• Please review our updated Terms and Rules here

Best 386-era machine?

Yeah, I can't really think of any really "awesome" 386 machines because that time frame was very dominated by clone PCs, as already stated. That said, I wholeheartedly agree with using a 386DX-40 w/ say 128K on board cache and 8 MB of RAM, more if the board will support it, but being a 386, I figure you're looking mostly at DOS and maybe Windows 3.11 apps, for which that config should work fine.

The Compaq suggestion is good. Heck, maybe you can find a System Pro and get really lucky (i.e. first EISA-based PC, but has slots to asymmeterially process two 386s if you run SCO or NT 3.1); I also seem to remember Northgate Computers making boatloads of 386 and early 486 clones during that time frame as well, but the 40s came late in the 386 line (1991 I believe because of the AMD vs. Intel lawsuit), so if you go with that, a "clone" board is probably all you will get.
 
Anonymous Coward said:
AMD 386DX-40 with cache is a nice 386 machine.

You can also get 486DLC3-100 with VLB slots, which is technically still a 386...but probably a little overkill for what you want.
Click to expand...
Back in the mid-'90s I upgraded my 386SX computer with another 386SX-based processor -- this time in the guise of an IBM 486SLC2-66. It provided a cheap alternative to the Intel 486DX2-66, with all of the clock speed but none of the 32-bit performance. Mine even had "32-bit" VESA Local Bus slots, but the I/O speed was cut in half due to the CPU's 16-bit external data bus. Also, all 386SX-based CPUS are inherently limited to a maximum of 16 MB of RAM, which is fine for DOS and Windows 3.1, but not much else. Anyway, that 486SLC2 had the smallest CPU cooler fan I've ever seen: a tiny but very loud fan sized at a whopping 1 inch square.

About a year later, IBM hit the market again with the "Blue Lightning" 486BL3-100 -- the first ever x86 CPU to reach 100 MHz. It was really a clock-tripled 486DLC with 16 kB internal cache -- itself based on the 386DX, and thus an external math co-processor was still required. Of course, performance again paled in comparison to the Intel 486DX4-100 which hit the market a few months later. IBM later used the "Blue Lightning" name on Cyrix-built 486DX4 chips, which were finally full 486DX-based chips in every sense of the word.

Note that Cyrix also made their own versions of the 486SLC and 486DLC, but with only 1 kB of internal cache, as compared to IBM's 16 kB. My dad had a Leading Edge laptop with a Cyrix 486SLC-25 in it, and again it was really just a glorified 386SX-25, with that tiny 1 kB cache and a 486-compatible instruction set added.
 
Unknown_K said:
I think I just snagged one of the slowest 386 computers made, a Samsung 386SX 16.
Click to expand...

I have a NEC 386SX-16 with 1 MB of RAM onboard. For additional memory you're supposed to use NEC's proprietary RAM board, which I don't have, so instead I put in an ISA RAM card with 8 MB of SIMMs installed. So now Windows 3.1 does run in 386 Enhanced mode... but very slowly, since all RAM above 1 MB is being accessed through the ISA bus at a blistering 8 MHz!
 
Have you tried benchmarking your 386SX/16 against, say, a comparably equipped 286/16? If so, who won? In any case, on a machine like that, I think a nice choice is running DesqView 2.X and "multitasking" DOS apps on it......even games.
 
I have a 286/16 (has a 20 Mhz chip but not a jumper setting to go that high) I could benchmark my 386sx/16 to, but I expect the 286 to be faster. A 386SX/16 would be perfect for old DOS games that were meant for a 286 and it allows you to use QEMM to get the most out of the base 640K memory.
 
glitch said:
If you can find one, the Compaq Portable III was a pretty cool 386. You could get an expansion box for it, that would accommodate two ISA cards. The orange plasma screen was pretty impressive. A friend of mine used to have one, which he used as a terminal to connect to his Linux server over Telnet.
<snip>
Click to expand...
Hmm... wasn't the Portable III a 12MHz 286? Mine's in pieces in a box somewhere but that's my recollection.
 
My current project is fumbling around with a couple of these PCs.

The Portable III was a 286/12.5 while the Portable 386, as the title would imply, was a 386DX/20. The III didn't have a box with it, but the 386 did, and I've got a graphics card and a NIC in there. These are very sturdy machines.
 
Hi,
If I may reply, I used to have a very nice 386 back in the end of the year 1999 as a teenager, wich was my second computer (The first was a tandy 1000, with a 286 upgrade, 640 k ram, internet capable (in network with my older brother's pentium 200).

