I have been tracking the different chipset permutations in this thread, preferably in powerful looking motherboards.
ALi stuff in particular has been a point of interest to me because of how volatile it is in the different assemblies.
For example, M1429 is present in some of the worst and best 386 mobos.
Similarly for the 486-class M1489, including some ridiculous boards such as PC-Chips M918i.
Naturally wanted to check the in-between version used in this MSI-4143 motherboard.

Tried some hotkeys with the AWARD bios but nothing worked.
The AE95E bios is from AMI. Everything is fine there - able to adjust timings and other settings.

I dont know why, but the ALi 486-class chipsets dont like p24t processors.

Xinetron X/LAN Xi-386/486-SC rev.A based on Xinetron X/LAN Xi-210-B also known as USA SC-9204-A
A riddle of motherboard.

A powerful looking 386/486 hybrid assembly:
- plenty of CPU/FPU sockets
- the two VLB slots are especially interesting
- two extra level2 cache chips sockets suggest write-back capabilities
- CAN-4 crystal oscillator socket makes it easy for overclocking
- limited number of jumpers is always a plus
- an unknown (to me at least) X/LAN chipset - what can it do?

While my expectations were high, the board came hard dead - plenty of damage in the upper left corner from an old battery leak.
Repaired corroded rails on both sides, replaced elements and sockets, replaced keyboard controller and connector. Scrub.
Monitor eventually lit up, much to my satisfaction after long period of iterative fixes.

The board came with Intel 486DX-33 CPU, 32Mb 70ns rated RAM (8x4Mb) and 20ns rated 128Kb L2 cache.
After confirming the system's basic functionality with the existing setup, my curiosity turned to exploring the full potential of the 386 hardware paths it offered.

This motherboard is not documented anywhere so took a moment to figure out what the few jumpers do.
JP1 (486SX/DX) is obvious.
JP12 (NPX IN/OUT) does not make any difference.
JP13 (486-50IN) does not make any difference.
JP20 (U42 Disable) does not make any difference.
JP21 (386) is obvious. Must be closed for 386 processors and open for 486 ones.
JP17 (VESA) is obvious. VLB wait states. Didn't explore them. Notes below.

EDIT
About the chipset - a post i made a while ago in the old hardware giveaway thread shows a motherboard that is almost identical to this one here. The main CI on it is "USA integrations" instead of Xinetron X/LAN. Quick search in TheRetroWeb for motherboards based on "USA i" chipset revealed this 3 boards of the same design. One of them had the entire manual scanned.

Xinetron X/LAN Xi-210-B is relabeled USA Integration SC-9204-A
/EDIT

Plugged a 386DX-40 processor, closed JP21, inserted Trident 8900C ISA video card, standard IDE controller with CF card, 80MHz crystal oscillator.
Default BIOS settings.
Lights on, everything seems to work.

It is worth mentioning that there is a discrepancy in how the base frequency is derived from the crystal oscillator - it is divided by 2 for 386 processors and used as is for 486 ones. First time seeing this kind of setup.

Naturally the first thing i wanted to try was to see how VLB graphics cards work with 386 processors.
This quickly went nowhere.
I usually start with Ark1000VL and S3 Trio64 based adapters but both of them didn't produce an image this time. The speaker started beeping some sequence upon power on which indicated that not all is well. Fetched few more cards from the stash to test with. Still no go. Eventually tried pretty much each and every card from this extensive list. With the exception of ATI Match32 and Cirrus Logic GD-5428 cards none of the rest worked. Even the two that produced a picture made the system very unstable.
Switched to 486 CPU to see how that goes - everything just worked out.
Story was the same for VLB IDE/EIDE controllers. Didn't check VLB SCSI ones.
So, it looks like this board has some major integration issues with 386 processors and VLB cards.

