Thanks everybody for your help! I see now that DX motherboards are indeed rare. Will look for 'classic' mbs that are available (will be assembling 486-era PC some time later, so additional versatility would be an overkill, but thanks for detailed explanation!)

As for CPU/FPU, I have one restriction (kinda like a challenge 😀 ) - it must be the fastest Intel-manufactured CPU and/or FPU, so that narrows down my search even more (no AMD DX-40 motherboards).

Were all 386 motherboards designed with barrel batteries? I found a picture of EVEREX STEP 386 REV. E (presumably) motherboard, and it has only CR-2032 slot attached (possibly later modification by owner?)

I ended up bying M321 rev 2.6 motherboard (https://stason.org/TULARC/pc/motherboards/I/I … -386-M-321.html)

Strange thing, it had no keylock (+power led), turbo switch, turbo led, reset switch and speaker connectors (J10 to J14 on scheme), looks like they were pulled off (why?).

I Installed speaker, power led and reset connectors, and everything works except of power led.

By looking at motherboard images, I saw that mine is missing also some other components - C47 and R38, also located near J10-J14. Can someone, who owns the same MB, tell me C47 and R38 nominals?

I think that R38 might be the ballast resistor for +5v for power led, am I right?

R38 appears to go to pin 1 of J13 so yes is probably for power LED. From the pictures it is a 150 ohm (brown-green-brown), cannot read the Tantalum C47.
https://www.ultimateretro.net/en/motherboards/3059

A bit larger picture of rev. 2.7 - http://www.amoretro.de/wp-content/uploads/pc- … motherboard.jpg

C44, C45, R37, R50 are also missing...

Horun wrote on 2022-05-01, 17:09:

R38 appears to go to pin 1 of J13 so yes is probably for power LED. From the pictures it is a 150 ohm (brown-green-brown), cannot read the Tantalum C47.

Wattage of R38 should be 0,25 watt or more?

lalakobe wrote on 2022-05-01, 18:14:

Horun wrote on 2022-05-01, 17:09:

R38 appears to go to pin 1 of J13 so yes is probably for power LED. From the pictures it is a 150 ohm (brown-green-brown), cannot read the Tantalum C47.

Wattage of R38 should be 0,25 watt or more?

yes 0.25 watt should be fine. actual current be about 25mA so about 125mW

Replaced R38 and got power led working.

Still wondering what was the purpose of other missing components.

I want to upgrade cache to maximum (now mb has 64 kbytes). Two small questions:
1. Is ISSI IS61C256AH-15N memory a good choice?
2. This motherboard has a slot for 16x4 chip (U32) which enables cache write-back. Which mode performs faster on 386s - write-back or default (is it write-through?)?

ISSI chips should work fine. It would be best to make sure you get real ones. There are some fake ones from China that are also viable, but you need to be prepared to weed out the defective ones. There's about 10-15% failure rate with those.
It's possible that the dirty bit in U32 may have to be upgraded to 64kx4 when upgrading to 256k.

Thanks!

I will try first with 16k x 4. What should I buy? DIP 24 and 15 ns ones? Need some recommendations to start search.

Got some ISSI 15ns, installed them and got 256kb of cache. Nice! Had to solder JP3, it was missing from the motherboard.

Also got two QS8888-15 (DIP22) and tried to install it in U32 slot. Slot is DIP24, but I saw somewhere here that you still can install DIP22 in it, aligning with pin 1.
However, I got no performance increase nor decrease in CACHECHK and System Speed 4.78 benchmarks.
It is possible that both chips are dead/fake OR DIP22 is not compatible with DIP24 OR pin configuration of QS8888 is not compatible with what motherboard needs OR 64kx4 is needed there.

I think that pin configuration is the main culprit. What other 16kx4 chips should I try?

Also, is there some way to test if write-back is enabled?

