Chaintech 486SPM M104 (seems to be the latest revision), based on SiS 85C496, 85C497.

In crisp condition. Really well preserved.

Feels like slightly early design compared to the very late 486 motherboards.
No support for EDO RAM.
No 3.45V to CPU option.
Very easy to setup jumpers. Very nice.
There is not much more to add - very straightforward assembly.

Test setup:
Matrox Millenium PCI 4Mb
1Mb level 2 cache, 10ns rated chips
32Mb or 64Mb 60ns FPM RAM
CF adapter and card connected to the onboard IDE

--- Am5x86 at 160MHz (4x40)

All BIOS settings on max.
PnP BIOS auto-config can hang the system during POST. Requires BIOS reset to resolve it.
Turning that parameter off resolves the problem entirely.

SpeedSys:

benchmark results

Not the fastest motherboard, but not bad either.

--- Am5x86 at 180MHz (3x60)

There are no clock generator jumpers for 60/66MHz, but this simple mod unlocks these frequencies.

Different tests pass with arbitrary tighter BIOS settings and performance is really good, but had to slow things down for a fully stable system:
CACHE WRITE CYCLE = 3 CCLK (best is 2)
CACHE BURST READ CYCLE = 2 CCLK (best is 1)
DRAM WRITE CAS PULSE = 2 CCLK (best is 1)
DRAM SPEED = FASTER (best is FASTEST)
ISA BUS CLOCK = 1/4 PCLK (best is 1/3), the 1/3 option works, but not every key stroke is captured and processed, which is annoying

Speedsys numbers look nice.

All 180MHz metrics will be combined soon into a single chart, so for now posting the test results in a textual form only:

1Wolf3D: 141.5 fps
2PC Player: 29.8 fps
3Doom: 69.7 fps
4Quake 1: 19.9 fps
5WinTune 2: 15552 KPixels/sec
63D Studio R3: 103 sec (1:43)
7LightWave3D 4: 1099 sec (18:19)

Numbers are below what PC-Chips M919 with 1Mb L2 cache module shows (which is currently the best option in this category), but still - the fact that 3x60 is a viable for this motherboard is great by itself.

With smaller level 2 cache size tighter BIOS timings will most likely be possible. Something i may explore later. For now will stick to the maximum L2 cache buffer.

--- Am5x86 at 200MHz (3x66)

The system hangs during POST no matter what.

--- Am5x86 at 200MHz (4x50)

POST completes but always hangs during BOOT to DOS.
It is related to the level 2 cache. If turned-off or chips not present - all is good.
Tried hard to overcome the issue but without success.

--- Intel Pentium Overdrive P24T (POD100) at 100MHz

All BIOS settings on max.
No jumper changes required between Am5x86 and POD100 CPUs, which is pretty nice.
Either the motherboard supplies 5V to CPU by itself, or POD100 works fine at 3.3V on this board.

Updated the POD100 comprehensive summary post with the related information for this motherboard.
The short summary:
Intermediate performance in interactive DOS graphics.
Number 3 in Quake 1, number 2 in Windows GUI and one of the top 3 in complex offline computing.
Quite satisfying.

---

The SPM model is a clear upgrade over the previous SOM model which does not handle POD processors and shows overall lower performance metrics.
Also, very easy to work with system.

Simple mod by Feipoa to enable 60/66MHz here.

Thats a very nice Board - stable and reliable. I have this Board in Rev 102. It seems my BIOS is newer from 1996, maybe you can use it for your revision.

I have uploaded the BIOS here: https://dosreloaded.de/forum/index.php?thread … pci-mainboards/

BIOS Chaintech 486SPM M1.02B (1996/01/24)

Hi again,
I have (unfortunately only very) small updates:

I made benchmarks with the PVI @ 160MHz with 256kb Cache configuration (double banked) and the VLB Trident 9440AGi (with the jumper for 0 waitstates) and came closer to your values.
Some values still too far off .....
I switched then to single bank 512kB cache config - that was surprising for me - some benchmarks improved a lot:
PCP 640x480 from 8.5 fps to 9.9fps (more than 10%!), also Chris' 640x480 from 18.6 to 20.7 (also more than 10%). Quake 640x480 improved from 16.3 to 16.8 fps.

Still I had with Windows 98 SE only a bit more than 3000 in Wintune 2 graphics (so half your value, tried all drivers I could find including the one you sent), so something fishy with the VLB Trident and Windows 98SE ....

3D Studio chevy bench is a good benchmark. I got 1:11: with both 256kb and 512kB (sometimes), also the lightwave 3D test now runs, but I have to investigate more, values are too high in comparison to yours.

