• Please review our updated Terms and Rules here

$10 386/486 (Cyrix Cx486DLC installed + math co-processor) adventures

retro-pc_user

Veteran Member
Joined
Oct 13, 2017
Messages
719
Location
SE Michigan, USA
Bought me another computer and this time a 386 system. It currently has a 1.44MB 3.5" FDD, 1.2MB 5.25" FDD, and some size IDE HDD installed along with a generic (possibly a WinBond chip) ISA I/O controller card and some VGA video card.

Specs of the system:
Unknown motherboard (will look for any markings to identify the board later on)
Cyrix Cx486DLC CPU (either 33GP or 40GP)
Math Co-Processor (might be a Cyrix Cx487DLC, will get a better photo)
Board has 2 EISA slots, 5 ISA slots, has cache chips populated, 4 sticks of SIMM-30 installed, and a bad clock battery (will do a swap and check for any damage and if there's minimal damage, it'll be easy to fix and do a CR2032 battery swap, but I'll follow traces to install a diode to prevent the charging circuitry from blowing the battery up)

The system itself was only $10 at a consignment shop close by, which $10 for a 386/486 system is actually a great price (better than eBay prices).

Going to let the system set out for a while and do some more with it. My plans are to use it as a gaming system with 3 sound cards, plus a CD-ROM drive (proprietary one) to use it as a multimedia system and to play games that are overly speed sensitive for certain MIDI modules and games that were meant for 80386 machines. The Cyrix CPU is a 386 chip (PGA-132 LIF) with 486 instruction sets and 1KB cache.

Stay tuned for further updates, adventures, and computer lab setup for my older systems.
 
The math co-processor is a CHiPs 387 and the CPU is a Cyrix Cx486DLC-33GP. Battery damage is minimal, so that'll be an easy cleanup and the innards, minus the battery are really clean.
 
The math co-processor is a CHiPs 387 and the CPU is a Cyrix Cx486DLC-33GP. Battery damage is minimal, so that'll be an easy cleanup and the innards, minus the battery are really clean.

You got a fairly rare math coprocessor. C&T wasn't very popular because it was short lived.

If you want to play speed sensitive games, DLC might not be the best choice. Maybe a standard 386-33 would be better.
 
Ordered an Intel 386 DX-33 and swapped the 3.5" FDD and cable so that I could use the 3.5" as drive A instead of B. Need to figure out the schematics for this board so that I can solder a diode to the circuitry diode for the CMOS battery (already cut the original trace off from the positive terminal area to prevent shorts/explosions from occurring, but I'll verify to make sure again in case I done messed up). Funny thing is, the BIOS sees the 4.3GB HDD, but DOS begs to differ since LBA doesn't exist on a 386 board unless an enhanced IDE BIOS is used or XT-IDE Universal BIOS.
 
And the clock won't keep up. When I power it off, the time is 9 minutes behind (if I leave the system off for 9 minutes) or the settings disappear. I already checked the connections and they are all good.
 
And the clock won't keep up. When I power it off, the time is 9 minutes behind (if I leave the system off for 9 minutes) or the settings disappear. I already checked the connections and they are all good.
Symptomatic of low battery voltage to motherboard chip/s.

CR2032 = 3 volts, then that will be reduced by the diode. Are you sure that is adequate for your motherboard ?
 
Yeah, the CR2032 wasn't sufficient enough. Apparently, the diode drops the voltage by 0.5V-0.58V, which is insufficient and the RTC requires 2.7V-4.5V exactly. I decided to solder a 3x AA battery pack by using a GameBoy DMG-01 rear chassis and the positive terminal from the mainboard of the DMG-01 since the board was dead anyway and soldered two wires: 1 for positive, 1 for negative and got 4.5V at the battery leads and 3.8V at the other end of the diode and to the RTC chip, plus, the capacitors and other points that are on the same circuitry. I soldered the battery pack to the external battery header since that is a safer method and I ran out of spare diodes from other boards that broke in the past.

I then tested it out and the system is keeping the time, date, and settings. All in all, everything is working properly and I also changed the Sound Galaxy NX Pro H/W address to 240h so that the SAAYM (CMS/GameBlaster Clone + YM-2151 and YM-3012 Stereo DAC) can use the H/W address 220h. The only thing I need is to figure out how to disable the COM2 port so that I can use IRQ3 for the YM-2151 chip for a VGM player software so that I can play some VGM files.
 
Got COM2 disabled now, received an Intel i386 DX-33 and installed it, yet HIMEM test fails: says unreliable XMS memory at address 00113884. XMS driver not installed. With the Cyrix CPU, no error messages. The RAM itself has gold plated pins, yet the RAM slots are tin plated, which I believe is one of the issues, but I don't think that's the case. The system detects the CPU as an AMD Am386 DX-33 according to NSSI. So I'm not sure if it's because it want an AMD CPU installed or if it's something else I'm missing.

