Considering that the inability to overclock occured at the same time as using the incorrect RAM, I'd say that your memory controller has some questionably burned out transistors due to overcurrent (inside the chip). Depending on how long you had the wrong voltage RAM in there, the memory controller (Northbridge) may always act flakey. Surprisingly, the M919 I put the wrong RAM into 10 years ago works a little better now than it used to. You may need to reduce the cache wait states to get the board stable again.

This same principle can be applied to CPU's as well, which is why I don't like to run them at a voltage much higher than they need. For this reason, I use a variable voltage regulator on the CPU's to find the lowest possible voltage the CPU will run at stably while not overheating (aka, the sweet spot). It is quite a tedious process, but for these overclocked Cyrix CPUs, it is somewhere between 3.65 and 3.95 volts. I've got a thermocouple probe at the base of my Cyrix right now for characterisation.

Since you are using a Shuttle HOT-433 v4, your Northbridge memory controller is likely a UMC 8881F. These are not produced anymore, but you can probably desolder them from other motherboards. To properly remove SMD chipsets, you need a rework station, perhaps like this one,
http://www.youtube.com/watch?v=VySTIRoPtt4

Or you can use flux and a wick like this guy,
http://www.youtube.com/watch?v=0EUAEtri3h0

Before going to all this hassle though, have you thoroughly tested the motherboard and determined it is a failed memory controller? What happens if you put in an AMD X5 at 133 MHz and run it with 2-1-2 and 0ws/0ws (using 3.45 V)? Have you reset the BIOS with the reset jumper?

I think the HOT's prefer paired RAM, are you using FPM RAM in pairs? Two sticks of 60ns, low density, 8 MB RAM modules would be good for testing this.

I cleared BIOS again to see if that would make a difference, and it did. It made a huge difference. Now it won't POST even at normal speed and cache settings.

I tested other CPUs in there, including my Cyrix 100GP (that I was able to run at 120MHz before without trouble) and I got the same results.

Prior to clearing BIOS, I was able to overclock to a stable 60x2.

Yes the Northbridge is the 8881F, and I don't have any others (except for the one in the apparently dead 8433UUC) so I'm screwed at the moment... time to dig through my junk again for
another board to work with for now I guess.

Lol I guess I should learn to leave well enough alone, but I won't. I have too much curiosity.

At least I can say that I accomplished a few tasks with this board so far: 150MHz running Windows 98SE, broke the 80 point threshold on System Speed Test,
proved 60 and 66 MHz front side bus speeds, and even caught a glimpse of an IBM 5x86 100 at 180MHz (though that might have just been another straw on the proverbial camel's back)
all without going over 4v.

Heh, well, I never did have much long-term luck with HOT-433 boards. I think they are gypsey cursed. Do you have any SiS PCI 486 boards?

Lol I guess I should learn to leave well enough alone, but I won't. I have too much curiosity.

I too have itchy hands. There is no cure.

At least I can say that I accomplished a few tasks with this board so far: 150MHz running Windows 98SE, broke the 80 point threshold on System Speed Test, proved 60 and 66 MHz front side bus speeds, and even caught a glimpse of an IBM 5x86 100 at 180MHz (though that might have just been another straw on the proverbial camel's back) all without going over 4v.

Indeed, quite an accomplishment. I would say that you stiill have a working unit of the more valuable item -- A Cyrix 5x86 @150 MHz. The fact that you could run Win98SE on it indicates to me that your IBM 5x86C-100HF is better than the 2 I have. You could probably run it at 133 MHz without any of the issues I've been battling. If you are ever having a Zen moment and feel like getting rid of retro hardware, let me know; I'd be happy to adopt that Cyrix.

If you are ever having a Zen moment and feel like getting rid of retro hardware, let me know; I'd be happy to adopt that Cyrix.

A very generous offer, my friend.

Do you have any SiS PCI 486 boards?

