I really appreciate your work with these Turbo XT builds. There aren’t enough numbers out there.
Lots of anecdotes, but hard data is hard to find (no pun intended).
I really wish there were more 8086 AT boards. I’ve only ever seen one/two but it’d be interesting to throw them in the mix.
NEC V20/V30 are quite interesting. The V20 especially. It is a very “pre-Cyrix” type upgrade and, electronically, a fascinatingly simple idea. Pretty cool that the V20 is really not a “drop-in replacement” for the 8088, being a different sort of class of hardware, but that it works so well in the same assemblies.

Out of curiosity, on these systems, is there really any appreciable difference between a WD90C00-JK and ET4000AX?
Landmark and CheckIt scores seem strikingly close, despite the clock speed and architectural differences at play.

Agreed on your points regarding building an error-free system. It’s practically impossible these days.
In some regards, it’d be nice to have a NuXT-like 486 setup that’s “modern” but made with vintage vintage parts.
The money is where things are really getting ridiculous. Even Pentium III era hardware is getting scarcer—and pricier—by the day.
The VLI and PVI are unbeatable in terms of quality and features. It appears that many can come close in terms of quality and performance, like the SuperEISA, but there are always catches. I don’t think there’s anything wrong with that, but it depends on how dedicated one is to tuning their system.
EISA, for example, has extra demands that must be met. EISA config is easy, but doing it right is important.
VLB is tricky on certain boards, acceptable on others, and flat-out perfect on some.
At the end of the day, these tests prove one thing to me: it’s all about the implementation. Various configurations of BIOSes, registers, &c. are half the battle. The rest is in the combination of the right parts/cards and the right board.

Thanks for the kind words! I have a boxed Mach64 VRAM, and I have the 2MB VRAM expansion module.
It might be joining the ranks soon. I’m getting a bit of an itchy trigger finger when it comes to the unboxing. 😀

Great point on the CPU-bound performance. I want to re-run these tests with a POD100 on a different board. Probably going to see how an SiS471/OPTi895 board does with such a setup. Diamond Stealth64 DRAM T is the gold standard, and the best of all worlds. It’s right on the tail of the ARK in DOS, right on the tail of the 964 in Windows, and easy to use. The only card that might best it is something with a Trio64V+.

I wonder if it’s possible to hack a 0ws BIOS for the Stealth64 DRAM T, like the STB PowerGraph 64V+ BIOS by Madao.

The only truly appreciable difference in terms of all of these cards is the RAMDACs, and that’s where something like a Stealth64 Video VRAM is unbeatable with a 220MHz TVP3026 part. Great for CRT usage, but these days, that’s only part of the equation.

I will say that the image quality on all of these cards is great. The 964, however, is on a different level. Crystal clear picture quality, and it’s noticeable.
That’s where cards like the ARK are a bit of a wash. The Paradise Bali32 is dim and dull; the 964 is bright, colorful, and sharp. Perhaps that’s why I like the ATi Mach cards so much—they similarly provide great 2D quality. The best 2D card for my eyes still must be the MicroLabs Tseng ET4000W32i. Warm colors, soft text, and sharp/clear picture. Would love to find the VLB version of that beast, as my Cardex ET4000W32P is decent, but not quite that good.

Tests will come with a 968, and all of these numbers are going to be re-run on a POD100.
The 968 being about 10% faster sounds right. Someone else in this thread probably knows a bit better.
The 964 really shines at higher resolutions; the Stealth64 DRAM T keeps up until you get there, but then again, what SVGA games run well on a 486?
The Vision models do feel a little more “premium,” offering VRAM-based models.
The 964 is a weird card to me. Interesting that it supported 4MB of VRAM, but seemingly not a great-selling card.
Seems that the 864/968/Trio64 are more “common,” with the Trio64 and 864 being most common.

Truth be told, with how marginal the differences are between the 864 and Trio64, the 864 might be a better buy today.
The Trio64 is undoubtedly better, however.

Regarding BIOS swaps—this is where I think we can all gain the most performance.
Miro 20SV was decent in stock form, monstrous with Diamond BIOS.
Hopefully a replacement BIOS for the 20SD shows up. Would be curious to put the 868 to the test.

Reminds me a lot of the controversy around the AGX-014 cards.
Hercules hacked their BIOS/drivers to perform better on WinBench and caught flack.
Sometimes it seems the BIOS implementation is the key.
Maybe one day a hacked 0ws BIOS will show up for some of these other cards.

