I might have mixed some information (you also modded the voltage regulator), but next to the BIOS there are frequency selectors. On the board I have there were installed only two, out of four. Third one was hard soldered as short. I know from another board that there are 3 selectors, as described on the board it self. Does the 4th one do anything?

You are refering to the FSB selection jumpers that control which pin out of the PLL chip is used. I don't recall what the extra 4th solder pad jumper is for. I left mine unsoldered.

PhilsComputerLab wrote:

Anonymous Coward wrote:

What is the model of your desoldering tool?

ZD-985

Recently if bought one too. I was curious if this desolderstation would do my expectation.

I had received this unit last wednesday.. But am not totally happy with it.. Whats good and whats bad on this unit:

My ZD-915 clone came from poland (named `extreme`)

The good.

-My version came with metal casing, and the quality of that its good.
- Filters are descent quality. (beter then the aoyue ones)
- The tips are also good (and having an 0.8mm one (the aoyue dont)
- I also like the cleaning wires that comes with this unit
-Also like the windowed / cleared solder collector chamber (so you can look how full it actually is)
- The heating element is also good.

The bad

- Dont like the front of it.. It looks like a bad quality AT case style cheap front think with some unevenness / scares in it.. But if you dont beware of it it doesnt matter. The finishing of this front could be a little bit more higher quality.

- Really dont like the fan (its only noisy) and i really dont know why they add this to this unit.

- The pistol is very cheap. Once you open it it still a mess.. These things are wrong on the pistol: Very cheap plastics, plastics dont close enough on both parts (there is still a gap). The heater element sitts slanting in the pistol casing, so the tube to the tip isnt aligned properly with the solder collecting chamber, also this reducing the vacuum section power..

- The vacuum suction isnt very great.. (not much power on my unit)

Iam doubt if buying a new gun would solve my problem on the low vacuum thing.. It sucks ok when keeping a finger on the hole collar it self.. But when connecting the gun on the collar the fun is over..

The unit cost me 91 euro.. And buying a new gun i would pay 37 euros for it.. And still havent a good working unit here.
I was already thinking of buying a professional weller unit.. Only that would costs me about 1150 euros..

Would i recommend to buy any duratool or same clone or an aoyue desolderstation.. No i would recommend to go for the more expensive brands out there.. On the long run they are much better quality and cheaper.. I spend about 150 euros on the aoyue one and 91 euros on the duratool clone and still have nothing.. Yes two units that dont would do my expectation.. Only collection dust..

Hello, I am new to the forum.

I have a Biostar 8433-UUD-A Ver.2 with the same problem of Dallas chip dead. Does anyone know if the problem is solved by connecting an external battery in the JP42? This option would be easier to replace the chip.

I have another question: How 60Mhz bus is set on this board? I know there are people who have achieved and I would like to reach 180Mhz.

My 486 has a pretty extreme hardware. These are the components:

- AMD Am5x86-P75 ADW 133 @ 160MHz (9634GPD)
- 256KB L2 cache 15ns
- 256MB RAM EDO 50ns 4x64MB
- Graphics Creative 3D Blaster CT6850, S3 Savage-4 PCI 32MB
- MPEG-2 decoder RealMagic Hollywood DVD Playback Card PCI
- Sound card Creative Sound Blaster AWE 64 Gold ISA
- PCI Card 5+1 USB 2.0 Adaptec AUA-5100B chip Nec
- Network Card 10Mbps ISA
- Philips CD-RW 24x10x40
- LG 16x DVD
- Hard drive Seagate IDE 4'3GB (Windows Millenium installed)
- Power Supply 235W

I have found that doing hot overclock the performance is much better than starting to 160MHz directly. I put the turbo switch to JP15 for starting at 133Mhz and cause the overclock after BIOS boot.

I find this extremely slow motherboard with the transfer rate of the hard disk. One of my plans was to install a SATA controller, but today I received an ATI Radeon 9200 128MB PCI for this 486 but does not function on this motherboard, so I do not think I can run a SATA controller on it.

I have also tried to install Windows 2000 but do not have the IDE controller driver for this operating system, and it is impossible to install W2K without such driver. Does anyone know where to find it?

I await your answers. Greetings!

