So looking up the clockgen datasheet, it shows the clockgen supports:
20
25
33,33
40
50
60
66,67
80

I think these would all be made available if I put an actual jumper at the JP27 position.

JP24 is for something else. Looks like it jumpers to ground or not depending on the position.

MWAHAHAHAHAHAHA.. success!

I have at least a working 40Mhz fsb now.

Also have another setting that has the POD83 showing as 40Mhz.... guessing the BIOS isn't programmed for the different FSB settings that were not officially supported.

I now need to hook up a floppy and CF card and run some tests.

First time I powered it on after installing the jumper blocks it blew a Tantalum capacitor right next to the AT power connector. Stole a newer one off of a scrap modem module.

Let's see if this motherboard can come from reportedly sucky performance to good performance.

The DS12887 battery is dead so BIOS settings aren't going to save after power off but I already ordered a replacement from Glitch Works that takes a coin cell battery.

Well... I have not been able to get the L1 or L2 cache to work on this board. The POST info screen shows the L2 cache but cachechk speeds are definitely without cache and with the POD83 I only ever get a speedsys score of 8.58.... uggghhhh.

I've tried:
Intel 486 SX-33
486OD-75
POD83 - won't even POST if the L1 cache jumper is set to Write-Back.

Evergreen 5x86 133 (AMD 5x86)

I also tried the latest "beta" BIOS (0205) as well as the BIOS that came with the board (0202) because I thought maybe the BIOSes available from ASUS were only for the v1.3 board.

Could the Dallas battery being dead cause this?

Surely ASUS didn't release a board that the L1 and L2 cache were non-functioning... right?

Edit:.... I'm an idiot. I didn't have the Turbo jumper on. Caches are now enabled... 🤣

So after testing the jumpers and the POD83 and an Evergreen 5x86 I have a few results.

JP24 - stock is hardwired 2-3. Setting it to 1-2 seems to maybe be changing the reference clock????? At least it seems like it. On the one 16.66Mhz fsb setting, it looked like it changed it to 33.3 but I have not been able to get it past POST with JP24 in the 1-2 position so I can't really tell what it is doing.

1JP25  JP26  JP27
21-2      2-3     2-3   = 16.66Mhz 
32-3      2-3     2-3   = 16.66Mhz
41-2      1-2     1-2   = 25Mhz
52-3      1-2     1-2   = 25Mhz
61-2      2-3     1-2   = 33Mhz
72-3      2-3     1-2   = 33Mhz
81-2      1-2     2-3   = 40Mhz
92-3      1-2     2-3   = 40Mhz - unstable and was causing graphical corruption.

The Evergreen AMD 5x86 is not entirely stable at 160Mhz with all timings set to minimum. Will have to explore that later. Maybe a regular AMD 5x86 would be more stable.

Will do more testing later but so far Write back works on the AMD 5x86 as well. Write back in the POD83 raises sysmark scores significantly AND the RAM Move speed test skyrockets compared to the Write through mode.

I'll probably test a Cyrix 5x86 120 in there next to see how it behaves.

The one really weird thing about this board so far is that there are only 3.3v and 3.6v settings.

Well.. I accidentally found out how to get this board to drop to an 8.3Mhz fsb.

1. Set fsb to 16.66Mhz
2. Have the CPU multiplier jumper set to 2x or 2.5x
3. Boot system
4. Remove CPU multiplier jumper

I have only tested this with a Cyrix 5x86, which btw, the CPU multiplier 2x jumper setting works for on this board.

This halving of the fsb looks to work with the other fsb settings as well though I only tested on the 40Mhz fsb so far.

Edit:
The halving of the fsb trick doesn't work on the Evergreen 5x86 133 upgrade chip.
Guessing it will only work on CPUs that support a multiplier change via the motherboard settings.

So after a bunch of testing with 3 different Socket 3 PCI / ISA boards, this VIP board has so far come out the overall winner.

PCI speed is amazing, VLB speed is equal to other VLB boards running at 40Mhz fsb, and RAM speed is really good as well.... and it can take 1MB L2 cache.

The bad rap this board has gotten in the past is completely unwarranted from what I can tell... just have to install the 3rd fsb jumper to get the extra fsb settings.

PCI speed is over double the SIS based boards and the VIA based board I tried has nice, fast PCI, but the Voodoo 3 3000 will not work with it. Not sure if it is just an incompatibility or if it is a BIOS issue. I don't have any EEPROM chips that will work with this board so I am stuck waiting for some to arrive. It has a very early version BIOS from what I can tell so hopefully that will fix some things.

