Ever since i checked the Chicony CH-471B motherboard i felt uneasy about how flaky it was.
Spent some more time on it over the last days. Good i did that. I knew i was missing something there. Turns-out this mobo is actually pretty badass. Revisited the post. Updated the Am5x86 @160MHz section. Will go over 180, 200 and POD100 configurations soon.

It looks like the GX4 kicked it’s ass in every test, am I looking at the wrong thing?

I didn't expect CH-471B to outdo Asus VLI and PVI at their game, but sometimes the numbers don't tell the entire story.
I am still figuring it out now, when things are working as expected, but i can confirm already that this board is finding its place on my short list with notable 486 class hardware.
Apart from being kind of unique assemble, and despite its disagreement with Ark1000VL, it feels really good playing with it.

Some dump of thoughts on your illustrious Chicony CH-471B follow:

Based on what you wrote previously, I am surprised you continued testing this board. You wrote that you need to leave the ISA bus at 7.159 MHz or it misses key strokes. I guess you thought this an acceptable sacrifice because you'd be running IDE and graphics off the VLB? What's on ISA - sound and ethernet maybe?

You write that it doesn't work with L1 in write-back mode. That would be a show stopper for me.

You had a hard time getting 1024K working and had to run through your cycles of modules for it to work. That's not a great sign and hints to marginal stability. ARK 1000VL not working. That's not an issue because if you want any Windows GUI, you'd be using something else, but then you said Trio64 no go too. Does that Trident 9440 have any decent Windows GUI compared to the other brands?

Those ARK cards are really just for DOS max benchmark numbers and if you want primarly a DOS system.

No EDO - that's expected. Do any ISA/VLB-only boards work with EDO?

The onboard Adaptec IDE VLB is a nice novelty, but this doesn't really alter one's decision because there are two VLB slots and you can add whatever for IDE/SCSI.

You said you had to use your curtailed magic 16 MB modules, otherwise no POST. That's another red flag.

Keyboard controller can make the system hang 'sometimes'?!?! Oh man, I'd have stuffed this board in my bin by now.

Working at 200 Mhz with a VLB system is a nice niche to the board. Same for 180 Mhz. From what I've read, these would be the only reasons to make a system out of the board, assuming the Asus VLI system cannot do these speeds.

It is a sexy looking motherboard, that's for sure. Hands down it would win the Ms. VLB award.

pshipkov wrote on 2022-12-17, 05:11:

A small comment - CL GD-5434 ISA is the top dog in ... well ... ISA VGA/SVGA.

I can confirm that CL GD-5434 VLB is sick fast card and loves 40MHz and 50MHz VLB BUS clocks. I have two of those but the other one broke down. It didn't always give picture on boot and after some time no picture at all.

@dj_pirtu
Agreed.

@feipoa and @Chadti99
Spent more time on CH-471B.
Made it work at 200MHz (4x50). Render me really really impressed.
So far the only SiS 471 based assembly that can do it.
Also, the board takes Ark1000VL if LOCAL BUS READY = SYNCHRONIZED.

Check the original post. The 200MHz section looks very different now.

To answer some of Feipoa's questions.

What's on ISA: Sound and and LAN cards. Nothing that perf critical.

Graphics adapter: Check my notes on Ark1000VL.
In a nutshell Ark1000VL requires CACHE BURST READY = 2T (best is 1T) and LOCAL BUS READY = SYNCHRONIZE. This is less than ideal for 4x40MHz - overall perf is lower than maxed-out BIOS with S3 Trio64. At the same time these two parameters are mandatory for 200MHz system, so Ark1000VL is back.

RAM size: 4x40 and 4x50 takes selected 32Mb modules.
Tested with single SIMM module most of the time, but ran complex compute with 2x32Mb - works fine.
At 3x60MHz had to use a selected 16Mb module otherwise things don't go well.
In general, performance above 160MHz is limited by the lacking qualities of FPM RAM.

The only real hardware problem of the motherboard is a missing jumper to force WB mode on the level 1 cache for Am5x86 processors.
Tried bunch of different configurations without success. It is possible to be some unholy combination of different jumpers, but at this point i doubt that.
Still, the simple interposer does the job just fine.

Needing to use synchronize is a common problem with VLB graphics cards. I'm surprised you didn't test this requirement earlier.

I thought L1 write-back support was contingent on the motherboard's BIOS and PCB being wired to work with L1 write-back? Simple interposers with a write-back jumper, aren't they just setting a CPU's WB/WT# input pin to VCC? Your last statement indicates that WB is working properly. This is surprising. I remember some posts with EISA/VLB setups with fancy interposers but without L1:WB support. If it is indeed true that you just need to set WB/WT# to Vcc for L1:WB to work perfectly on this board, it would be easy enough to add such a jumper to the motherboard.

Usually SYNCHRONIZE itself is enough but in this case CACHE BURST READY = 2T was also a requirement, so initially i had no idea about it.
Only after ARK1000VL worked in 200MHz i was like "why it does not at 160MHz ?" and quickly discovered the dependency.

I need to look some more at the Am5x86 L1 cache WB/WT business to better understand what exactly is at play.
This setting appears in BIOS only if Intel Pentium Overdrive CPU is present, but does nothing. One of the jumpers must be set in a specific position for the POD to operate with L1 cache in WB - in this case system is unstable.
For Am5x86 BIOS offers only L2 cache setting which functions properly - L2 cache WB/WT mode can be switched from BIOS.