Welcome to the forum D.Tape.

I am not sure about using JP42 with an ext. batt. That may have been for an early prototype design which contained a RTC module without an integrated battery. Or it may work fine. I have never tried it.

Most people end up desoldering the RTC module and installing a DIP socket for a new RTC. Or cutting open the plastic on the existing RTC and attaching a new battery to the dead battery's lead. However, even this latter approach is best done with the RTC out of the motherboard.

Concerning a 60 or 60 MHz FSB, please reference my custom manual for this board found in the World's Fastest 486 link (see my signature). While I have extensive plans to overhaul this manual, the information you seek is already contained within. Namely, JP15=open, JP16=closed, JP17=open.

Do you possess an Am5x86-P75 which is known to function reliably at 180 MHz? I have at least 6 of these chips and none of them are reliable at 180 MHz. I have one chip which first give the impression that it might be usable in DOS, but extended tests, particularly a boot into Windows, confirmed it was not sufficiently stable.

If you are only using 256 KB cache in Write-back mode, then only 32 MB of your RAM is cacheable, meaning using more than 32 MB of RAM in Windows will results in a 20% performance loss. If you are going to stick with 256 KB of cache, I suggest you install my modified BIOS, which allows you to run the cache in write-through mode, and limit your RAM to 64 MB so that all of it is cached.

What hardware do you run with the USB 2.0 card? This motherboard has a working PS/2 mouse port, so there would be limited need for a USB mouse on the USB 2.0 card. Maybe you are using a flash drive on the USB 2.0 card? If you are looking to increase the overall system performance, particularly in Windows, I suggest removing the USB 2.0 card and installing a DMA-capable hard drive host controller, such as SCSI, UltraDMA, or SATA.

What cache timings are you using?

I have determined that EDO RAM on this motherboard does not allow for as fast cache/RAM timings as FPM if you want a stable system.

Could you provide your results for hot overclock vs. starting at 160 directly? If your hot overclock performance is better than starting at 160 MHz from the onset, I suspect you have your cache and or RAM timings are set to auto. Setting these CMOS settings up manually should yield the same end results.

I noticed you were trying a Radeon 9200 on this motherboard. If you are interested in the latest possible graphic cards which work on UMC 8881-based motherboards, I suggest you view the list contained in this thread, Modern graphics on a 486

My favourite graphics card combination on this motherboard is a Matrox G200 mixed with a Voodoo2. This allows you to play glide and D3D games that normally only early Pentiums and newer could play.

The Promise Ultra100 TX2 works on this motherboard, however the soft reset won't work when using this host controller. You must hard reset. On the other hand, an Adaptec 2940U2W SCSI controller works very well on this MB. A Promise SATA150 TX2 plus sorta works, but after numerous reboots I had some odd issues in Windows 9x, which were not apparent using the Ultra100 TX2, nor within Win NT4 and the SATA150.

I was able to install Windows 2000 successfully with the integrated IDE controller and the W2K-built-in IDE controller. I'm not sure why you are having an issue. However, the built-in IDE controller will be really slow, particularly with Windows 2000. I do not recommend it.

EDIT1: If you do end up replacing the RTC with a new one, I think DS12887A is the one you want. If I remember right, the A suffix is preferred because it allows you to set a jumper which resets the data in the RTC. The DS12887+ chip removes the pins which allow for the jumper reset.

EDIT2: I forgot to point out that just having the USB card enabled in Windows will lower your benchmark score on a 486 system. Use CPUMark99 to confirm this.

feipoa wrote:

Do you possess an Am5x86-P75 which is known to function reliably at 180 MHz? I have at least 6 of these chips and none of them are reliable at 180 MHz. I have one chip which first give the impression that it might be usable in DOS, but extended tests, particularly a boot into Windows, confirmed it was not sufficiently stable.

I'm not sure, I just have two of these chips and didn't know how to set the motherboard to 60Mhz FSB. I'll test them when I can.

feipoa wrote:

If you are only using 256 KB cache in Write-back mode, then only 32 MB of your RAM is cacheable, meaning using more than 32 MB of RAM in Windows will results in a 20% performance loss. If you are going to stick with 256 KB of cache, I suggest you install my modified BIOS, which allows you to run the cache in write-through mode, and limit your RAM to 64 MB so that all of it is cached.