The reason why I don't give a nice complete list of values is the following: Testing with Windows 98 SE and in DOS with Autodesk 3D Studio revealed instabilities with 512kB cache. I started then swapping around chips, but did not succeed to reach full stability. Acutally I made it worse 😁. So I even have problems with single banked 2-1-2 at 40MHz. Believe it or not, I even corrupted my Windows 98 SE installation, I must re-install it.

I switched back for now to known good config 256kB double banked.
=> What I learned: Installing and running Win98SE (and some apps like the Lieghtwave from you) AND running 3D Studio in DOS are pretty good stability tests.

Sticking now to DOS 7.1 only, I tried DOS benchmarks with a PCI S3 868 2MB (instead of the VL - 9440). As expected everything was slower, except one:
Quake 1 640x480 increased from 16.3fps to 16.6fps! Still a bit lower than 16.8 fps with VL Trident and 512kb Cache. Maybe that is interesting for you?

So either I now try again to find a 100% stable set of 512kB Cache with the PVI ..... or I now switch back to playing with the VL/I with 1MB cache (getting that stable seems to be more easy for me @160MHz, most likely because of the double banked cache).
I have some more VL cards, that I can try out, but no special ones (Mach 32 VRAM, Cirrus and a "Chips")

@PC-Engineer
Always good to pass the cross-check.
Thanks for the link. Good information there. (the VLI note about issues with L1WB and DMA - i think this is related to Adaptec SCSI controllers only, everything else seems to be ok)
Cannot remember right now the BIOS version i tested with.
BIOS sticker says 1998 but that does not necessarily implies latest micro codes. I think it was the latest visible online ...
Away from home for few more days. Will check once back.

@CoffeeOne
Actually you touch on quite a few things that i find important (well, in the context of a retro computing hobby i mean).

It is good that you are running more comprehensive tests instead of spinning Phil's benchmarks only and call it a day.
As you can see for yourself - everything checked-out with the DOS tests, but things stall with the additional ones i suggested.
Let's say that i spent good amount of time distilling that stuff before settling down on my current methodology.
First coarse approximation - Phil's tests. If something does not go well there - big red flag. But also, passing these tests means little.
Second approximation Watcom C compile of small-medium size project and offline 3D graphics tests in DOS. That's the 80% mark on stability.
Then Win9x/NT overall stability, installation, drivers, VC++ compile with a medium size project, Internet Explorer - surprisingly sensitive in some cases, offline 3D graphics. For late 486 boards with PCI slots - GL Quake and maybe 1-2 more early 3D accelerated games.
So far didn't have a case where code compilation and IE fail if offline 3D graphics pass, so i share here only the later.
For complete stability i run in parallel (in Windows) code compilation, IE, GLQuake and offline 3D graphics tests. Takes a while for things to complete, but if passed the system is as stable as it can be.
I don't run composite testing for each and every piece of hardware. The individual tests are good enough approximation. But the few retro PCs that i assembled got grinded like that. No half measures. Things must be real.
: )

Few more words on tests.
Code compilation is a good stability test, but not a good performance one - CPU only + lots of disk I/O.
WinBench is in the same category with other synthetic tests such as SpeedSys, NSSI, Landmark, CheckIt. Meaningless overall.
WinTune is mostly in the same department. What i like there is the quick accelerated graphics run - drawing bunch of primitives and text. Ok approximation.

About your PVI board with 256 and 512 kb level 2 cache buffers.
PVI likes 512Kb single banked more than 256Kb double banked. Go with the bigger size. Tightest BIOS timings are possible.

VLB Trident cards are not very fast for Windows GUI. I am not sure what you are comparing against, but if you look at Matrox Millennium numbers that i shared in previous posts - no chance you can match that with the Trident card. If you compare to Trident TGUI9440AGI or -1 and still falling short - make sure you are in 16-bit color depth. For most VLB cards 16/32-bit colors result in very big performance difference.

Guys I could have sworn at one point I was able to boot my 8434UUD @60mhz FSB with the PCI ratio being 1:1. If memory serves I had to use a promise pci ide controller and I only have the one. I even have some photos with the bios settings and benchmarks. Now it won’t work 🤣. Am I imagining this?

It works, but with third party PCI IDE you have to turn off one of the BIOS options. Away from home now and cannot check, bit it was PRE-something.

“Preempt PCI Master Option” maybe?

Yep that was it! Needed to be disabled, thanks!

PC-Engineer, could you upload your BIOS for the 486SPM board to this thread? I receive some error message in German when trying to download it from the dosreloaded.de link you provided. It states I need to be a registered user.

