So that last installed failed. After the reboot during the install, I got an "ODBC Installer DLL is not installed on this system" and after the next reboot I got registry corruption errors.

I deleted the files off the drive, copied over fresh scripted install files from a contemporary computer, did an error check on the SD, turned off UDMA in the BIOS, and then restarted the scripted install.

The install stopped here with the mouse still working but the keyboard num lock unresponsive:

douglar wrote on 2025-08-23, 15:12:

So that last installed failed. After the reboot during the install, I got an "ODBC Installer DLL is not installed on this syste […]
Show full quote

So that last installed failed. After the reboot during the install, I got an "ODBC Installer DLL is not installed on this system" and after the next reboot I got registry corruption errors.

I deleted the files off the drive, copied over fresh scripted install files from a contemporary computer, did an error check on the SD, turned off UDMA in the BIOS, and then restarted the scripted install.

The install stopped here with the mouse still working but the keyboard num lock unresponsive:

The attachment Photo Aug 23 2025, 11 29 02 AM.jpg is no longer available

For any motherboard that has symptoms like this that seem to be memory related that has onboard level 2 cache - try slowing it down or entirely disabling it in bios and see if the problem goes away, or if it makes no difference

bertrammatrix wrote on 2025-08-23, 19:03:

For any motherboard that has symptoms like this that seem to be memory related that has onboard level 2 cache - try slowing it down or entirely disabling it in bios and see if the problem goes away, or if it makes no difference

Disabling the external cache fix things, thanks!

douglar wrote on 2025-08-24, 14:49:

bertrammatrix wrote on 2025-08-23, 19:03:

For any motherboard that has symptoms like this that seem to be memory related that has onboard level 2 cache - try slowing it down or entirely disabling it in bios and see if the problem goes away, or if it makes no difference

Disabling the external cache fix things, thanks!

Ok, well that's not really a fix but good to hear it at least made a difference. Performance will likely be pathetic without it. You could try cleaning the board around the cache, or if that doesn't help you could try cooling it with a little heatsink on it, or a fan.

Sometimes these chips were already borderline adequate for the upper FSB speeds SS7 boards could do when they were new, 30 years later the situation is obviously only worse. It may still work fine at 66, 75 or 83 but not be able to do 100 reliably

Edit - since you are using this motherboard with a k6-3 that has it's own L2 cache and effectively uses the motherboard cache as L3 - I wonder if the penalty for disabling the motherboard cache may be fairly minimal

bertrammatrix wrote on 2025-08-24, 15:21:

Edit - since you are using this motherboard with a k6-3 that has it's own L2 cache and effectively uses the motherboard cache as L3 - I wonder if the penalty for disabling the motherboard cache may be fairly minimal

There's a hit, but it's < 10%. It isn't a deal breaker. I was running the 03/08/1999 BIOS. It is supposed to know about the K6-III, but I see 01/03/2000 rev A is newer. I'll give that a shot.

douglar wrote on 2025-08-24, 20:42:

I was running the 03/08/1999 BIOS. It is supposed to know about the K6-III, but I see 01/03/2000 rev A is newer. I'll give that a shot.

Here's how it went:

1. I installed the new bios which returned settings to the default values, including external cache = enabled
2. the system rebooted and installed a couple new drivers
3. before I could restart, the registry got corrupted beyond repair
4. I'm back to reinstalling Windows 98se

douglar wrote on 2025-08-24, 23:50:

Here's how it went: […]
Show full quote

douglar wrote on 2025-08-24, 20:42:

I was running the 03/08/1999 BIOS. It is supposed to know about the K6-III, but I see 01/03/2000 rev A is newer. I'll give that a shot.

Here's how it went:

1. I installed the new bios which returned settings to the default values, including external cache = enabled
2. the system rebooted and installed a couple new drivers
3. before I could restart, the registry got corrupted beyond repair
4. I'm back to reinstalling Windows 98se

The attachment download.jpg is no longer available

Haha, yup, typical results 😀 that's the best when it looks really promising before repeat fubar 😀 well at least now you know

So after reinstalling windows I was still having trouble. I wouldn’t get a windows desktop, just a black screen in text mode. I noticed that after the crashes, the CF was unresponsive until a power cycle. The Bios upgrade had reenabled udma for the device. Turning off udma and forcing pio4 seems to have fixed that issue.

OK, I'm remembering why these super 7 boards are so addictive. Because you think you can fix them if you just do a little more work !!

Here's some speedsys benchmarks:

1) No cache / loosest memory timings. SDRAM = SLOW / CAS 3.

2) Cache Disabled - Memory throughput increases from 152MB/s to 168MB/s. SDRAM = Turbo / CAS 2.

3) Cache Enabled -- No change to the CPU score. Everything must fit in the on chip cache. SDRAM = Turbo / CAS 2.

