Yes they are compatible, but you need 9 pieces of UM61256FK-15 for double-banked, 256K cache. You can mix them with other 32Kx8 bit chips, such as, W24257AK.

feipoa wrote:

Yes they are compatible, but you need 9 pieces of UM61256FK-15 for double-banked, 256K cache. You can mix them with other 32Kx8 bit chips, such as, W24257AK.

Great! Found UM61256FK-15 [DIP-28W] locally. 160 here i come.

Where are these parts sold locally?

feipoa wrote:

Where are these parts sold locally?

Found one shop in Warsaw [PL]. They have only 4, so I suspect they are salvaged from old MB.

I just ran into few aditional problems:

1. The PS/2 out on the mobo has 5 pins, i only have 4 pins converter for back of the case. That is small problem, already found few 5pin versions online.
2. How do i connect the Turbo SW on this board? In manual: JP1 pins 23 & 24, problem is my turbo SW have 3 pins connector (pc case have 3 digit MHZ display), when I connected it wrong it started to smell, so quickly i just turned off PSU.
3. I noticed that the Geforce is veeeeeeery hot. My version is passive, think i will have to add fan.

The PS/2 header is 5 pins, but only 4 are used. Some MB's have 6 pins, but still only 4 are used.

Sounds like you connected the switch to the wrong pins if it started to smell. You can use either of 2 adjacent pins on the switch, depending on if you want turbo to be turbo, or low, and visa-versa.

feipoa wrote:

The PS/2 header is 5 pins, but only 4 are used. Some MB's have 6 pins, but still only 4 are used.

Sounds like you connected the switch to the wrong pins if it started to smell. You can use either of 2 adjacent pins on the switch, depending on if you want turbo to be turbo, or low, and visa-versa.

Hmm, the weird thing is that no matter how i connect the 3-pin turbo, the mhz display is not working 😒 according to manual next to turbo SW, JP1 pins 25 & 26 are +12V ground and just before Turbo SW is the pin22 with no connection. I think the 12V is the problem, everytime i see that kind of switch it's connected to 5V.

Few aditional obserwation:

Win95 won't install with 128Mb of edo ram [all 60ns] "SUWIN caused an Illegal instruction in module..." Acording to Microsoft (https://support.microsoft.com/en-us/kb/224836), this behavior can occur when any of the following conditions exist:
- There are incorrect basic input/output system (BIOS) settings for a built-in peripheral device on your motherboard (for example, cache settings, CPU timing, hard disks, and so on).
- Your computer has bad or mismatched random access memory (RAM) or cache (for example, if you are using EDO and non-EDO RAM, or you are using different RAM speeds).
- The motherboard in your computer is not working properly.

I'm using 2 sets: 2x32MB from one manufacturer and another 2x32 from different one. Even with 64MB the mentioned error sometimes occurs.

COM ports embedded in pc case are not working, switched them with other one i have, and it works. Still it's weird, never had a broken COM ports 😐

Try loosening the Ram timings in the Bios and see if that helps with Win95 install.

Some motherboards used non standard com port pinout, so standard cables may not work in that scenario.

Imperious wrote:

Some motherboards used non standard com port pinout, so standard cables may not work in that scenario.

Good to know, lucky that i had free com port that works with this mobo.

Regarding RAM, timings in bios are on the lowest values 😒 After removal of another stick of RAM, leaving only one 32MB the system [win95] was still very unstable. Constant explorer errors, register errors, blue screens. I decided to install Win 98 - so far no errors, works pretty stable, will see what happen after ram upgrade.

I've heard that this mobo works best with FPM RAM, found one guy who is selling ram with VG2617400DJ marks on chips, think it's FPM, someone can confirm this?

opiate wrote:

Good to know, lucky that i had free com port that works with this mobo. […]
Show full quote

Imperious wrote:

Some motherboards used non standard com port pinout, so standard cables may not work in that scenario.

Good to know, lucky that i had free com port that works with this mobo.

Regarding RAM, timings in bios are on the lowest values 😒 After removal of another stick of RAM, leaving only one 32MB the system [win95] was still very unstable. Constant explorer errors, register errors, blue screens. I decided to install Win 98 - so far no errors, works pretty stable, will see what happen after ram upgrade.

I've heard that this mobo works best with FPM RAM, found one guy who is selling ram with VG2617400DJ marks on chips, think it's FPM, someone can confirm this?

If the last number is 0 (like it is in your VG2617400DJ), then it's probably FPM.
And if you're still in doubt, you could use one of those older VIA Socket 7 Pentium boards that support single SIMMs (like the FIC PA-2005) or just look up the part number on a website named "Chipmunk" (dunno if this one still exists though).

Do not use EDO RAM on this motherboard. Did I not mention this already? If you are going to use EDO RAM, then set your cache timings to SLOW, like 3-2-2-2. You may also need to add some RAM wait states. I'm not sure if you can still use the slow refresh option either. I have not experimented much with EDO because it just doesn't work well on this MB. Get yourself a good set of FPM memory, preferably 32 MB and 64 MB sticks. Make sure they are 5 V by looking up the part number. A lot of eBay sellers just assume they are 5 V when they are, in fact, 3.3 V. This is especially true of 64 MB sticks.