I cannot find any jumper settings for the board, so I think it auto detects the CPU, so that shouldn't be an issue and the crystal clock oscillator is 66.67MHz, so it's the right speed.
 
I checked for bugs, and sure enough, there is a bug: POPAD bug: ! ERROR ! according to HWiNFO. No bugs in the Cyrix CPU, so maybe that explains the issue there. And there is a bug-free one (double-sigma ΣΣ) out there that I see (SX211 CPU ID marker). Going to order that CPU and some RAM to play it safe. Even MemTest86 won't load as the system restarts after showing the blue screen after the loading MemTest86 text appears.
 
I'm kind of scratching my head here; why do you want to replace the Cyrix CPU so badly if it works fine? I doubt there are many games that are speed sensitive in a way that they would fail to run on the Cyrix CPU yet work on a 386/33. I mean, sure, I could see the argument for reverting, say, a 5150 that'd been upgraded with a V-20 back to an 8088 in order to make it run "cycle-counting" games and demos correctly, but is there really anything appropriate for a high-end 386 that's going to fall into that category? The 386/33mhz and the 486/25mhz were practically contemporaries (both announced in 1989, but realistically you probably had to wait until early 1990 to get your 486), which the Cyrix chip slots somewhere in between in terms of performance, and on the other end the 386 went all the way down to the lowly 386SX/16, so it seems unlikely to me there were many games that would fail to run on the Cyrix chip unless it was some really obscure issue with the built-in cache or whatever.
 
I don't see any point in replacing the CPU either. The Cx486DLC40 supports the full 486 instruction set and, according to Wikipedia, performs similarly to an i486SX/25 or Am386/40. Replacing it with an i386/33 is actually a significant downgrade - both in generation and performance.

All 386 contain the POPAD bug. There is no bug-free version. The Cx486DLC is not a 386, so does not have this bug.

All 386 with double-sigma are fully 32-bit capable; to my knowledge, this includes all chips faster than 16 MHz. Without double-sigma (i.e. early 16 MHz chips only), there are severe bugs when mixing 16- and 32-bit code. Such chips cannot run Windows 95.

The 486 instruction set adds instructions for more fine grained cache invalidation (INVD, INVLPG) and some atomic instructions necessary for decent SMP support. This may very well be the reason that memtest86 doesn't run on your 386.

So... I think replacing the CPU is a step in the wrong direction.
 
It particularly seems like a step back if the BIOS in the board is properly recognizing the Cyrix chip and has full support for it. I know the upgrade version of the Cyrix 486DLC was sometimes a little finicky and could have some compatibility gotchyas with the cache, but with full BIOS support it should be a perfectly decent CPU and arguably more interesting than a plain 386. I'd sure be thrilled to have found a complete system with one for $10. ;)
 
Good thing I archived the old 203MB Maxtor drive as that had the Cyrix software for the cache on there. I'll try that out and see what happens. Had a hiccup with the Diamond SpeedSTAR VGA rev. C6 (ET4000AX) card as it wasn't getting detected (got detected once and that was it), and after changing the DIP switch #3 to OFF, it started to work (backed up the ROMs as well to be safe and make spare ones in case the original ROMs die mysteriously). Also got 8MB SIMM-30 Parity 70ns (8x 1MB) for the system as the RAM might need to be replaced (the 386 DX-33 has a 32-bit instruction bug, which is causing HIMEM to act crazy if /testmem:eek:n is in place in config.sys, but Memtest86+ v4.10 reports no issues, which is weird), but better to have spare RAM in case the original 4MB SIMM-30 (technically SIMM-32, but the first and last pin are not connected to anything, except for common ground) stops working one day.
 
An update on the system: I kept the 486DLC-33GP installed and swapped the I/O controller card with a Promise EIDE Pro Super I/O card and removed the XT-IDE Universal BIOS from the 3Com EtherLink III 3C509TP ISA card and needs an HDD reformat, but I'm going to copy the files over from the HDD to my laptop, format the HDD, and copy the files back over to the HDD.
 
Update: got the files copied to my laptop's hard drive, reformatted the 4.3GB HDD, and copied the files over to the drive. I then swapped boards with the M326 V5.2 that had the Cyrix Cx486DLC-40GP and Cyrix FasMath 8D387-40GP installed, but that board would lock up and I believe it was because of the BIOS issues, but I'm not certain. I then swapped both the CPU and FPU from the 33MHz Cx486DLC-33GP and C&T 8D387-33 to the Cx486DLC-40GP and FasMath 8D387-40GP, and installed an 80MHz clock crystal to get the full 40MHz. Runs pretty well.
 
Back
Top