Yes, I know, but with only 32-64MB of RAM is necessary to have the swap memory available in Windows and the hard disc is much more slow than the RAM chips, so I prefer a lot of RAM without the cache benefits and let the swap file OFF in Windows cause the general performance is better (not in benchmarks, it's true).

feipoa wrote:

What hardware do you run with the USB 2.0 card? This motherboard has a working PS/2 mouse port, so there would be limited need for a USB mouse on the USB 2.0 card. Maybe you are using a flash drive on the USB 2.0 card? If you are looking to increase the overall system performance, particularly in Windows, I suggest removing the USB 2.0 card and installing a DMA-capable hard drive host controller, such as SCSI, UltraDMA, or SATA.

Now it's 4 years ago that I don't use this PC because I was working out of my city (and in this time the Dallas chip has dead), but the USB card had the function of pendrive ports and wireless USB card for internet connection and sometimes for connect a multicard reader (SD/MMC/CF/etc).

Anyway, the USB PCI card has a Nec chip which somebody said to me that it doesn't decrease the microprocessor performance because it works by itself (not like VIA chips). I don't know if it is really true but I never noticed a worst performance with this USB card installed, it works fine.

feipoa wrote:

What cache timings are you using?

I have determined that EDO RAM on this motherboard does not allow for as fast cache/RAM timings as FPM if you want a stable system.

I don't remember, but the RAM SIMMs are 50ns and I suppose that the timings would be better than with FPM SIMMs. I can't verify at this moment because the PC doesn't start with the Dallas chip dead.

feipoa wrote:

Could you provide your results for hot overclock vs. starting at 160 directly? If your hot overclock performance is better than starting at 160 MHz from the onset, I suspect you have your cache and or RAM timings are set to auto. Setting these CMOS settings up manually should yield the same end results.

I can't provide results for now, you know the problem with the Dallas chip. I know that in someplace I saved a capture with the benchmarks of CrystalMark running at 133 and 160Mhz, but not with the difference between starting at 160Mhz or making hot overclock. I'll look for those capture images and will upload them here.

feipoa wrote:

I noticed you were trying a Radeon 9200 on this motherboard. If you are interested in the latest possible graphic cards which work on UMC 8881-based motherboards, I suggest you view the list contained in this thread, Modern graphics on a 486

My favourite graphics card combination on this motherboard is a Matrox G200 mixed with a Voodoo2. This allows you to play glide and D3D games that normally only early Pentiums and newer could play.

The S3 Savage 32MB that it has installed lets a very good performance in every aplication or game that this PC can execute, but I like to play DVDs on it and the MPEG decoder is taking a PCI slot that would be amazing for installing a SATA controller. I bought the Radeon 9200 expecting this card could decode MPEG-II by itself without needing the help of RealMagic Holliwood decoder PCI card, but it doesn't work on the 486 so I'll try to install it on a Pentium-Pro.

feipoa wrote:

I was able to install Windows 2000 successfully with the integrated IDE controller and the W2K-built-in IDE controller. I'm not sure why you are having an issue. However, the built-in IDE controller will be really slow, particularly with Windows 2000. I do not recommend it.

I tried everything that was in my hands for installing Windows 2000 and it was impossible. Tried with every version (from first to SP4), tried installing on other PC, tried with 64MB to 256MB RAM, tried disabling cache... and there was no way.

I told every adventure with this 486 project in this link from year 2006 to 2011. I had a lot of problems, the first mainboard (Octek Hippo 12 VIP) died when I had everything prepared for assembling and it took 2 years for find the actual Biostar MB-8433UUD, then was very dificult to find 4x64MB RAM and many others stories (sorry, it's spanish): https://www.hard2mano.com/topic/14240-el-486- … ente-del-mundo/

(Sorry for my poor english, I try to do it better that I can)

Hey Guys,

I recently got my hands on a Biostar MB-8433UUD-A Ver2. Upon arrival it didnt work which turned out to be a faulty bios chip. I flased a new bios onto an SST chip and its now up and running. I have also desoldered the odin chip and fit a socket and new dallas timeclock with external battery mod.