There were two common COM pin-outs for socket 3 systems. I find that SiS boards tend to use one style, whereas the UMC boards use another style. There are other threads on this topic.

Are you aware that slowest RAM and cache timings are those with the largest numbers?

Do a search online for VG2617400DJ and find the datasheet. The datasheet will say if it is EDO or FPM. It will also say what voltage the chips run at.

EDIT: I do not think you should be connecting any pin of the 3-pin turbo switch to Vcc or GND. Usually the MHz display has a separate cable for power.

I have added revision 2 of the MB-8433UUD manual to this thread, The World's Fastest 486

It is the first downloadable link.

feipoa wrote:

I have added revision 2 of the MB-8433UUD manual to this thread, The World's Fastest 486

It is the first downloadable link.

The attachment Clickthis.png is no longer available

Great manual! Big time saver!

Turbo SW.

Just noticed that the mhz display have separate power cables and they go straight to PSU. Will have to check how many V the cable is supplying, ideally the same as molex so I could make quick 'mod'. My ATX PSU doesn't have this sort of connector (it's the same as other mb connectors for power sw, reset, leds, etc.).

I think the 3 pin turbo sw is build in a way so it could also be turbo LED. I will have to scrap this cable and solder new one with separate 2 pins for SW and 2 pins for turbo led.

I'm just wondering - does the mhz display take data from SW? how does it know what to display?

RAM.

Yeah you did mentioned the FPM memory - it's just that they are so hard to get 😒 and I already have 128MB in EDO. Nevermind - will buy FPM. Just one thing - manual claims that 128 MB is max, and max bank capacity is 32MB - so is 64MB per bank confirmed? 256MB in 486 mobo would be nice 😀

Does the "Fast Page Mode access" listed in Features of the VG2617400DJ confirms that this is in fact FPM?

Here is datasheet: https://www.digchip.com/datasheets/download_d … r=VG2617400DJ-6

If so, I have one seller here who is selling new (some corporate preventive supplies) FPM 32MB Simms for ~5USD. Think it's good price. Access time : 50/60 ns. Problem is with voltage, seller doesn't say anything aboult that, and acording to manual, there are two versions: 3.3 and 5V. That is problematic.

Stability.

Now this is interesting. All my test with windows 95 and later 98 were conducted on standard HDD. Windows 95 completly unstable, after playing around with 98, errors started to show up too. Decided to replace standard HDD with CF.

Surprise-surprise ALL problems disappeared! No more errors, no more blue screens. Windows 95 runs stable, played few games for hours - no problem.

Tested Quake without 3dfx: 11.8 fps, and with 3dfx: 16.0 fps. [S3 Virge 4MB + 3dfx voodoo1, Am5x86@133, 32MB EDO]. Decided to go with modified driver from here: http://www.falconfly.de/voodoo1.htm

So after few hours with just 32MB I decided to try 128MB. First try - 80% fine, just one blue-screen, changed few settings in bios, and to this point, no errors, no blue-screen. Picture of options attached.

Here are benchamarks after RAM upgrade:

Without 3dfx + newer Direct X: 9.4 fps - 😀 lower than before
With 3dfx + newer Direct X: 13.9 fps - again: lower than before

All test under Win95 USB edition, parameter used: -nosound -noipx -nojoy -nocdaudio +timedemo demo1

Trivia

Tested my standard HDD with 'long' scandisk - no errors. Weird.

For some reason I am unable to restart my machine. Every time i try to restart from Windows, PC freez - not really freez - i have flashing "_" in right corner, but nothing happens. Weird 2.0.

Most of your questions can be answered if you read the new Biostar manual, cover to cover.

You are using horrifically slow CMOS timings. Stop using EDO RAM. Do not go beyond the cacheable limit.

As for the turbo switch, all my 486 boards only use 2-pin turbo headers. The turbo switch is single-pole, double-throw, so you can use any of the two pins, depending on what you want the display to be: turbo and de-turbo. I have a 386 MB which uses the 3-pin turbo header. I beleive it sends some kind of signal that the computer has passed POST because before POST, it reads turbo LO, after POST it reads turbo HI. Of course, if I hit the turbo button, it reads LO or HI again.

Your 486 case's MHz display needs to have the speed display set manually by jumpers behind the display itself. You can set it for, e.g. 160 and 40 MHz to signify HI and LO. The button is connected to the display, so it knows when you pressed it. The cable header flips a switch on the MB which adds a bunch of wait states, thereby slowing the system down.

feipoa wrote:

Most of your questions can be answered if you read the new Biostar manual, cover to cover.

You are using horrifically slow CMOS timings. Stop using EDO RAM. Do not go beyond the cacheable limit.

FPM is on it's way, will check it with 256kb of double banked cache and cpu @160.

feipoa wrote:

As for the turbo switch, all my 486 boards only use 2-pin turbo headers. The turbo switch is single-pole, double-throw, so you can use any of the two pins, depending on what you want the display to be: turbo and de-turbo. I have a 386 MB which uses the 3-pin turbo header. I beleive it sends some kind of signal that the computer has passed POST because before POST, it reads turbo LO, after POST it reads turbo HI. Of course, if I hit the turbo button, it reads LO or HI again.