For my Chaintech 486SPM board, I have BIOS dated 03-06-1996. I don't have the system running, so I'm not sure if that date is 3 June 1996, or 6 March 1996.

pshipkov, the Chaintech 486SPM is a very fast PCI SiS 496 board. I've noticed that you are comparing it with a VLB based board using the ARk1000VL. I think a fair comparison would be to compare it against other PCI 486 boards.

pshipkov wrote on 2022-12-28, 19:38:

About your PVI board with 256 and 512 kb level 2 cache buffers.
PVI likes 512Kb single banked more than 256Kb double banked. Go with the bigger size. Tightest BIOS timings are possible.

How many different cache modules did you test with 256K double-banked to arrive at this conclusion? I haven't run into a case whereby 512K single-banked was more stable with SRAM/DRAM tightest timings on a socket 3. Perhaps you just found a very nice set of 128kx8 and that this conclusion cannot be generalised? Do you have more data on this?

Chadti99 wrote on 2022-12-30, 02:12:

“Preempt PCI Master Option” maybe?

Yep that was it! Needed to be disabled, thanks!

Did you check my manual for the 8433UUD ? On page 22, I indicate that Preempt PCI Master Option should be disabled.

However, I have later pencilled in on my printed manual "disabled if using onboard IDE. Promise Ultra100 can have this option enabled... check again - confirmed, it can be enabled w/Ultra100"

I ran the tests with a BIOS from 1995.
Will give the later BIOS a quick spin and will update numbers if something notable comes out of it.

---

Just looked up some of the collected perf numbers.
From the tested so far 35 486 motherboards Chaintech 486SPM M104 ranks as follows:
Wolf3D: 21st place (9th place if we ignore VLB boards/adapters)
PC Player: 4th place (3th place)
Doom: 11th place (5th place)
Quake 1: 4th place (3th place)
WinTune 2: 5th place (5th place)
3D Studio R3: 5th place (offline compute, video adapter does not matter)
LightWave3D: 7th place (offline compute)
Visual C++ 6: 12th place (offline compute)
Watcom C/CPP: 6th place (offline compute)

Will share the combined data soon.

---

About Asus PVI-486SP3 + 512Kb L2 cache + tightest BIOS timings.
Usually it takes 2 or 3 times the number of chips to find a good combination.
When i was setting up this board remember buying 10 pieces of those 10ns ones.
It didn't think long to find the 4 right ones + the tag chip, which i think is 12ns.

---

Biostar 8433UUD works here with third party Promise UIDE controllers with PREEMPT PCI MASTER OPTION set to ENABLED, but i happen to remember that Chadti99 had to turn-it off for some reason, so reminded him about that.

Updated the BIOS in the post above. It’s from Janurary 1996.

@chadti99
Tried again Asus VLI at 200MHz.
System is not fully stable.
Will spin it few more times, but writing is (still) on the wall.
At 4x50 it matches Chicony CH-471B. Maybe slightly faster in some tests.
CH-471B at 3x66 is the faster board.

Thanks for trying pshipkov! Did you find a jumper combo that worked at 200Mhz and 5volt?

I tried the promise controller again on the 8433UUD at 1:1 and I wasn’t getting consistent results. I was able to boot with PREEMPT disabled but discovered a reboot would cause a hang on the next boot. I’ll give it another shot tomorrow and re-read Feipoa’s guide. I was def referencing it when I was working on this build, it’s a fantastic write up.

@pshipkov have you tried the 4SA2 at 4x50? Tempted to move my 200MHz 586 over for testing. I think it has some potential in conjunction with TweakBios. Unknown if it’s Windows stable at that speed.

Asus VLI Revision 2.1 of the board offers jumpers for 3.3, 3.45, 3.6, 4, 5 volts.
With one of the CPUs i can get to 200 even at 3.3V but needs 5V Peltier.
The system can get in weird state no matter CPU voltages.

Didnt push Soyo 4SA2 at 200MHz past the brief confirmation that it works since you went over it.
I you feel there is a potential - we can give it a second chance.

Chadti99 wrote on 2023-01-01, 01:21:

Thanks for trying pshipkov! Did you find a jumper combo that worked at 200Mhz and 5volt?

I tried the promise controller again on the 8433UUD at 1:1 and I wasn’t getting consistent results. I was able to boot with PREEMPT disabled but discovered a reboot would cause a hang on the next boot. I’ll give it another shot tomorrow and re-read Feipoa’s guide. I was def referencing it when I was working on this build, it’s a fantastic write up.

@pshipkov have you tried the 4SA2 at 4x50? Tempted to move my 200MHz 586 over for testing. I think it has some potential in conjunction with TweakBios. Unknown if it’s Windows stable at that speed.