4) Using Sandisk Ultra CF instead of Innodisk iCF 4000. SDRAM = Fast / CAS 3.
* Curiously, the CPU benchmark takes a hit with the Sandisk. That wasn't a one off. It was completely repeatable and it was the same regardless of the main memory timing.
* Sandisk didn't perform as well in Dos, but it wasn't really noticeable

5) The Sandisk doesn't have the UDMA issues.
* Operations < 8kb are slower, Operations > 8kb are faster
* The general feel of windows 98 is better. Fewer pauses. Fewer crashes.

Mother-fudgecicles! The board isn't seeing IDE devices anymore. This is definitely a toxic relationship.

One of my first computer builds was a AMD k6-2 ( 300mhz ) from what I remembers.
I was running Win95/98
It did not run very good because I was new at building computers and was just learning.
I think i may have had the jumpers set wrong and I was using the wrong type of ram too.
So it would crash every now and then.
I was going to get around to fixing it but I really wanted a BX440 motherboard with Pentium ll-400
So I put away the K6 and Built the BX440 instead.
That is when I decided to only use Intel CPUs and I never went back.

Until, I got into retro computers I wanted to rebuild that AMD K6 computer so I did.
It works okay with all the right settings and ram but I still think the Pentium computer is more stable and reliable .

My install process
:
I install Win98se
Then SP3 core updates only
Reboot
Then install Directx 9 from SP3 package
Reboot
And then install voodoo and Sound card software and drivers

Works pretty good.
Stable and Reliable.

Wish I knew this back in 1998
Would’ve have saved me allot of frustration

Intel486dx33 wrote on 2025-08-25, 20:14:

One of my first computer builds was a AMD k6-2 ( 300mhz ) from what I remembers. I was running Win95/98 It did not run very good […]
Show full quote

One of my first computer builds was a AMD k6-2 ( 300mhz ) from what I remembers.
I was running Win95/98
It did not run very good because I was new at building computers and was just learning.
I think i may have had the jumpers set wrong and I was using the wrong type of ram too.
So it would crash everybody now and then.

That's how I felt except I'd been building computers for 10 years at that point!

I ended up keeping an AMD and and Intel computer until the Athlon 64 came out and then I want AMD. When the Core 2 E8400 came out, I went with 2 intel PC's. When the Ryzen came out I ended up going back to AMD.

I had a loose connection in my IDE socket. A quick reflow seems to have fixed the IDE issue. Tempted to put one of those Promise ATA 100 cards to work in this computer.

douglar wrote on 2025-08-25, 22:06:

That's how I felt except I'd been building computers for 10 years at that point! […]
Show full quote

Intel486dx33 wrote on 2025-08-25, 20:14:

One of my first computer builds was a AMD k6-2 ( 300mhz ) from what I remembers. I was running Win95/98 It did not run very good […]
Show full quote

One of my first computer builds was a AMD k6-2 ( 300mhz ) from what I remembers.
I was running Win95/98
It did not run very good because I was new at building computers and was just learning.
I think i may have had the jumpers set wrong and I was using the wrong type of ram too.
So it would crash everybody now and then.

That's how I felt except I'd been building computers for 10 years at that point!

I ended up keeping an AMD and and Intel computer until the Athlon 64 came out and then I want AMD. When the Core 2 E8400 came out, I went with 2 intel PC's. When the Ryzen came out I ended up going back to AMD.

I had a loose connection in my IDE socket. A quick reflow seems to have fixed the IDE issue. Tempted to put one of those Promise ATA 100 cards to work in this computer.

When I've used Promise IDE cards with MVP3 motherboards they have always worked, but at some point in installing Windows 98 drivers Windows would get stuck when rebooting and I would have to move the card to another PCI slot. Once all of the drivers were installed this would no longer happen and I could leave it in the same slot from then on. Just an FYI.

Repo Man11 wrote on 2025-08-25, 23:23:

When I've used Promise IDE cards with MVP3 motherboards they have always worked, but at some point in installing Windows 98 drivers Windows would get stuck when rebooting and I would have to move the card to another PCI slot. Once all of the drivers were installed this would no longer happen and I could leave it in the same slot from then on. Just an FYI.

Thanks for the tip.

K6-III + Super 7 is like Columbo. There's always just one more thing--

Welcome to the MVP3 nightmare!!

One lesson I learned the hard way, MVP3 IDE controller+SD card adapter or CF adapter = trouble

The solution is to use a very short IDE cable or even better a 80 wires IDE cable

Hope it helps

I got the Promise ATA100 TX2 card working and I was starting to make good progress.
https://theretroweb.com/expansioncards/s/prom … se-ultra100-tx2

I tested Voodoo 3 2000 PCI & AGP, then Permedia 2 AGP & 2V PCI. I'm benchmarking with Doom & Quake 1 for DOS and 3d mark 1999 & 2000 for DX7. It was going pretty well.