The simple interposer is how you describe it - simply connecting the corresponding to VCC.
The mod i am thinking is to just wire the CPU socket pin to VCC on the back of the PCB and done.

I'm not surprised the L1:WB setting does nothing for the POD. I think Intel kept changing their requirements and many board manufacturers were scrambling to re-adjust. At some point, they had to push their product though. No matter, 200 MHz on a VLB system is more interesting than another POD-100.

Ok, one last note on Chicony CH-471B.
Got it properly working at 3X66MHz.
Performance is impressive.
Updated the original post with details.

Brief summary:
The fastest ISA/VLB system to date. That's for sure.
Not the fastest clock-to-clock motherboard but its capability to overclock to 3x66MHz without bus divider place it in a class of its own.
It is starting to take on LuckyStar rev:D in the 200MHz department.
Of course LSD can go much further on a narrow overclock but we are talking complete stability here.

Finally something that outdoes Asus VLI.
VLI is the better integrated assembly overall, but man, it does not stand a chance with CH-471B hitting the 200.

In addition:
Fastest Wolf3D score to date.
Fastest Doom score to date.

Any chance you have an old floppy drive you can see if it works on this board at 3x66 ?

The requirement for complete stability at 3x66 is 16 MB DRAM? That's disappointing. Which timings are stable with 64 MB DRAM?

floppy drives work in all configurations.
tested with gotek. i have an authentic FDD but it is somewhere in the garage.

DRAM timings - there is a bios screenshot in the original post.
2 parameters lowered from their optimal values.
too bad sis 471 does not support EDO RAM, things would be even better.

Going to search some more for 32Mb FPM sticks that can handle the 66MHz ...

Interesting find pshipkov! Did 66x3 require a peltier? What CPU voltage was best?

Something I was thinking of that would take this thread to the next level would be sharing jumper settings, as tedious as it would be, such a pain sometimes to figure these out.

Motherboard offers only two CPU voltages - 3.3V and 5V.
The presence of Peltier depends on the processor used.
At the moment i have 3 Am5x86 chips that can handle 200MHz on air.
2 of them can do it at 4V. Actually one of them can do it at 3.45V, but not fully stable, anyway.
They can operate at 5V as well, but over time, especially in warm/hot days, they will eventually hang, or at least cause some tests to misbehave or quit unexpectedly.
If i use one of these processors a Peltier element is necessary.
If i use the third CPU which needs no less than 5V for 200MHz, it operates stably and does not require Peltier.
Hope this makes sense.

To Feipoa about RAM requirements - figured i should further clarify.
For DOS games and casual computing tasks (in both DOS and Windows) you can use 24-32Mb RAM and will never notice that something is off.
It takes additional probing to find the instabilities. For example - Watcom C can hang or quit early during compile, offline 3D graphics tasks get flaky, sometimes heavy multitasking in Windows can lead to a BSOD, the participation of IE4/5 increases the chance for that.

I have been thinking for a while to scrub and reorganize this thread.
Similar to how all POD100 benchmarks are now in a single post instead of small snippets in the individual posts, i will do the same for all inspected 286, 386, 486 (160, 180, 200 MHz) mobos, as well as the bunch of Dual Slot1/2 data. Also, there are couple of fat posts about VGA and IDE comparisons that will need to be grouped together as well.
Then split the dictionary in the first post by hardware class with the "summary" post at the end of each.
Right now i need to search and open multiple posts to superimpose the scattered results and findings.

To your point about jumper settings - i have been thinking about this as well.
Most of the shared pictures of the items are indicative for jumper configs for the inspected processors, but sometimes multiple processors are tested and that is not represented.
I have photos for many of the jumper settings for the different motherboards, but not for all.
Can add them as a following step after the reorganization.
Actually i already did the work for 486@160MHz which had the most entries, but want to post data for 2 more 486 motherboards with 1Mb L2 caches first.

You’ve got to let me in on your secret for finding 200MHz air capable 5x86 😀 .

Have you tried one in combo with an ark1000 at 50x4 on the VLI? Use these settings:

“Local BUS Ready had to be changed from Transparent to Synchronize. Also Cache Write Cycle needed to be 3T to boot to DOS.“

I don’t think the VLI supports 5v in combination with the correct jumpers for the AM5x86 but I could be wrong. I’ve not been able to test these settings thoroughly because the one air capable 5x86 I have requires 5v.

No secrets on this one.
About 20-30 of these Am5x86 chips passed through here over the years. 3 had the right silicon for 200MHz on air.

The positively unexpected turnout with CH-471B reminded me that we left the VLI@200MHz business kind of unfinished.
Will give it a proper try in January and see what's there.

About VLI voltages - going by memory here - revisions 2.0 and 2.1 can supply 5V to Am5x86.

I don't know who got what for Christmas, but one of the kids gave me this.
Now i need to figure-out if there is a hidden message, or it is all based on genuine love for the parent (and his peculiarities ... ahem hobbies).

(custom print with the picture of the actual 386 rig)

Haha, that’s pretty cool, merry Christmas!

Nothing says love more than such a t-shirt! Merry Christmas!