My 8433UUD manual is due for a major overhaul, but I haven't had the time or motivation to do it. Now I noticed you used the word "reboot" causes it to hang on next boot. If by "reboot", you mean pressed the hard reset button, then that is odd. If by "reboot" you mean a soft reboot, or CTRL-ALT-DEL, then the behaviour you witnessed is expected. The 8433UUD cannot soft reset with any of the Promise IDE controllers that I've tried. I think I tried the 100/133/SATA-plusTX2. I don't recall if I tried the Ultra66, or if I even own that card. If you need to reboot, hit the reset button when using a Promise controller. if you insist on having the soft reboot, then you should use another controller card, like the Adaptec AHA-2940UW or 2940U2W. I've done most of my testing with the later.

Okay this makes sense, thanks for confirming this behavior! I’m shopping a 2940. Does it handle 1:1 at 60Mhz FSB?

Chadti99 wrote on 2023-01-01, 04:25:

Okay this makes sense, thanks for confirming this behaviour! I’m shopping a 2940. Does it handle 1:1 at 60Mhz FSB?

I don't know. I think all PCI 2.1 spec cards are supposed to support 66 Mhz, but whether or not the motherboard works properly at 60 MHz, I have not tested this condition. I run my 66 MHz FSB 8433UUD with PCI at 1/2*66.

There were a lot of different versions of the 2940. I would guess some of the early ones might only be PCI 2.0 compliant? The 2940UW is probably a safe bet though.

Chicony CH-881A ver:1.0, based on UM8886F / UM8881F chipset.

Still in great condition:

Convoluted jumpers setup.
Was not able to find which jumper forces WB mode for Am5x86 processors.
Used the simple adapter from this post to enforce it.
BIOS offers settings for L1/L2 cache modes, but the L1 one does nothing.

Not pretentious about level 2 cache chips.

Supports EDO RAM.
Very picky about memory modules for tightest BIOS timings.
Using 2x16Mb selected sticks in slots 3 and 4.
Unstable with bigger amount of RAM (considering tightest BIOS timings).

Used Promise EIDE2300 PLUS VLB adapter.

Voodoo 3 graphics cards don't work with this board.

--- Am5x86 @160MHz (4x40)

All BIOS settings on max.
Ark1000VL requires 1WS for stable system.

SpeedSys:

benchmark results

Not bad, but not great either.

--- Am5x86 @180MHz (3x60) and $200MHz (3x66)

Cannot complete POST no matter how hard i tried.
Disappointing.

--- Am5x86 @200MHz (4x50)

For DOS all BIOS settings can be on max, but for Windows and some of the more complex offline computation tests the next settings must be adjusted:
ISA BUS CLOCK OPTION = PCICLK/4 (best is PCICLK/2 )
L2 CACHE WAIT STATES = 3-2-2-2 (best is 2-1-1-1)
DRAM READ WAIT STATES = 1 (best is 0)
DRAM WRITE WAIT STATES = 1 (best is 0)
HOST TO PCI WRITE = 1WS (best is 0WS)
PCI BUS DIVIDER = 1:1/2 (best is 1:1)

benchmark results

Again, not the best results out there, but not too bad really.

--- P24T @100MHz (2.5x40)

Inherently unstable with POD100.
Works fine with POD83.

If JPR22 is in position 2-3 according to manual performance is very-very bad, but things are more stable.
Setting it to 1-2 brings performance to the expected levels, but the system becomes less stable.

Updated the POD100 extensive summary post with details for this motherboard.

---

Pretty ok at 160MHz with Am5x86, but overall not the best 486 class hardware that passed through here.

------

(Wanted to be done with the 1024Kb level 2 cache 486 motherboards before the end of 2022 but one more is left on the list. That last guy can be potentially special. Symphony Haydn SL82C461/362/465 + VLB. Hope it does not disappoint ...)

pshipkov wrote on 2023-01-01, 01:54:

Asus VLI Revision 2.1 of the board offers jumpers for 3.3, 3.45, 3.6, 4, 5 volts. With one of the CPUs i can get to 200 even at […]
Show full quote

Asus VLI Revision 2.1 of the board offers jumpers for 3.3, 3.45, 3.6, 4, 5 volts.
With one of the CPUs i can get to 200 even at 3.3V but needs 5V Peltier.
The system can get in weird state no matter CPU voltages.

Didnt push Soyo 4SA2 at 200MHz past the brief confirmation that it works since you went over it.
I you feel there is a potential - we can give it a second chance.

Hi,

I only have a VLI Revision 2.0, but still .....

Where do you have a jumper for 5 volts???

Revision 2.0 of of the Asus VLI motherboard does not have these jumpers. Only 2.1.