The other thing about these other boards is that the RAM speed plain sucks on the VIA board when at 40Mhz fsb.. like 1/2 the speed of when it is at 33Mhz.

The SIS boards are super finicky with the BIOS settings as far as RAM speed goes and give the fastest results when setting the "SETUP DEFAULTS"... Changing it to manual yields significantly lower results even when the timings are lowered to the best settings that are stable.... figure that out.

I do have a few more PCI/ISA boards to try out including a Mtech (Rise) R418W2 but it is currently not working at all due to damaged traces on the back of the board. Yay for scrap lots.

I found that the 486AP4 is very picky about PCI cards, especially graphics cards. A lot of cards that I tried on that board, even older ones, did not work properly. They mostly lock up in Win95 or when switching to a VESA mode in DOS.
S3 Virge and S3 Trio64V+ are such candidates. Interestingly, a Trio64 (without V+) works perfectly, as well as a Matrox Millenium.
A Hercules Dynamite VLB (ET4000W32) works, whereas a Dynamite *PCI* does not...very strange.
That board also only has one dedicated PCI master slot, the second one is tied to the VL slot and the third is slave only - not very useful for most PCI cards.
I also have a 486SP3G, but I cannot remember having these issues with it as well...

That's very good info. I just happen to have such a mobo in build with DX4-100 or Am5x86-133.
I will try your trick to get 40MHz FSB.

The one really weird thing about this board so far is that there are only 3.3v and 3.6v settings.

Which one you use with 5x86 and have you measured it? Yet I'm using 3.3V and a bit curious is it enough at 160.

Well, the 5x86 I was using was one of the Evergreen upgrade chips that will work with either 5v or 3.3v. I left it at 3.3v and it seemed to work just fine at 160Mhz from the little bit of testing I was doing.

Way back in the day when I had an AMD 5x86 in my main system, I was running that at 160Mhz and I am pretty sure I never changed the voltage from stock for that either.

Really just depends on the particular CPU whether or not it is stable at that speed on stock volts.

I was testing a different board with one of my Cyrix 5x86 120 CPUs and little did I know that the fsb jumpers are labeled incorrectly on the board and accidentally powered it on at the 50Mhz fsb setting.... at 3.45v no less.... so I have some small hope that that CPU may be super overclockable.

Do you happen to have manual for PVI-486AP4? Manual on the ASUS site is just stupid, missing a lot of pages e.g. no BIOS flash instructions. After a long search found the right flasher file from finnish site. Also had to remove BIOS chip sticker to see the chip type.

fool wrote on 2020-10-13, 15:25:

Do you happen to have manual for PVI-486AP4? Manual on the ASUS site is just stupid, missing a lot of pages e.g. no BIOS flash instructions. After a long search found the right flasher file from finnish site. Also had to remove BIOS chip sticker to see the chip type.

I don't have the original manual yet. I was just using the lame one form the ASUS site.

I used my USB programmer to update the BIOS.

I did just order a manual off of eBay though and will get it scanned and uploaded to archive.org and vogonsdrivers.com once it gets here.... probably won't be till next week sometime.

cyclone3d wrote on 2020-10-13, 06:22:

Well, the 5x86 I was using was one of the Evergreen upgrade chips that will work with either 5v or 3.3v. I left it at 3.3v and i […]
Show full quote

Well, the 5x86 I was using was one of the Evergreen upgrade chips that will work with either 5v or 3.3v. I left it at 3.3v and it seemed to work just fine at 160Mhz from the little bit of testing I was doing.

Way back in the day when I had an AMD 5x86 in my main system, I was running that at 160Mhz and I am pretty sure I never changed the voltage from stock for that either.

Really just depends on the particular CPU whether or not it is stable at that speed on stock volts.

I was testing a different board with one of my Cyrix 5x86 120 CPUs and little did I know that the fsb jumpers are labeled incorrectly on the board and accidentally powered it on at the 50Mhz fsb setting.... at 3.45v no less.... so I have some small hope that that CPU may be super overclockable.

What jumper settings did you use to get 50mhz FSB. I'm trying to get a 486dx 50 working at 50mhz fsb. Also you mentioned PCI bus speed was good, what did you use to benchmark it.

I've got the original manual for the board, bought it in december 1994 (really expensive board back then ,$240). I will do a complete high res scan and upload it to this thread. Might save you buying the book.

AvalonH wrote on 2020-10-13, 22:50:

cyclone3d wrote on 2020-10-13, 06:22:

Well, the 5x86 I was using was one of the Evergreen upgrade chips that will work with either 5v or 3.3v. I left it at 3.3v and i […]
Show full quote

Well, the 5x86 I was using was one of the Evergreen upgrade chips that will work with either 5v or 3.3v. I left it at 3.3v and it seemed to work just fine at 160Mhz from the little bit of testing I was doing.