Jumping in to say a huge thank you for all the knowledge sharing. My 486 build is ongoing with the strange beast that is the M919.

Per feipoa, these are known BIOS revisions:

9190506.BIN -- 05/06/1996
9191016S.ROM -- 10/16/1996
9190914S.ROM -- 09/15/1998

Do I parse the thread correctly, that the general recommendation for this board is the 10/16/1996 code?

I do not own an EEPROM programmer as yet, but I have some SST 128kb EEPROMs, and I expect that I can probably convince my 3c905 card to burn the PCChips ROM. Reasonable?

@doogie
What happened with Asus VLI ?

I use the 9191016S BIOS for M919. Works well. Tried the rest. The Award one is not good. Cannot remember if there are notable differences with the later AMI one.
If you mean to do the hot-flashing of the BIOS - that will work.

---

@WJG6260

There were no observable difference between AT4000AX and WD90C00-JK that is not in the rounding error on these 8088 motherboards.
That's why tested with ET4000AX. But there is a difference in the 8086 board. Not sure why. Just accepted it.

I keep searching for that perfect motherboard (tm). Identified few so far. Chances for new positive surprises are getting real slim.
But after you opened to EISA subject that made me scan what's in that corner. As a result i spotted one motherboard in the "retro web" database that blew me away. Maybe that's an unicorn.
Hope i manage to obtain one eventually but i know it didn't show-up in the usual e-stores during the last 5-6 years so chances are not looking very good.

Please expand your post with beefy 4Mb Match64 and Vision968 when you can - it will become my go-to place for this class hardware.
Cannot remember seeing before a complete set of Vision/Trio/Match VLB cards rounded-up with a more complete set of tests, or did i miss something ?

Agreed, HW/SW integration is a key.

pshipkov wrote on 2022-10-10, 21:48:

Not sure if new bios will help, but give that a try.
Something else must be going on. To clarify - did you confirm that each and every jumper on your motherboard was setup like the photo-reference i provided early ?

......

Hi again 😀

I am now happy. I bought a XGecu Pro TL866II-Plus and some Winbond W27E512 chips.
And then successfully programmed the famous Feipoa Write Back Fix BIOS aaaaand success!

So I was right: With the 3.06 BIOS the board did not boot at all with an AMD 5x86 in 4x setting.

I still use the jumper settings for Intel DX / DX2 / DX4.
The computer now boots and shows 133MHz (33 * 4), before there was only an empty screen.

So now: 200MHz I come!

@CoffeeOne

Awesome work! Glad to see the VLI is up and running.
Interested to hear what you're able to squeeze out of it--please keep us apprised of your progress.

-------
@pshipkov

I find this to be interesting. The ET4000AX should by all means be superior, but it makes sense that they'd perform identically if the 8088 was limited solely to 8-bit/16-bit ISA transfers and the 8086 could do 8/16/32-bit transfers (or something of that sort)? I am curious if that might be the case; I am not too familiar with the ISA implementation on pre-286 hardware. Does that perhaps sound like it could be the culprit?

There's definitely a perfect motherboard out there somewhere 😁! It's probably hiding in plain sight.
EISA definitely is an interesting avenue of implementations and brings its own quirks. Will add more later, but just to inform those following this thread: I broke out the Mach64 VRAM from its tomb and found that it had some issues with the SuperEISA.
Notably, the board struggled to properly configure the Mach64's memory aperture, resulting in awful DOS performance.
There are EISA CFG files for the Mach64, and yet it underperformed even the Mach32. Troubling, and seemingly implying issues with VLB/ISA/memory implementation on the SuperEISA, whether BIOS-wise or otherwise.

I intend to re-run the numbers with the Mach64 and Vision968 on the SuperEISA, and add the same set of 20+ tests on another chipset, just for overall refernece. My aim is to use an SiS471 board, so that we can standardize and extrapolate a few things: (i) chipset performance metrics; (ii) VLB compatibility/implementation; (iii) CPU performance on a per-board basis; and (iv) S3 accelerator performances on a couple of popular chipsets.

I think there may be some other numbers; I believe FGB once compared Vision864/964 and Vision868/968, so I won't lay claim to being the first to do so. But I hope that these numbers give some idea as to how S3 accelerators stacked up, and what we can expect from them.