Ok, fine. Had to step down to ISA extension cards.
Started with Cirrus Logic GD-5434. Didn't go well - no video signal.
Tried ET4000AX and W32i - no video signal or highly unstable.
Finally, a Cirrus Logic GD-5428 based card (Diamond SpeedStar Pro) enabled the system to operate stably.
Tested several more adapters - turns out that anything faster than 5428 is a no-go. At the same time all Trident and older models Cirrus Logic cards, as well as Headland/G2, OTI, etc., worked fine.
This is not related to jumper and BIOS settings but is prevalent.
Ok, CL GD-5428 will be used for testing going forward.

Standard ISA IDE controller worked right away already.

Next up - level 2 cache setup.
The board came preconfigured with 128Kb single bank 4x32Kb 20ns and 1x16Kb 15ns. That's ok but 256Kb buffer is strongly preferred.
That didn't go anywhere either.
Populated the sockets with trusted 9x32Kb chips.
There are 5 jumpers to configure the level 2 cache.
Spent good amount of time testing all possible configurations.
Varied the TAG chip - 16/32kb - just in case.
Multiple passes with different set of chips.
No and no.
Tried with 486 CPU hoping for a VLB cards -like pattern. Not this time.
It is safe to say that no stone was left unturned, so with decently high level of confidence i can say that the board does not support 256Kb level 2 cache.
Set things back to the original settings for 128Kb buffer size but replaced the slow 20ns chips with trusted 15/12 ns rated ones.

EDIT
From the found above scanned manual i confirmed the jumper configuration required for 256Kb level 2 cache.
After quite a bit of musical chairs with bunch of chips finally managed to get a working set. My previous attempts failed because the board is quite picky and requires specific chips in specific sockets, does not just take any chips anywhere. This combined with me guessing the jumpers config prevented me from succeeding.

Moving to 128Kb to 256Kb level 2 cache resulted in minor performance improvement only.
/EDIT

At the same time populating the two WB cache sockets worked instantly and improved performance by few percents.

16Mb (4x4Mb) 60ns rated RAM.
Cyrix FasMath 40MHz rated FPU (black top).

BIOS is decently advanced but within the 386 standards. Offers about 10 parameters for L2 cache / memory wait states, ISA bus frequency divider, memory remapping, shadowing, etc. The usual stuff.

--- IBM BL3 at 96MHz (2x48), ISA at 12MHz

Complete stability in Windows required to step down to 90/45MHz.

All BIOS settings on max except:
CACHE READ WAIT STATES = 1 W/S (best is 0)
BUS FREQUENCY SELECT = CLKIN/4 (best is /3)

Board is picky about RAM - requires good 60ns rated chips or system is unstable.

There is something weird with the level 1 cache support of the BL3 processor.
First clue is a discrepancy between what SpeedSys and CCT report.
Notice how SpeedSys does not show the expected higher throughput from level 1 cache but flat lines for the entire level 2 cache range.
At the same time CCT shows the peak.

EDIT: Forgot to retake the CCT screenshot with working 256Kb level 2 cache.

This seems to be the reason for low test results from PC Player Benchmark (4.1 fps), Doom (7 fps - ouch), Quake 1 (2.1 fps), WinTune2 (~1000 KPixel/sec). Some of these are worse than bare 386DX running at 40MHz.

REVTO486.SYS hangs no matter what but LIGHT486.SYS seems to be working. This is unusual.

Not sure what to think about this but the BL3 support is clearly lacking.

--- Cyrix 486SXL2 at 45MHz (1x45), ISA at 11.25MHz

Anything above this frequency results in unstable system.
This is a bit surprising since the motherboard was happy at 48MHz base frequency with the BL3 CPU.

All BIOS settings on max except:
CACHE READ WAIT STATES = 1 W/S (best is 0)
BUS FREQUENCY SELECT = CLKIN/4 (best is /3)

This just worked. Nothing more to say really.

SpeedSys hangs which is common for many motherboards/SXL2 combos.

Intermediate performance.

