Hey,

I have a Forex 386 board too. It came with 20Mb RAM so it should be pretty sweet, alas, it is very unstable, and freezes continously during various tasks (running linux, freebsd, qnx, and most other stuff too). Tried every possible bus clock setting. Memtest is okay.

http://data.gabucino.be/386FOREX/speedsys.gif

Seems a very nice and fast 386 MB.
Is that an EISA slot below? and is it working on 386 also?

It's likely not an EISA slot (though that is indeed an EISA connector).

It's possibly a proprietary local bus slot....or perhaps just a regular ISA slot if they were trying to use up spare connectors.

Gabucino wrote:

Hey,

I have a Forex 386 board too. It came with 20Mb RAM so it should be pretty sweet, alas, it is very unstable, and freezes continuously during various tasks (running linux, freebsd, qnx, and most other stuff too). Tried every possible bus clock setting. Memtest is okay.

Interesting. Might put Doom or Quake up overnight and see what happens 😀

Do you have a 387 FPU? The benchmark for Quake Timedemo 1 is 1.8fps for a 386DX-40.

I achieved also 1.9 fps with a different 386 system.

dirkmirk wrote:

Do you have a 387 FPU? The benchmark for Quake Timedemo 1 is 1.8fps for a 386DX-40.

Sorry 🙁

why use quake on a 386 ?, the 0.1 fps difference says nothing and could be counted as error.
The previous benches Doom included seems best to me in those systems.

True. And even Doom IMO is not a game you would play on a 386. But I guess there are no other games with such a built in testing capability.

JaNoZ wrote:

why use quake on a 386 ?, the 0.1 fps difference says nothing and could be counted as error.
The previous benches Doom included seems best to me in those systems.

Of course it is just a proof of principle that it runs, nothing more.
The 0.1 fps however is not due to a measurement error. The time such a run takes is so high that an error from the timer precision can be excluded. The amount of frames is exact. The test runs ca. 525 seconds and the 0.1 fps would be an error of about 5%. This means the timer error has to be above 26 seconds in this time interval.

Had memtest running overnight and doom since this morning and no issues.

I tried an OAK video card and it's very slow 😀 Can be used to slow down a machine that's a bit too fast if needed.

Mau1wurf1977 wrote:

Had memtest running overnight and doom since this morning and no issues.

Thanks for trying though.

I also have one of the Forex 386 boards.

It looks like this one:
http://stason.org/TULARC/pc/motherboards/M/MI … ml#.UqHN6uJcX_Z

Did a few tests with it and this is the configuration:

AMD 386DX-40 + IIT 387-40
256KB 20ns cache
32MB 70ns (8x4MB 30pin)
Cirrus Logic 5429 2MB ISA
Soundblaster16 ISA
CompactFlash 1GB formated to 420MB.

Speedsys:
9.27 (NoNPU)
6.31 (with NPU)
Memory Bandwidth: 50MB/s
VESA Memory: 5374kb/s
L1 Cache: 31.32MB/s
Memory Throughput: 23.03MB/s

3DBench 1.0... 17.5
PCPBench DOS VGAmode... 4.3
PCPBench DOS VESA100 (640x400)... 1.9
Quake (nosound)... 1.7
Quake (sound)... 1.6
Doom (nosound)... 7820 = 9.55
Doom (sound)... 8893 = 8.39
Norton Sysinfo... 43.1

WinTach (Win95 OSR2.1) 640x480/16bit (A16 mode):
Word... 14.60
CAD... 21.12
Spreadsheet... 23.12
Paint... 15.84
Overall... 18.67

Winbench 3.11 (Win95 OSR2.1) 640x480/16bit: 8010081

I was quite surprised how well Win95 works on this machine. Of course CL5429 graphics accelerator and CF card helps a lot 😀

Next I'll test 486DLC-40 with IIT 487-40 but it's on another VLB motherboard:
http://stason.org/TULARC/pc/motherboards/U/UN … ml#.UqHS7-JcX_a

Quick test showed 3DBench is faster with VLB card over CL ISA (+2FPS), but Quake is a bit slower. Also memory throughput in Speedsys is slower when VLB graphics card is installed. 😕

Very nice, thanks for sharing!

I must say I'm positively surprised by how well these 386 parts work. Very few jumpers, not much to adjust wrong in the BIOS, auto detecting of HDD works well and everything seems stable as well. Very pleased 😀

Gabucino wrote:

Mau1wurf1977 wrote:

Had memtest running overnight and doom since this morning and no issues.

Thanks for trying though.

In other news, mine seems to run stable at the moment. Either due to the winter cold, or the insertion of an AdLib clone, or it will fail later regardless...

What is this interesting looking application?

Some kind of stress test?

Mau1wurf1977 wrote:

What is this interesting looking application?

Some kind of stress test?

No, it's "top", a process monitor utility for most UNIX(-like) OSes. Running in a remote X terminal.

Top using 25% CPU? That's a 386 alright 🤣

Is there anyone on here that has this FOREX motherboard?

http://www.ebay.com/itm/Forex-386-Motherboard … V0AAOSwU-pXulJn

I see it pop up all the time. The main reason I haven't bought one yet is the apparent lack of support for 256kb cache. Has anyone ever tried using two 32kx8 tag chips and eight 32kx8 SRAMs to see if it will go to 256k? I know the silkscreen and TH99 both say 128k is max, but for a 1991-1992 board it seems silly to limit it like that.