I'd also like to do Quake 2 & Quake 3 for Open GL but that's been a bit of a struggle for me. Anyone have a walk through for how to set up those for benchmarking? Do I need to bet full editions of the games or is there a publicly available edition for benchmarking?

Then I was ready to start Nvidia when things curdled.

I have 2 TNT cards. Neither of these would even display a picture when in my K6-III board--
https://theretroweb.com/expansioncards/s/nvidia-p009
https://theretroweb.com/expansioncards/s/nvid … -p0002-0000-b01

Then I have a pair of Mx4000 cards that do display a picture. I put this one in the system: https://theretroweb.com/expansioncards/s/msi- … 000-t64-ver-200
I was getting ready to install the driver when the thought occurred to me that maybe I should see how the Innodisk CF does with the Promise ATA100 TX2. That would be cool, yeah?

First I pulled the Sandisk Extreme 8gb 60MB/s CF and put in the Innodisk CF. The Innodisk CF shut down after starting Wind98se, requiring a power cycle to return to life, like before.

Then I put the Sandisk Extreme in IDE port 1 and the Innodisk CF in port 2 on the card. The system booted and the innodisk was visible as D:. Making progress! But I was running in 640x480 which is no good for ATTO screenshots, so I went to install the Nvidia driver. It gets 80% of the way to decompressing the Forceware driver and locks up.

Oh, it's the innodisk somehow, yeah? Power off, remove the innodisk, restart the system. It gets 80% of the way to decompressing the Forceware driver and locks up. Bah!

So I take the Sandisk Extreme to my contemporary computer, and it doesn't read. That's strange. I remove and reinsert. It comes up fine. Aren't these old devices whacky? How unpredictable and capricious! I'll get through this. So I manually extract the forceware driver using winrar and it all goes fine. I'm all set to try installing the driver.

I take it back to the K6-III and it starts scandisk. OK. Unexpected, but you do you, Win98. Do the needful. Whatever you want. Then there is a glitch in the matrix when it was up to 90% completed and then skipped back to 70%. I get that sinking feeling in my stomach when I see that, but it does finish scandisk without errors. But things don't boot. Then Promise firmware stops detecting the CF. I try moving the Promise card to a different slot. The Sandisk CF is detected again! But it doesn't boot. Then it isn't detected any more. And it won't read on my contemporary computer any more either. I left the old PC's and went up stairs to watch the Netflix documentary about "the poop cruise", because it seemed appropriate.

I'd been backing things up regularly because I figured the mess of drivers I was rolling through was going to foul my win98se install, so I'm not too put out. And I was able to grab the last couple screen shots off the CF today because of course it works this morning. But do I trust it? Nahh! I just want to finish this project. I still have the MX4000, Rage 128, and Radeon 7000 to get through.

So one CF adapter has a pin that got pushed thru. The other developed an odd sticky smear on the back. I'm going to stop using this brand of adapter for a bit.

The innodisk is still fussy as can be and earned another "timeout!" Probably needs fed or a diaper change.

Got things working with the Sandisk again in a different brand of adapter. Well, "working" is a relative term. I mean it's booting Win98se ok.

These MX4000 cards don't get along with 20th century versions of 3d bench very well. I probably need to go back earlier than Forceware 71.

Hi douglar,

I've made the experience with Forceware 71.84 and 3DMark 2000 and 2001 that the benches either won't start at all or beeing very unstable.
That was with a GeForce4 Ti4200 as well as an FX5200 on a more modern 775 Via platform.
With Forceware 53.04 and 45.23 all went fine.
Also, 3DMark99 was fine with all three driver versions.

Regards.

ChrisK wrote on Yesterday, 06:33:

Hi douglar, […]
Show full quote

Hi douglar,

I've made the experience with Forceware 71.84 and 3DMark 2000 and 2001 that the benches either won't start at all or beeing very unstable.
That was with a GeForce4 Ti4200 as well as an FX5200 on a more modern 775 Via platform.
With Forceware 53.04 and 45.23 all went fine.
Also, 3DMark99 was fine with all three driver versions.

Regards.

Thanks for the info. I was able to get things to work with Forceware 45.23 and forcing the driver as an MX440. Turns out that my AGP card only has a 32bit memory bus. That's never a fun discovery and kind of invalidates my testing. Even though the colors are bright, I'd describe the video quality on both MX4000 cards as "a little fuzzy"
https://theretroweb.com/expansioncards/s/msi- … 000-t64-ver-200
https://theretroweb.com/expansioncards/s/pny- … 00-64mb-ddr-pci

I wasn't planning on testing fx 5200's but maybe that's something to do. I'm pretty sure I have a PCI fx 5200 card somewhere. I forgot that they were older than the MX4000, even though I'm pretty sure the MX4000 is just an Mx440 with a new PCI ID.

And where is my Diamond STEALTH S60 PCI? It was "new-in-box", so I put it some place safe so that I wouldn't accidentally damage it after testing it. It seems I put it some place too safe.