Why, yes, I do. One of my favorites (mostly because it's the only 486 board that seems to be functioning fully) would be my Soyo 4SAW2. It's a great board for testing VLB cards in it too if need be.
Although right now it's setup with an AMD 5x86 133 and some other parts that I consider very nice, so I'll have to bring myself to tear it apart before I can test it with the Cyrix/IBM CPUs.

Well, that, and I want to wait until I have at least one more IBM 5x86 before pushing the limits again... the only other one I have is in a sealed Gainbery 5x86 120 (at least I'm pretty sure it's an IBM 100HF in there, but I still can't bring myself to rip it open yet).
I have such a tiny collection of old hardware left, it's pathetic.

So you do have a spare Gainbery.... no point in keeping the box all sealed up; you're on a role now! And best check for DOA. ;)

From my previous readings, this is the same 4SAW2 used on a German forum for his AMD X5-200.

SiS 4SAW2
512 KB L2 cache max
256 MB RAM max
SiS chipset
4 PCI slots
PS/2

The specs look good. I've been keeping an eye out for this motherboard, but one never seems to show up. On paper, this motherboard is better than my Biostar board in that it has 4 PCI slots. The Biostar only has 3. The rest of the specs are the same except for the chipset. The SiS chipset should work with Voodoo 3/4 cards, where the UMC chipset will not. The best graphics card for a UMC is a Matrox G200, which is already CPU-limited anyway, so anything better won't make a differance.

Again, if you're having a Zen moment and are looking for a suitable benefactor, look no further.

I hear that some SiS boards have the 60/66 MHz FSB setting; this may be one of them. Where did you find this motherboard? Are there more?

If you don't need the PS/2 port, there's a similiar SiS board which supports 1024 KB cache, its the Soyo SY-045A2. To date, I have not found a working 486 board with 3+ PCI slots, which supports 1024 KB cache, and has a working PS/2 port.

I think that before you jump straight into a 150 MHz Cyrix, confirm that 2x66 MHz works well with your Cyrix. You'll be shockingly pleased by the speed of the 66 MHz front-side bus. I have a whole cased setup of a 2x66 MHz system now and have recently ordered some double-wide slot fans to keep it extra cool.

EDIT: If you are going continue with Cyrix 5x86 testing, you might be interested in reading my PDF document on the Cyrix register enhancements, found here:
Cyrix 5x86 Register Enhancements Revealed

Well, that's a fountain of knowledge that I didn't already know. Thank you for the resources.

I bought the motherboard a few years back (when I first decided to get back into the roots of my PC experience) from a website that specialized in obsolete technology. There were a ton of them back then, but now there's none to be found anywhere. Honestly, I didn't really know what I was getting, and was pleasently surprised when I discovered just how good the board was overall. I have a second one that I might try to resurrect someday (northbridge chip has some damaged legs, among other issues).
Other than that, those are the only 2 I've personally seen.

I opened the case that stored the 4saw2 and discovered that I have quite a bit of corrosion to deal with before I start testing it. I had forgotten that there was an old barrel battery soldered onto the motherboard, and never replaced it with a CR2032.

More details later I guess, I have a lot of work to do.

Soyo 4saw2 had no problems recognizing the Cyrix 100GP that I threw in there, but I'm not sure how to check for 60/66 MHz FSB on this thing. I'm a bit skeptical of it's abilities to push that, but I want to know for sure.

Also with the layout, it's a bit tricky figuring out how to change the dividors and stuff... I really want to get my Shuttle HOT-433 up and running again 🙁 .

I'll update again when I get something interesting going on it.

I thought it was dead... at the very least mortally wounded. I was about to begin piling the dirt on it's coffin when I detected a faint glimmer
of life in the old Shuttle HOT-433 (Rev. 4). As I leaned closer, its angry head erupted from the soft earth, spraying dirt everywhere as it
roared fire and defiance at death itself.

I gazed in awe as it unfurled it's ragged wings and, slowly at first, but then gaining speed, raced across the ground flapping it's wings and
finally gaining flight. It dissapeared over the horizion for a few moments and I thought I might have lost it again, but then in and instant, it
was triumphantly soaring among the clouds at 133MHz.

