th, 2008, 07:58 PM
As a general rule, a Weitek cpu IS a 387, unless otherwise noted on the cpu. You'll find it listed in most bios of the era as a Weitek Processor, which can be enabled/disabled in the bios. I've only seen one actually in a motherboard. They see heavy use in satellite tracking applications, as well as some other heavy duty math functions, hence the name "math co-processor". They will usually remain cold unless called upon by the running application. The last 387's I got were $2.00 each at a local store on clearance, and I had a stack of boards to put them in at the time. The good old days.

ronaldSteward
March 3rd, 2008, 09:48 AM
The last 387's I got were $2.00 each
Wow certainly don't cost what they did in the old days do they. A tenth of the price of current mid market cpus. Maybe it's time to load up an old system with a linux system :P

fred333
April 10th, 2008, 07:42 AM
That is very true. you can find some pretty nice deals out there.

SHODAN
June 21st, 2008, 12:21 PM
SO...again, does anybode have an idea if the multiprocessing is supported...coz i have MG-80386 and the same Weitek socket. But probably not, since cpus like 486DLC adn SLC were made, there were some exceptions with doublecocking tech - 33/66 but no NPU. If anybody discovers something, please let me know here or here - vault131@hotmail.com

TroyW
June 21st, 2008, 04:56 PM
SO...again, does anybode have an idea if the multiprocessing is supported.

I'm pretty sure multiprocessing is not supported on the 80386 class CPUs, and I think that mobos with both a soldered on CPU and a CPU socket like has been shown and discussed in this thread are designed that way in case of CPU failure - so if for some reason the soldered on CPU stops working, it's as simple as installing a replacement in the socket to get it running again. But hey, I could be wrong, lol :)