Hmm, since the default mode for this mobo is 'turbo on' i think i can skip this switch. Therefore Turbo LED is not important either. I connected the mhz screen to the turbo led pins for power - it works fine.

feipoa wrote:

Your 486 case's MHz display needs to have the speed display set manually by jumpers behind the display itself. You can set it for, e.g. 160 and 40 MHz to signify HI and LO. The button is connected to the display, so it knows when you pressed it. The cable header flips a switch on the MB which adds a bunch of wait states, thereby slowing the system down.

Then i'am screwed, there is no way in hell i can guess correct jumper settings for that screen. Searched online for manual - no results.

If you are going to use 256 KB cache, do not go beyond 64 MB of RAM, you will have a big performance drop in Windows. And if you are using 64 MB RAM with 256 KB cache, you must set the L2 cache to Write-through and the Alt bit to 8+0.

If you want to use more than 256 KB, 512 KB single-banked will not work with the fastest cache timings at 40 MHz. 512 KB double-banked might be stable, but then you will need to perform the cache mod.

I recall seeing at least one post which has the pin-outs for the MHz display.

Not sure if I'd connect the MHz power connector to the Turbo SW control. Might be better to use pins 25/26 (Vcc/GND). Ref. pg 18.

feipoa wrote:

If you are going to use 256 KB cache, do not go beyond 64 MB of RAM, you will have a big performance drop in Windows. And if you are using 64 MB RAM with 256 KB cache, you must set the L2 cache to Write-through and the Alt bit to 8+0.

If you want to use more than 256 KB, 512 KB single-banked will not work with the fastest cache timings at 40 MHz. 512 KB double-banked might be stable, but then you will need to perform the cache mod.

OK - so my cache chip arrived.

Plugged them in, jumpers set [acording to manual: JP5,JP6,JP7], mobo boots, bios set to 2-1-1-1 and L2 W-T. It' just works.

Then... i thought that this could be right time to set cpu @ 160. Jumpers set: JP15 open and JP16 closed. PC boots, I've used ISA graphics this time. Then weird thing happened - after minor bios changes, computer acted just like before: no screen.

So I changed CPU speed to 133, and PC boots normally.

I use 64MB EDO instead of FPM, (they are ordered, will be at my house in few days).

I think it possible that I am one of the few people with am5x86 than won't work @160 🙁

BTW: It is writted in manual that PCI @40 can be problematic, but does it apply if i'm not using PCI cards?

feipoa wrote:

I recall seeing at least one post which has the pin-outs for the MHz display.

Not sure if I'd connect the MHz power connector to the Turbo SW control. Might be better to use pins 25/26 (Vcc/GND). Ref. pg 18.

I use pins 10+11, they are for Turbo LED, it works. Managed to set "110" and "170" 😁

Problem with mhz display is that there are as many jumpers settings as PC case manufacturers back in the days. Found one webpage that tries to sums it up, but it's just one drop in the sea.

"after minor bios changes,"
What minor BIOS changes?

EDO + 40 MHz is particularly problematic. If I were in your shoes, I'd set the system aside until you have some FPM memory. I believe your issues are related to the EDO.

PCI at 40 MHz is mostly problematic when you have PCI graphic cards installed.

You should be able to achieve 40/160 on your MHz display. Spend an hours to study the jumper table and play around. There is a logic to it. I also find it frustrating, but the results are achievable.

feipoa wrote:

"after minor bios changes,"
What minor BIOS changes?

"Onboard Parallel Mode" and "Slow Refresh (1/4 Freq)" was Disabled - changed it to Enabled.

feipoa wrote:

EDO + 40 MHz is particularly problematic. If I were in your shoes, I'd set the system aside until you have some FPM memory. I believe your issues are related to the EDO.

Ok - I will wait for FPM. BTW: is there a software that can tell me what kind of memory I am using [edo/fpm/other]?

For 72-pin SIMMs, I do not believe any software can determine the FPM/EDO RAM t ype, but I may be wrong. It doesn't take that long to find the RAM type by doing an IC part number search.

Have you read the manual now? Do you think the cache mod is something you could do?

feipoa wrote:

For 72-pin SIMMs, I do not believe any software can determine the FPM/EDO RAM t ype, but I may be wrong. It doesn't take that long to find the RAM type by doing an IC part number search.

Have you read the manual now? Do you think the cache mod is something you could do?

I'm torn. At one hand I think I would be able to perform this mod (from a soldering skill point of view), on the other side - there is a risk I could break mobo with very little performance to gain. I think i'll wait until I get another Biostar mobo. Problem is - I was searching for this one for months and don't see any other on the horizon 😒

Anyway...

My FPM arrived. Plugged in 32MB, jumper set for 160. PC boots, and that is it. Another boot - no screen. Used ISA graphics, everything else plugged off [sound card, hdd, cd rom, etc].

I think my am5x86 just won't go past 133. Sad, sad, sad.