Way back in the day when I had an AMD 5x86 in my main system, I was running that at 160Mhz and I am pretty sure I never changed the voltage from stock for that either.

Really just depends on the particular CPU whether or not it is stable at that speed on stock volts.

I was testing a different board with one of my Cyrix 5x86 120 CPUs and little did I know that the fsb jumpers are labeled incorrectly on the board and accidentally powered it on at the 50Mhz fsb setting.... at 3.45v no less.... so I have some small hope that that CPU may be super overclockable.

What jumper settings did you use to get 50mhz FSB. I'm trying to get a 486dx 50 working at 50mhz fsb. Also you mentioned PCI bus speed was good, what did you use to benchmark it.

I've got the original manual for the board, bought it in december 1994 (really expensive board back then ,$240). I will do a complete high res scan and upload it to this thread. Might save you buying the book.

That was a different board that had a 50Mhz setting. This board doesn't seem to have one even with the 3rd jumper installed.

I guess what I should do is hook up jumpers directly to the clock gen and see if I can find a 50Mhz setting. My guess is this is a custom version of the clock gen though as mentioned in the data sheet.

For the PCI speed, I was using a 3dfx Voodoo 3 3000. Sysmark reported around double the throughput of the other boards I have tested. This board was getting over 44MB/s while the other boards were getting around 22-26MB/s with the same card at the same 40Mhz fsb.

One of the other benchmarks in Phil's benchmark pack was also showing significantly higher throughput on this board. Same for the vidspeed benchmark.

Still limited to 90.9 fps in 3DBench though. Guessing that is the max that is going to be seen at a 40Mhz fsb as different CPUs gave the same result.

I also already ordered the manual. A good scan would save me some time though.

Did some research on those jumpers.

Clockgen MX8315PC
Juper positions 1-2 are HIGH and 2-3 LOW
JP24 and JP25 are not changing the clock speed, at least not from the clockgen.

JP26 is S0
JP27 is S1
S2 is tied to ground, I could not trace where it goes. Maybe it's directly to ground plane (?).

EDIT:
Looking the table, clockgen output frequency is divided by 2 at some point. Maximum FSB is then 40 MHz. Need to check where the output actually goes, is there some IC doing the dividing...
Pulling pin 13 (S2) up (internal pullup) would not help getting higher frequency.

With setup:
JP24: 2-3
JP25: 1-2
JP26: 1-2 -> S0=1
JP27: 2-3 -> S1=0
I got post showing 150. I did not have time to check out was real frequency 160MHz. Had to go sauna 😀

Sweet. That was going to be my next step.

JP24 didn't really seem to do anything other than add instability to most fsb settings.

So maybe we can get all the other datasheet listed settings. Will have to mess with that next week.

Sounds like we may need to add a jumper in between ground and S2.

From what I can tell, the 1-2 position on the others is the same as them not being installed. Didn't check with a meter yet though.

It seems that clockgen output goes straight to pin 108 of S82426EX *CLK2IN*. Dividing is done in the chip.
This confirms the suggestion of max FSB 40 MHz.

Maybe JP24 and JP25 is related to HOSTCLK and PCICLK as it changes between 25 / 33 MHz depending on the FSB setting.

What SYSCLK might be with unspecified CLK2IN of 80MHz?

Ah, ok.. so adding the jumper for S2 should then give us these additional fsb speeds
10
12.5
20
30

With these settings, that should mean we should be able to get 5, 6.75, 10 and 15Mhz fsb when playing with the Cyrix 5x86 and the multiplication jumpers and possibly with a 486 dx4 as well.

What about adding a clock doubler circuit such as this?
https://www.maximintegrated.com/en/design/tec … tes/3/3327.html

That way we could play around to see if we could get 60 and 66Mhz fsb working.

Finished scanning the manual, took awhile, 94 pages!! lot of excercise running up and down stairs to the scanner and trying not to damage the manual.
I have attached it to this post, front inside page of manual states it is manual revision 1.2, dated August 1994, and for mainboard Rev 1.6 and up.

One thing, do you guys know how much memory this board (Aries chipset) can cache. Can't find any info on this in the forum or google. I only have 32MB but the board supports up to 128MB. Can it cache the full 128MB??

When I am back home, I can check the cacheable amount... I'll put 512KB cache in it and then test with cachechk with 128MB installed unless somebody else knows for sure.