Epilogue: I replaced the failed ide controller cards with an Adaptec SCSI controller (AHA2940-U2W) and put in a Barracuda 2.1GB temporarily to
patch it into use. I have a Promise TX2 SATA controller card in my possession to try soon. The major significance being that it now works at
66x2 at only 3.3v stable so far. I was never able to do that before the northbridge fried.

Here is a benchmark showing it running at slowest timings and no enhancements... getting more benchmarks now.

Uploaded with ImageShack.us

Lastly I would like to appologize for the overdone introduction to this post if it bothered you.

Hahaha, retro PC-revival excitement at its best.

It seems you now have the two finest PCI hard drive controller interface cards a 486 could ask for.

Your sputtering 486 acts a lot like my M919 after injuring the northbridge. I'm working on an UMC order now with the Chinese for replacement QFP parts. For a lot of them, it seems they see an e-mail in English and automatically assume they can rip me off. I'm sorting through 25 vendors now to find the best price. If you are so interested, I could add an extra 1997 area UMB northbridge/southbridge for you to the order. It would be ~$10. I haven't yet found a cheap way to get them on the motherboards yet though.

I'd recommend running your CPU at 3.60V instead of 3.3V. The HOT-433 v1-3 at least has the 3.6 V setting.

Maybe you can run it on the slowest memory settings, install Windows and ensure everything is running fine. Then start speeding up the memory timings.

Hahaha, retro PC-revival excitement at its best.

Perhaps I should write that novel.

It seems you now have the two finest PCI hard drive controller interface cards a 486 could ask for.

I may have an appallingly small vintage parts collection, but at least I have a couple parts to be proud of. 😀

I'm sorting through 25 vendors now to find the best price. If you are so interested, I could add an extra 1997 area UMB northbridge/southbridge for you to the order.

I'm interested. Message me with the details, if you please.

I'd recommend running your CPU at 3.60V instead of 3.3V.

Odd thing happened. It lost stability (wouldn't boot 8 out of 10 tries) and now the 3.3v is a little flakey.

Heh, sounds like you need some IC rework done on that HOT-433.

After e-mailing 100+ vendors, I got a list of respondants down to 25. I then filtered the database of costs from cheapest to most expensive. Unfortunately, not all vendors want to deal with low quantities, and not all had the cheapest prices.

After this, I then e-mailed the 25 vendors again to name my target price. A lot of them were like "no way, price too low." I now have a short-list of only a few remaining vendors who can supply my parts at the lower end of the cost margin.

In particular, I am going to order these items:

IS61C1024-10N (10 ns DIP-32 cache, 128x8). Good for 512 KB cache, single banked, or 1024 KB cache double-banked. Most of this size cache is 15 ns, so these, if they do exist, they would be a good find.

W24257AK-10 (10 ns DIP-28 cache, 64K x 8). To serve as the TAG for the above noted cache.

--------------for Biostar MB8433, M919, HOT-433------------
UM8881F - UMC chipset northbridge, QFP-208, 1997 datecode
UM8886BF - UMC chipset southbridge, QFP-208, 1997 datecode
UM8663BF - UMC chipset super I/O, QFP-100, 1997 datecode
UM8667 - UMC serial port controller, SSOP-48, 1997 datecode
UM9515-01 - UMC Phase-lock loop chip to set motherboard FSB, DIP-14, 1996 datecode
--------------for Biostar MB8433, M919, HOT-433------------

UM8670F - UMC super I/O, QFP-100, 97/98 datecode. Used on some newer M919 motherboards.

M5818PB1 - ALI keyboard controller, DIP-24, For M919 motherboards, 1997+

PH29EE010-150-3CF - The flashable SST BIOS used on Biostar MB8433 and Shuttle HOT-433 motherboards, DIP-32, 1996+

UM8002 - UMC, SSOP-52, Not sure what this is for, but the HOT-433 has 5 of them onboard. 1996+

W29EE011-15 - Winbond flashable BIOS. A good replacement for the M919 and other motherboards. 1997+

I'm planning on getting this order out very soon. It has been a week in the making. If anyone wants these chips, please PM me ASAP what exactly you want. I'll try to postpone the order for another 24-Hrs.