-------
Speaking of HW/SW integration, and speaking of VLB implementations and chipsets, a new challenger has entered the ring:

The Culprits:

The Story:
-------
Note: Numbers were limited to DOS interactive graphics due to stability issues in Windows.

EDIT: Windows numbers stabilized and ARK1000VL provided the following results:

ARK1000VL
WinTune2: 7963 KPixels/sec (7906 KPixels/sec @ POD100)

All BIOS settings were set to minimum, all waitstates were disabled, and the FSB was reported at 39.9MHz in ChkCPU.

-------
Enter: the QDI V4P895GRN/SMT1.0.
This board sure is a looker, and it's got a story to tell.

This is a relatively late 486 board, and it shows. The brain of the operation is the OPTi 895 chipset. Dated to 1995, this chipset is something of an oddity. It possesses a high level of integration, being a single-chip solution and seems to be--on paper--on par with SiS471 and UMC8498. Feature-wise, it's--again, on paper--superior, possessing a VLB 2.0 implementation, theoretically enabling better performance on VL burst reads and writes, increasing VL Bus stability, and overall allowing for higher speed operation at 0ws. VLB 2.0 is spec-ed for 66MHz and provides an implementation for 64-bit data transfers, as well. Unfortunately, the OPTi 895's paper prowess does not tell the whole story. This assembly, and all 895-based parts, are effectively interchangeable with OPTi 802-based boards. The 895 is simply an 802 with power management features and (presumably) bug fixes. Moreover, the 895 is not the cache fiend that the SiS471 is; this board--like all other 895-based assemblies--is limited to 512k single-banked cache, or 256k double-banked cache. For interleaving, performance, stability, and overall ease, the latter is the preferred setup on this board. This board can handle 3v CPUs, and with a BIOS update to v1.4 can work just fine with an Am5x86 in write-back mode with 4x multiplier. Do note: the earlier v1.3 BIOS allows either WB or 4x, not both. RedHill seemed to like the board back in the day; it is definitely an easy one to work with, although it's not the friendliest due to its clunky AMI WinBIOS.

The Good:
This board is stable as a rock.…in DOS
Coin cell battery.
At the tightest BIOS timings, it works great... in DOS.
It will take any VLB card/controller/you name it.

The Bad:
It is not stable enough to reach Windows with WB L1 and an Am5x86 at 160MHz.
EDIT: Instabilities in Windows seem due to aggressive BIOS timings, but slashing timings even slightly kills performance.
It seems that the root cause of the instabilities of this board in Windows are related to DRAM being run at 3-2-2-2. Will investigate further.

EDIT 2: Windows numbers stabilized by swapping around cache chips. Turns out that the original chips were fine, even after testing a few sets. Curious, but not necessarily unusual for this unruly, cranky older hardware.

VL master/slave slots show no difference in performance/stability for IDE controllers.

VL EIDE performance measured with a Promise EIDE2300+ v3 yielded excellent results. See below.

The Pentium OverDrive:
During some private discussion with pshipkov, he recommended I try a POD on this board at 100MHz. I bought a POD years ago as an upgrade for an HP Vectra 486/33N, but never was able to make the CPU work due to the lack of a proper BIOS (which I only recently found, ironically). This POD is a special one. It was unused, and for some reason it can handle 100MHz without modification and with only 3v. That is the precisely how I configured this CPU on this QDI board. The 3v undervolt was accidental, and was facilitated by my failure to close JP21; despite under-volting the CPU by 40%, it seemed stable, and I left things as-is.

For those seeking to recreate the same on this board or its derivatives, a note: V1.0 does not officially state support for the POD and does not have any references in the manual regarding WB L1 for a POD. This is easy to manage, as the board has an unmarked, reserved jumper--JP9--which correlates with the P24T WB jumper on the later versions and indeed works just the same here. Actually, the board also doesn't even have settings for a P24T in its manual; the settings, however, are the same as those for the P24D and Am5x86 (barring two jumpers, JP19 (WB) and JP21 (4x)).

The Numbers:

Note: Nice SuperEISA was tested with Am5x86 configuration and tightest BIOS settings. Numbers are from post a few pages back. Orange bar represents QDI w/ POD, and blue w/ Am5x86. Green is SuperEISA w/ Am5x86.