I have tried various bios files; the 2014, the lastest official ones and more. The problem im having is i just cant seem to get the performance it looks like Phil has achieved. I have a Gigabyte GA486AM/S which is very similar and achieves better results (1487 realtiks in doom vs the 8433UUDs 1620 realtiks.

I have set the board up with an AMD 5x86 133, 256Kb double banked cache and have tried various ram EDO and FP. The problem doesn't appear to be the cpu or ram as the scores for CPU, Ram and cache are basically identical to the Gigabyte

Any suggestions would be appreciated 😀 i have attached bios settings, i have tried tweaking various options to no avail

I was recently trying to squeeze the best possible fps from this board but clocking the amd 586 to 160. I was stuck at 16.6 fps and got nearly 1 whole fps extra by using the modified bios from feipoa in this thread The World's Fastest 486

His bios has the unhidden option for L2 wb cache which I see you do not have in your bios.

Since you have double banked cache it would be only a matter of flashing this bios and using L2 WB which should provide a better performance.

Apart from that your timings look quick. Anyway im sure feipoa will pop up sooner or later and he knows this board inside out.

For comparison quake timedemo I was getting about 14 fps on this board with 5x86 amd 133 then clocked at 160 with low timings I would get 16.6 finally with WB L2 and double banked cache I got 17.3 or 17.4

Is the graphics card the same between the Gigabyte and the Biostar? Did you check the part number on your memory chips to ensure that it is indeed FPM? Only use FPM on this board. Judging bo your BIOS screenshots, it doesn't appear that you are using the 2014 BIOS? In the 2014 BIOS, I also set the default optimal BIOS settings. For example, I noticed that you have Early Cache Write mode disabled, while I keep it enabled. I also have Burst Copy-Back enabled and IRQ assigned for VGA. This may not be related to your issue though.

I don't see the ISA Bus clock option in your screenshot. Might want to unhide this feature with the 2014 BIOS. Same is true for the missing L2 cache feature. When you unhide this feature, make sure that when you set L2:write-back that 7+1 setting for Alt Bit in Tag SRAM, and when using L2:write-through, that 8+0 is used.

Are you using the DOOM demo version on both motherboards? Is there perhaps a green frame around the screen on the Gigabyte board, which will yield faster results.

The Gigabyte and Biostar boards use the same chipset and have similar BIOSes and there shouldn't be much difference in the results.

I have not checked your jumpers, but please compare them with the information I provided in my custom manual for this board.

I did try the 2014 bios and tried tweaking the setting you guys are talking about thanks. I spent alot of time tweaking settings to no avail. My board simply can not crack 1600 realticks

I used exactly the same tests for both boards using phils benchmark pack. I used the same hard drive so there cant be any discrepancies from that. Also yes i used the same ARK video card

I actually had a look on Phils ultimate VGA database project and he has his identical board listed with a 5x86 133 With timings tuned and he has listed a 1614 realticks which is what im seeing.

Interestingly enough there is also a listing for a Gigabyte GA486AM/S also which returns 1413 realticks which is also near to what my board scores

I suspect the screenshot he posted on the first page with 1350 realticks has to be an amd 5x86 clocked at 160Mhz or perhaps some capacitors are worn on my board altering some timings or something?

I cant find any other benchmarks of the board at 133Mhz all i can find are results at 160Mhz.

A bit unfortunate. Such a shame the Gigabyte doesnt have PS/2 as it seems to be the faster board. Oh well might keep my eye out for another board with PS2

I don't see the frequencies listed on that chart.

I have both of these boards and I might investigate this.

You can use the serial to PS/2 protocol converters noted in this thread on the Gigabyte board, but it only works well if you aren't using a KVM. Or if you are using a KVM, you need to use the 19200 baud driver, but that driver only works in DOS.

Since the Gigabyte board doesn't have a DIP-40 KVM, it isn't easy to add custom PS/2 support.