--- AMD 386DX at 45MHz (1x45), ISA at 11.25MHz

All BIOS settings on max except:
CACHE READ WAIT STATES = 1 W/S (best is 0)
BUS FREQUENCY SELECT = CLKIN/4 (best is /3)

Performance is lacking - below medium compared to the bulk of previously tested motherboards.
Nothing more to say really.

---

benchmark results

I genuinely believed this motherboard would be incredible, but that turned out not to be the case. However, it is quite an intriguing assembly built around a lesser-known chipset, filled with peculiarities and limitations that provided me with an enjoyable challenge to explore.

Ouch. How many hours of your life did this board consume? Isn't it best to avoid motherboards which sport, both, a 386 socket and a 486 socket?

pshipkov wrote on 2025-02-20, 08:27:

[ This motherboard is not documented anywhere so took a moment to figure out what the few jumpers do. JP12 (NPX IN/OUT) does not […]
Show full quote

[
This motherboard is not documented anywhere so took a moment to figure out what the few jumpers do.
JP12 (NPX IN/OUT) does not make any difference.
JP13 (486-50IN) does not make any difference.
JP20 (U32 Disable) does not make any difference.

Would JP12 enable/disable an FPU (NumericProceXor)?
JP13 - maybe to do if using af 50MHz crystal for a DX-50?
JP20 - should be some IC on board called U32, that is enabled or disabled?

[Edit] It seems the text at JP20 is actually U42 rather than 32.?

@H3nrik V!
JP13 - Probably but didn't check.
JP20 - yes, U42 - fixed it above. Yes - the U42 component is a PLACE16V8H. Didn't check if the jumpers are connected in any way to the PAL, but simply switching them around results in no functional/perf difference.

@Feipoa
Actually, i had quite good time with this motherboard. I wanted it to be a success story but that was not the case. Instead, i was sent on an amusing journey.

Just curious when you did the VLB testing was that done with a real intel 386 or one of the meme ones?

yes, cross checked all possible combinations, including AMD 386DX-40 cpu running at 33 and 40 mhz, with and without fpu, with and without l2 cache, many different jumper and bios configs.

Smells like a lemon.

The most probable chance of improvement in the realm of 386+VLB is to remove the PLL on a AL486V-D and play with clock generator alternatives. It's on my list, but boy oh boy is time scarce as time grinds on.

My AL486V-D board is NOS so very hesitant to mess with it.

Btw, this post of mine reminded me that once i had a very similar board, which led to the recognition that Xinetron X/LAN Xi-210-B is relabled USA integrations SC-9204-A. From there i found few more motherboards of the same base design in The Retro Web . One of them had scanned manual provided. And from there managed to bump the level 2 cache from 128Kb up to 256Kb.
Updated the original post.
Anyway, still a lemon.

feipoa wrote on 2021-08-08, 07:34:

In looking at my notes in the Gigabyte GA-5486AL, I see there are working 50/60 Mhz settings, but for 66 MHz, I have noted that it won't boot, but there is a jumper for 66 Mhz. There is also a jumper for 80 MHz. I have noted that the system is fixed on write-back L2 mode and there is no write-through option.

Can you please share the jumper settings for 50/60 MHz bus on the GA-5486AL?

Thank you

I scanned the relevant notes from my manual.

I believe this was the conclusion for 50/60 MHz:

Below are some of my notes when testing all other jumper combinations. Don't ask me to make sense of it (too much time has passed):
[attachment=-1]GA-5486AL_test_all_jumpers_notes.jpg[/attachment]

feipoa wrote on 2025-04-21, 10:02:

I scanned the relevant notes from my manual. […]
Show full quote

I scanned the relevant notes from my manual.

I believe this was the conclusion for 50/60 MHz:

The attachment GA-5486AL_60-60_MHz_FSB.jpg is no longer available

Below are some of my notes when testing all other jumper combinations. Don't ask me to make sense of it (too much time has passed):
[attachment=-1]GA-5486AL_test_all_jumpers_notes.jpg[/attachment]

Thank you, I really appreciate this.