The Verdict:
This board has the chops to be great. It's got great performance, but private discussion with pshipkov yielded the observation that the OPTi 895 is limited in FPU throughput. pshipkov noted FPU-intensive apps like Quake fell behind and we together surmised that, while the VL-graphics-intensive transfers demonstrated excellent performance and a better implementation than that of the SuperEISA, it is clear that some performance and throughput was left on the table with the high integration. Compared to the SuperEISA's 406/411, comprised of 5 ICs, this board has 2 ICs and one is simply an OPTi602A, implementing an RTC and a few other features.

A good board. Fast, stable in DOS and Windows.
Limited by its design, but well-implemented.
Nice for accommodating longer VLB cards and ISA cards due to socket location.
Quirky, but neat. Certainly a "late" 486 board.
Has some undeniable creature comforts, but also some notable drawbacks, discussed above.
OPTi 895 is not bad. Certainly closer than farther from SiS471 and UMC8498.

WJG6260 wrote on 2022-10-29, 01:15:

@CoffeeOne […]
Show full quote

@CoffeeOne

Awesome work! Glad to see the VLI is up and running.
Interested to hear what you're able to squeeze out of it--please keep us apprised of your progress.

[...]

Hi,
I now tried all my (lately increased the number) 9 Am 5x86, but none of it posted at 50*4. Euphoria mode off 🙁
Most likely all of them work at 160MHz, so I have only the options 160 and 150MHz.
Of course from CPU performance point of view 160 is better, because it will also allow fastest timings with the cache, but I like external 50MHz, it is a challenge for the VL cards and the mainboard....
I will - as a last attempt - try to increase the voltage to 3.6V (there is only 3.45 and 3.6 on VLI Rev.2.0)

Also interested about 50MHz on your late Opti board, but I guess impossible for a POD 83, so you would need an Am5x86 or DX4-100. By the way, if I really go for the 150MHz route an Intel DX4-100 &EW would be nice, too. Can they do 150MHz? Maybe very unlikely, too; it's the same 50% overclock as 133 => 200.

UPDATE:
I tried all with 3.6V and again 50*4, none of it posted, so no "jackpot silicon" for me.

But for those 2 the monitor was turned on, so for a very short time there was a video signal at least.
So newer is better? Those 2 are from the year 2000 I believe.

UPDATE of UPDATE:
After reading more posts in this thread, it seems that 200MHz is impossible with 3.6V. .....
Is there a possibilty to force 5Volts on a 3V Am5x86 CPU on the VL/I-486SV2GX4? I am not really sure, if I want to do it though, poor CPU 😁

@CoffeeOne

Sorry to hear you didn't hit the jackpot.
It is improbable to find non ADZ/ADW models capable of 180/200MHz on air cooling.
So far i stumbled only one one BGC model that can do 180MHz on air - sort of kind of.
Unfortunately Asus VLI does not offer 60MHz base frequency.
Still, the chances of 200MHz on VLI are much higher than most other boards for some reason.

Why do you want to try Intel DX4 at 150MHz if you can achieve it without any problem using Am5x86 ?

I think you have revision 2.0 of the motherboard.
It does not offer jumper settings for 5V.
Manual talks only about JP32 which switches between 3.45 and 3.6 Volts.
There is an updated manual in electronic form only that talks about JP32 and JP33 that offer 3.3, 3.45, 3.6, 4 Volts.

Revision 2.1 has 6-pin jumper that seems to cover 3.3, 3.45, 3.6, 4, 5 Volts.
Search around - there are threads around here about that stuff, but cannot remember details anymore.

---

@WJG6260

Weird stuff with Match64. Not sure what to think about it.

Good info. Thanks for sharing.

You can potentially improve on the instabilities in Windows by finding better set of L2 cache chips.
This is usually the main problem for achieving complete stability, especially when the system is running within its specification.
But even if you don't find the time to rotate chips around to get it to that point, the shared info is pretty good and defines the perf characteristics of the motherboard pretty well.
When i hear OPTi the first thing i think about is compatibility but this one handles DOS interactive graphics really really well.
As you said - lacks in CPU intensive tasks. I am sure if you run 3D rendering test it will underscore that.

What is the local storage perf of the QDI board ?
You skipped the SpeedSys HDD test there.

Do you have your SuperEISA tested with POD100 ?