There are frequencies on the chart i just couldn't fit them in the screenshot i will attach a larger screenshot now that i'm on a PC. It also says further to right tested by MKau1wurf1977 being Phil, Bios "Modified UUD1212 BIOS from Feipoa", "2-1-1-1 Cache WS, DRAM read: 0WS, DRAM write: 0WS"

Interestingly i flashed the bios from the Gigabyte and tried running it. It seemed to run fine but returned the same results as your 2014 bios. So i guess kind of rules out the bios settings

If you do get around to testing yours out id be interested in knowing the results 😀

Biostar with Gigabyte bios

Gigabyte bios bios settings

Can you run the Quake timedemo on both systems and see if there is the same percent of difference between the boards?

have you tried clocking yours to 160mhz and running doom? if your cpu is able to do 160 and most amd133 do it could give some more information, for example if your cpu is getting correct results at 160 but lagging at 133 then its more isolated, lagging at 133 and 160, can rule out the cpu, or perhaps try another 5x86 if you have one

I know your cpu works on the other board at full speed, but maybye just maybye there is something about the particular cpu the biostar doesn't like

s.mouse - Can you also try using a different graphics card? Perhaps one of the cards I have tested below?

Tonight, I squeezed in the time to run the MB-8433UUD (original version, pre-2.0) with various graphics cards. Am5x86 @ 133 MHz, 32 MB FPM (single stick), double-banked 256K cache, Adaptec 2940U2W adapter with a 2 GB Seagate HDD (50-pin narrow), and MS-DOS 6.22. I'm using Doom timedemo 3 and Quake timedemo1. Ensure that you don't have a mouse driver loaded in DOS and that you only have HIMEM loaded.

Graphics card - Quake - DOOM
Matrox G200 - 14.2 - 1489
VirgeGX - 14.2 - 1507
Riva128 - 14.2 - 1513
VirgeDX - DNT - 1519
Vision968 - 14.1 - 1544

DNT = did not test

You mention that your target is 1487 and the Biostar with a G200 is 1489. You might want to try other graphic card combinations. I do not own the ARK, so I cannot test this combination. It is possible that the Biostar board doesn't favour your ARK card, while the Gigabyte does?

Did you run Speedsys to ensure that your L2 cache is enabled and working? If I, for example, disable the L2 cache, I get a score of only 1593 realtics, whereby L2:enabled is 1489 realtics.

EDIT: For further interest, I ran the Am5x86-160 on this motherboard w/G200. The results were such that DOOM = 1244 and Quake = 17.1 fps.

I pulled out my Gigabyte GA486AM board and ran some tests using the Matrox G200.

Am5x86-133
Cache - Quake - DOOM
256K - 14.2 fps - 1459
1024K - 15.2 fps - 1405

Am5x86-160
Cache - Quake - DOOM
256K - DNT - DNT
1024K - 18.2 fps - 1179

So the Gigabyte board does appear to be marginally faster than the Biostar board, that is, 1459 vs 1489 realtics 256K at 133 MHz, or 1 FPS. Similarly, for 1024K at 160 MHz, 1179 vs. 1244 realtics, or about 3 FPS. These differences are much smaller than what you witnessed. But why are there these differences? Perhaps the CLKGEN on one board is putting out 33.1 MHz, while the other is putting out 34.3 MHz? Perhaps the presence of the PS/2 mouse slows it down some?

I wanted to see how the Gigabyte board compared to my favourite SiS 496-based board, the MSI MS-4144, but unfortunately, the Matrox G200 does not work on this board. Using the VirgeDX, 1024K/133MHz yielded 1537 in DOOM and 14.9 fps in Quake. With 1024K/160MHz, we get 1291 in DOOM and 17.9 fps in Quake.

I never knew increasing the cache will make such a difference, I recall getting exactly the same scores in quake with L2 off and on in the bios (256k) maybye I remembr it wrong

s.mouse, you've gone silent. Did you try one of the alternate graphic cards I listed, preferably the Matrox G200?

My bad i was working away other the weekend. I have set the board up again to try your suggestions.

I installed a fresh copy of dos onto a spare hard drive copied Phils benchmark pack 1.4 then flashed the 2014 bios and tried various video cards.

Unfortunately i dont have a G200 so cant provide results for one. I do have a matrox Mga millennium which i tried but gave the same result around 1600 realticks. I tried some other cards - S3 Trio64V2/Dx, Tseng E6000 and Cirrus but all returned around 1600.

I find it strange the same board has such different results

try quake see what u get