If/when I actually receive this shipment, I'll need to figure out how to reliably and cheaply get the QFP-208 chipsets onto the motherboard.

feipoa wrote:

Heh, sounds like you need some IC rework done on that HOT-433. […]
Show full quote

Heh, sounds like you need some IC rework done on that HOT-433.

After e-mailing 100+ vendors, I got a list of respondants down to 25. I then filtered the database of costs from cheapest to most expensive. Unfortunately, not all vendors want to deal with low quantities, and not all had the cheapest prices.

After this, I then e-mailed the 25 vendors again to name my target price. A lot of them were like "no way, price too low." I now have a short-list of only a few remaining vendors who can supply my parts at the lower end of the cost margin.

In particular, I am going to order these items:

IS61C1024-10N (10 ns DIP-32 cache, 128x8). Good for 512 KB cache, single banked, or 1024 KB cache double-banked. Most of this size cache is 15 ns, so these, if they do exist, they would be a good find.

W24257AK-10 (10 ns DIP-28 cache, 64K x 😎. To serve as the TAG for the above noted cache.

--------------for Biostar MB8433, M919, HOT-433------------
UM8881F - UMC chipset northbridge, QFP-208, 1997 datecode
UM8886BF - UMC chipset southbridge, QFP-208, 1997 datecode
UM8663BF - UMC chipset super I/O, QFP-100, 1997 datecode
UM8667 - UMC serial port controller, SSOP-48, 1997 datecode
UM9515-01 - UMC Phase-lock loop chip to set motherboard FSB, DIP-14, 1996 datecode
--------------for Biostar MB8433, M919, HOT-433------------

UM8670F - UMC super I/O, QFP-100, 97/98 datecode. Used on some newer M919 motherboards.

M5818PB1 - ALI keyboard controller, DIP-24, For M919 motherboards, 1997+

PH29EE010-150-3CF - The flashable SST BIOS used on Biostar MB8433 and Shuttle HOT-433 motherboards, DIP-32, 1996+

UM8002 - UMC, SSOP-52, Not sure what this is for, but the HOT-433 has 5 of them onboard. 1996+

W29EE011-15 - Winbond flashable BIOS. A good replacement for the M919 and other motherboards. 1997+

I'm planning on getting this order out very soon. It has been a week in the making. If anyone wants these chips, please PM me ASAP what exactly you want. I'll try to postpone the order for another 24-Hrs.

If/when I actually receive this shipment, I'll need to figure out how to reliably and cheaply get the QFP-208 chipsets onto the motherboard.

This is the same problem I sometimes have. Those who have the parts for a good price want you to buy in bulk. Everyone who will sell smaller quantities, want too much money.

Not for these Chinese IC vendors; they are a dime a dozen. I find that a 10 count is usually sufficient, with the odd ball requesting 20.

When I was refering to low quantities before, I was refering to less than 10 pieces, in which case the vendor pool fell by 75%.

I promised that I'd post some results when I tested more hardware combinations. Well, here's one of 'em.

Uploaded with ImageShack.us

CPU: IBM 5x86 100 @ 133
HDD: CF card on TX2 controller (first time I've used this controller and it's actually pretty good)
RAM: still the 4MB x 4 sticks.

I noticed that speedsys stopped posting the 66x2 as 133MHz on this machine not too long after using the wrong RAM. Just something I thought might help in some way with identifying what exactly fried on the motherboard.

An interesting test would be how an Ultra2-LVD PCI adaptec controller w/Ultra320 HDD compares up against the TX2 w/200X CF card in real world applications.

Update: I tested the first controller card in the Soyo 4saw2 motherboard (can't believe I didn't think of this before) and it worked... now I have 2 SATA/IDE controller cards that work in 486 motherboards. I'm thinking that the blown ICs on the HOT-433 is what held it back from working on the Shuttle.

I know the HOT-433 has PCI SCSI bus mastering issues, at least in Windows 2000, so it may also have PCI ATA bus mastering issues.