Having a P24T unicorn running at 100MHz at 3.3V on its own - that is nice.

pshipkov wrote on 2022-10-29, 19:08:

@CoffeeOne […]
Show full quote

@CoffeeOne

Sorry to hear you didn't hit the jackpot.
It is improbable to find non ADZ/ADW models capable of 180/200MHz on air cooling.
So far i stumbled only one one BGC model that can do 180MHz on air - sort of kind of.
Unfortunately Asus VLI does not offer 60MHz base frequency.
Still, the chances of 200MHz on VLI are much higher than most other boards for some reason.

Why do you want to try Intel DX4 at 150MHz if you can achieve it without any problem using Am5x86 ?

I think you have revision 2.0 of the motherboard.
It does not offer jumper settings for 5V.
Manual talks only about JP32 which switches between 3.45 and 3.6 Volts.
There is an updated manual in electronic form only that talks about JP32 and JP33 that offer 3.3, 3.45, 3.6, 4 Volts.

Revision 2.1 has 6-pin jumper that seems to cover 3.3, 3.45, 3.6, 4, 5 Volts.
Search around - there are threads around here about that stuff, but cannot remember details anymore.
[....]

The Intel DX4-100 write-back is a bit faster than the AMD 5x86 clock by clock. But I agree it is not worth the effort for hunting one that can make 150MHz (for a very few % more performance).
I started playing a little bit with the board (installed MS Dos 6.2 on compact flash) and copied sysinfo, hwinfo555 and speedinfo over from floppy.
But I already regret my own statement, like "external 50MHz is so nice": I installed DOS from floppy with a cache timing read 3-2-2-2 and write 3 (so slowest). After installation I changed cache timing to 3-1-1-1 and: error with himem.sys external memory test. OMG. Maybe my cache chips are not the best.
Maybe I should use 160 like everybody else 😉 and use doubled banked 256kB cache config as a starting point. I don't know.
Next step should be installing Windows 98 (when that runs I am confident that the system is stable).
I am also missing a compact flash reader on my main workstation, so it is difficult to move data to the VL/I system now. I used to transfer data to my 486 PCs via network (FTP), but ok, so I first need Windows 98 installed as mentioned above.

About 5V. The question is if the CPU voltage detection on the mainboard could be easily disabled. But I think 200MHz is out of my scope anyway, as I don't want to use Peltier.

@CoffeeOne

I am not sure about 150MHz on the intel part. It might not be possible.
Most struggle at 120MHz, and without a Peltier, 150MHz seems unlikely.
Clock-for-clock differences between the intel part and Am5x86 are not that drastic afaik, likely within margin of error.

------
@pshipkov

Mach64 is annoying, for sure. Working on a few software solutions and considering alternatives.
It seems that late model VLB cards were sort of unique; lots of instabilities/incompatibilities, not unlike the SpeedStar64 ISA and various 386 boards.
Seems S3 had everyone beat there--compatibility and performance; the best of both worlds!

Great idea on the cache chips. I did not think about that, but will give it a go with some faster ones.
Unfortunately, I do not have 12ns parts, but can try and source them sooner and revisit this board to see.
I think you're right on the money that OPTi is normally associated with compatibility. 895 is a late enough model that things got interesting.
895 is sort of the gateway to OPTi's later weirdness with Pentium/VL chipsets, which I hope to cover later.
It's a shame they never had a PCI 486 chipset. That could've been interesting.
Agreed re: DOS interactive graphics. It's nice to see that this board can do so well, and prove that OPTi made some good parts.
I need to stabilize the board to get 3D Studio running; I can't seem to get into the software and instantly experience crashes, likely due to the aforementioned L2 issues.

Will add some storage perf metrics.
Only ran tests with an ISA adapter, but will grab a VLB one and get some metrics.
Should be interesting to see how it does.

Unfortunately, I can't get the SuperEISA going with a POD100, as the SuperEISA v1 is a PGA168 ZIF board.
Unless there's a way to adapt the extra pins of the POD and get things going, I will have to use another EISA/VL board to compare.

I wonder why my P24T was able to do 100MHz on 3.3v.
Not questioning things and knocking on wood hard, but more so curious.
Could these parts degrade with age? Were the later revisions more capable silicon?
Do you know the datecode of yours?
Sorry about the barrage of questions!