It was an Am80386Dx 40mhz
387 math co-processor
32 Mb of Ram (Max)
3.2 gb Hard drive (with ez-bios)
1.44 mb floppy 3.5
1.2 mb floppy 5.25
52x cd-rom drive
Ati mach 32 Isa 1mb
Sound blaster 16 Isa
10/100 Cogent Em 110/tx ethernet adapter ISA
Windows 98 (Had to put the hard drive on the pentium 200 to install it, then put it back on the 386 to auto-detect hardware after being prompted to restart it.

When my uncle gave it first to me, it only had 4 mb of ram, 80 mb hard drive, no sound, no network, 512k svga realtek, windows 3.11 for workgroup. I then installed windows 95 OSR2 without IE, then upgraded to 20 mb of ram, added sound, network and better graphic, etc.
 
Question for Captain Midnight (or anyone knowledgeable):
Why would a 286@16mhz be faster than a 386sx16mhz?
Just curious to find out.
 
pianoman72 said:
Question for Captain Midnight (or anyone knowledgeable):
Why would a 286@16mhz be faster than a 386sx16mhz?
Just curious to find out.
Click to expand...

Well the SX uses a 16 bit bus which is the same as the 286, I would think clock for clock the 286 (both using a 16 bit memory bus) might be faster.
 
(Oh, I'm going to get flamed for this one--but I've got the docs to back it up.)

For the same clock speed on 16-bit real-mode code, the 386SX is slower in general than a 80286, particularly in memory reference and I/O instructions, but some differences are fairly dramatic (e.g. LOOP (8/4 on the 286; 11 on the 386SX).

The point of the 386SX was to get a 32-bit capable processor on a 16-bit bus, not to get a huge jump in performance over the 286.

By the same token, there are some instructions on the 486 that, clock for clock, run slower than the 286.
 
So the execution of real mode programs in general, is somehow less hindered on the faster 286, compared to the same clock speed of a 386sx.
 
pianoman72 said:
So the execution of real mode programs in general, is somehow less hindered on the faster 286, compared to the same clock speed of a 386sx.
Click to expand...

You've got to remember that the 386 in comparison to the 286 is hugely more complex--not only does it support real- and segmented-protected mode operation, but also V86 and 32-bit paged mode operation, as well as supporting 32-bit instructions. There's a price to be paid there. The saving grace is that 16MHz is the bottom of the 386 speed range, while 16 MHz represents the top of the 286 speeds. So perhaps the comparison isn't fair. Using a 16-bit bus interface further hinders the SX.

The big game changer was the 486 with a completely redesigned execution unit.
 
Chuck(G) said:
The saving grace is that 16MHz is the bottom of the 386 speed range, while 16 MHz represents the top of the 286 speeds. So perhaps the comparison isn't fair. Using a 16-bit bus interface further hinders the SX.
Click to expand...

There were 20 and 25 MHz 286 CPUs as well -- just not from Intel.

Also, there were some very early 12 MHz 386(DX) chips made, before Intel was able to get the 16 MHz yields high enough (and before the 386SX even existed).
 
ISTR that the 20MHz 286s were CMOS, not NMOS and generally given the part number 80C286 or such--and had the capability to be clocked down to DC. Intersil sold a lot of these.

The "ain't as fast as was promised" was the case for the 80286 also. Initially, we were told that the 286 was going to start out at 8MHz, since production 8086s were being shipped at the speed. They couldn't do it, so we got 6MHz parts for quite a while.

I think Intel really felt the heat to get a 32-bit CPU out. The 432 was a disaster and Motoroala was making great headway with the 68K line; and there were a bunch of 32-bit RISC CPUs out that might have eated away at the high-end market share.
 
I had a Compaq 286/20 system. Yeah, it was a Harris/Intersil chip in CMOS process. I don't recall whether the AMD 20 and 25MHz units were CMOS, though--best I recall was that they were NMOS speed bumps they did when Intel wouldn't license the 386, but I wasn't watching the eighters closely by that time--I was overloaded with DG Nova, 68K and 32K work, and looking forward to the 88K. (Which bombed when it appeared.)

My favorite of the 386s for a general purpose system was the HP Vectra. In its last incarnation, mine had a 50MHz 386 in it that was still smoking my 486DX3s years later. It was a compute engine for most of its life, then became an ftp server since it moved data across ethernet faster than any other system I had to spare for that job up till a Pentium II was freed from database duty.
 
The AMD 20 was also CMOS, as was the Harris/Intersil 25:

N80C286-20.jpg


NMOS had problems running much faster than about 16MHz (and those ran pretty hot).

The 286 was a huge improvement over the 8088 in terms of performance. I can't recall any other new chip design that gave that kind of advantage over its predecessor.

I still have a Micron 386-16 system here with *no* memory on the planar--it's all in a plug-in card. It runs Windows 95 slowly, even with 13MB of memory.
 
You must log in or register to reply here.
Back
Top