I have this mobo https://theretroweb.com/motherboards/s/microm … -tech-80386-wbh with an ISA ATI Mach64
I upgraded the board to use MR BIOS which made it much more pleasant to use

I expanded the cache. Took some trial and error to find chips that worked. These worked for me:

Speedsys gives a pretty good indication if writeback is enabled when you look at the memory speed graphs.
If the move speeds are higher than the read speeds, writeback is almost certainly enabled.
I used cyrix.exe to enable & disable the Cyrix on chip cache. when I enabled the onchip cache, it looked like the mobo cache was in write back mode.

I did a bunch of benchmarks with different CPU's and Cache settings.

• All bench marks are with Hidden Refresh, 0 wait state, 10Mhz ISA bus, 40Mhz CPU
• "-e" enabled the onchip cache, "-e-" disabled the on-chip cache
• -k (Ken) slowed things down a lot, -f (Flush) and -b (Barb) had little to no impact. Probably need to mod the board to get any impact there.
• Going from 0KB L2 to 32KB cache has a bigger impact than 32KB to 256KB L2 cache.
• The Cyrix was faster in DOOM with the L1 disabled & 32KB L2 cache than it was with 1KB L1 cache enabled and the L2 disabled.
• Doom can take advantage of bigger caches while most DOS synthetic CPU benchmarks fit pretty well in a 32KB cache

1CPU        Mobo   Cyrix    Doom   Speedsys  Read    Write  Move   Avg
2           Cache  Tool   High FPS CPU Index MB/s    MB/s   MB/s   MB/s
3Ti486DLC   256KB  -e -b    10.9    10.29    30.63   37.48  25.53  31.21
4Ti486DLC   32KB   -e       9.81    10.20    30.55   37.47  25.48  31.16
5Ti486DLC   32KB   -e-      7.48     9.47    30.55   37.46  38.16  35.39
6Ti486DLC   No     -e       7.08     9.88    18.57   37.05  13.48  23.04
7AM386-40   256KB  N/A      7.04     6.88    25.53   30.07  38.30  31.30
8AM386-40   32KB   N/A      6.70     6.88    25.47   30.05  38.16  31.23
9AM386-40   No     N/A      3.16     5.19    16.53   29.87  16.49  20.96
10Ti486DLC   No     -e-      2.93     5.76    18.57   37.05  16.49  24.03

There is a thread Cache chips problem about similar motherboard M321 but different revision (mine is 2.6), I expected to see similar behavior after installing WB cache in U32.

It has reports of using DIP22 CY7C164 with success.

In the second-to-last message we see SpeedSys screens, one with U32 installed and WB enabled, second without. Differences I see:

1. Memory bandwidth with WB - 54.45 MB/s, without - 30.47 MB/s
2. Memory throughput with WB - 20.11 MB/s, without - 18.27 MB/s
3. Moving (green line) is above Read (yellow) in cache zone (128 KB), but goes below after - with WB
Moving (green line) is always above Read (yellow) - without WB.

In my case Memory bandwidth stays the same with or without U32 chip.

I think that my QS8888 has different pin configuration than CY7C164 (just look at address lines) and that is the source of problems.

CY7C166 arrived today, installed it in M321 and WB seems to be working. QS8888 does not work.

1. Memory bandwidth with WB - 44.84 MB/s, without - 25.27 MB/s
2. Memory throughput with WB - 16.81 MB/s, without - 15.33 MB/s

Here is an idea for your build.

Custom interposer module for TI486SXL2-66 PGA168 to PGA132 - HELP!

I will be building up my 386 with this when testing is done on the interposer 😀

1) Intel CPU

2) Intel third party motherboard
Maybe 1993 bios or newer.

3) 410mb IDE hard drive

4) 8mb ram ( FPM , 60ns )

5) Cache - 128kb motherboard cache or 256kb ( I would not waist my time on more cache. As long as you have 128kb you are fine
It really does not improve performance much. Going from 64kb to 256kb only improves performance to about 3% to 5% )
For a Real Performance boost upgrade the CPU.

6) 32x CDROM or faster
Try to find a 52x

7) SVGA 1mb ISA Card

😎 Sound Blaster Pro or compatible sound card

9) 3com 3c509b Network card.

10) TP-Link Network extender for Ethernet to Wifi Networking

11) Gold Star IDE Controller card

12) Floppy drives ( 3.5 and 5.25 )