WJG6260 wrote on 2022-10-29, 20:04:

[...] I wonder why my P24T was able to do 100MHz on 3.3v. Not questioning things and knocking on wood hard, but more so curious […]
Show full quote

[...]
I wonder why my P24T was able to do 100MHz on 3.3v.
Not questioning things and knocking on wood hard, but more so curious.
[...]
My 2 cents about this. When you are able to set the voltage to 3.3V manually (not possible with VL/I-486SV2GX4, because there is auto-detection), you are not really "undervolting" the CPU. It's more like the voltage regulator on the POD is doing nothing then. So normally 3.3V are created by the mounted voltage regulator. But still a good idea, because there is less heat on the CPU.

Sorry for answering to my own message. After the disappointing setting with external 50MHz, I now switched to the setting 4 times 40MHz. I also configured the CPU type to 5x86 L1 WB.
All timing settings set to fastest setting, upcoming Windows installation test will show stability....
Before I continue with meaningful tests, I just made a norton si test (not very meaningful) and a speedsys test. Please ignore the harddisk part from speedsys for now.
I guess that are the expected values, am I correct?

Yes I'd say you're right on target for 4x40. I found better overall system performance with 3x5o though

@WJG6260
You don't need faster chips for 40 or 50 MHz.
15ns rated ones are more than enough.
Just find the less flaky chips in the pile that will allow you tighter timings.

I also noticed that on some motherboards POD100 works just fine at 3.3V.
Maybe the mobo figures there is an overdrive chip installed and provides the necessary 5V.
TBH, i am blanking at the moment about all this.

@maxtherabbit
In everything i have seen so far 4x40 is always faster than 3x50. This class hardware is hard CPU bound.
Where and how you see the opposite ?

pshipkov wrote on 2022-10-30, 05:22:

@maxtherabbit
In everything i have seen so far 4x40 is always faster than 3x50. This class hardware is hard CPU bound.
Where and how you see the opposite ?

basically everything that is not quake, massive improvements in memory, video and disk throughput

keep in mind I'm not using the PCI divider

I think i know what you are talking about, but wouldn't call it massive.
Games/apps that are CPU/FPU-light can reach a point where the PCI/VLB bus becomes the bottleneck.
Things like Wolf3D and Doom for example.
For the rest there is either no difference or arbitrary small advantage for 4x40.
The biggest benefit for 3x50 in some cases is for local storage metrics.
Can result in something like +1Mb read/write speeds compared to 4x40.

---
@CoffeeOne
Looks good.
One small note - successful Windows 9x installation is not indicative for system stability.
But since 4x40 is well within the normal operational boundary high-chance everything will just work.

Picked up this socket 3 board marked PCI400C-C. It supports 1024k cache and 60/66MHz fsb if you add a header to JP5. Too early in testing to give an opinion just yet, so far not promising. If anyone has any clues/ideas on how to get L1 into WB mode let me know.

I believe these are the jumper settings:
https://stason.org/TULARC/pc/motherboards/J/J … -PCI400C-C.html

The instruction mention to close J7 for an AMD 5x86, curiously there is no header on this jumper. If I jump it manually it won’t post.

this one looks very interesting.
looking forward to what comes out of it, despite your warning hint.

forcing L1 cache in write back mode:
Re: 3 (+3 more) retro battle stations
Re: 3 (+3 more) retro battle stations

i am going to post info about 2 more turbo xt motherboards and then will share stats/benchmarks for 5 486 boards with 1Mb L2 cache, so your data will be a nice addition.

WJg6260 has bunch of stuff to share as well, from what i gather.

pshipkov wrote on 2022-10-30, 17:54:

[...] @CoffeeOne Looks good. One small note - successful Windows 9x installation is not indicative for system stability. But sin […]
Show full quote

[...]
@CoffeeOne
Looks good.
One small note - successful Windows 9x installation is not indicative for system stability.
But since 4x40 is well within the normal operational boundary high-chance everything will just work.

Hi, sorry for jumping around, but ....
A few days ago I got a complete PC with an Asus PVI-486SP3 board in it. Seems to be working, only the CD ROM drive is dead.
Do you know if there Is there a 60MHz or 66MHz option?
Board revision is 1.2 and the clock chip is a VIA VT8228.

sounds pretty good.
Asus PVI mobos are great, even the early versions.
Where did you find it ?

pshipkov wrote on 2022-10-30, 21:30:

sounds pretty good.
Asus PVI mobos are great, even the early versions.
Where did you find it ?

A former working colleague (now retired) hoarded stuff and he knew that I like 486 computers. He gave the computer to me, I just had to pick it up.

I found in the thread "Test:486 undocumented jumper settings" here on Vogons
there should be a 48MHz setting (?)