myne wrote on 2026-09-16, 12:30:

myne wrote on 2026-09-15, 04:18:

https://www.youtube.com/watch?v=nk8uEFpYbBw

Short 2 pins.

Did you see this?

This mode will "convert" EDO SIMMs to FPM mode but not allow it to work in EDO mode in FPM-only motherboards
So no performance boost

myne wrote on 2026-09-16, 12:30:

myne wrote on 2026-09-15, 04:18:

https://www.youtube.com/watch?v=nk8uEFpYbBw

Short 2 pins.

Did you see this?

Sorry, yes.
I don't wanna take that route, no hardware moding, no bus crystal change and so on.
I can change voltage regulator if I damage it, but that is it.

Off the shelf components. Next steps are:
- try a different BIOS in the next few days.
- try getting some EDO modules that work and refine timings.

Babasha wrote on 2026-09-16, 12:48:

myne wrote on 2026-09-16, 12:30:

myne wrote on 2026-09-15, 04:18:

https://www.youtube.com/watch?v=nk8uEFpYbBw

Short 2 pins.

Did you see this?

So no performance boost

It's faster ram. 60ns vs 70.
Like running ddr5 4000 at ddr4 3200. Sure, it's only ddr4, but its got a higher ceiling by default.

Single bank cache is going to be the Achilles heel speed wise of this setup. Without bank interleaving it's going to be a real struggle to get something workable at 512k - those higher densities of sram usually leave zero margin at fast FSB speeds. You may very well be better off with L2 disabled if this enables you to get rid of any memory wait states.

For me the ultimate test after successful quake runs is a windows 98 installation where unpacking compressed content early on usually confirms go/no go of the current configuration

myne wrote on 2026-09-16, 16:07:

It's faster ram. 60ns vs 70.

These FPM modules are 60ns ones. 2x16MB

JAKra85 wrote on 2026-09-16, 08:26:

Quake is very memory and cache sensitive. I am running FPM, with a fast EDO I could probably get better score. The problem is t […]
Show full quote

amadeus777999 wrote on 2026-09-16, 08:10:

Great to see some AM5x86-200 based content!

Maybe try a Matrox MilleniumII - I got 21.3- in Quake and 30.7 frames in PC Player Benchmark(but with an UMC 888x based board).

Physical Cache configuration is an extremely hairy subject. In some cases you have to keep the order of the chips in mind... permutations aplenty. Also having a huge array of ICs with matching production dates is important... sometimes 15ns chips can outperform 12ns ones. I had a 586 SI5PI-AIO where only 1 set of 20ns sram chips enabled the fastest settings - to my surprise my heavily tested sets of 15- and 12ns chips would not allow proper function.

Quake is very memory and cache sensitive.
I am running FPM, with a fast EDO I could probably get better score.
The problem is that this board does not eat EDOs. 😁
One EDO module (out of a few) POSTs, but fails to boot to OS.

My 512K cache chips are quite mixed regarding matching production dates.
The 256K are all the same, but produces lower score (because it's 256K).

I often had Quake stable but Doom crashing in relation to cache issues.
Interesting - the multitude of possible sram failure / success profiles seems like a case for the X-Files...

amadeus777999 wrote on 2026-09-16, 16:49:

...failure / success profiles seems like a case for the X-Files...

This is exactly how I feel working with this setup! 😁
I have no idea why moving the VGA from one PCI to another would suddenly make 512K cache work.
Reported that IDE1 is not working, not POSTing when I connected something: System dead.
I just figured out that the optical drive (NEC) caused this, connecting an LG drive and it works perfectly.
Never seen anything like this before and I have no good explanation.

bertrammatrix wrote on 2026-09-16, 16:27:

For me the ultimate test after successful quake runs is a windows 98 installation where unpacking compressed content early on usually confirms go/no go of the current configuration

Windows 98 installation is in progress...

Truth to be told, I had to relax memory timings.
The first time I tried I got an error seconds after switching to: "Copying Windows 98 files to your computer" just as you predicted.

JAKra85 wrote on 2026-09-16, 18:43:

The first time I tried I got an error seconds after switching to: "Copying Windows 98 files to your computer"

I guess installing @ 200 MHz was OK, was it?

I wonder if SIV32L utility will do a Menu->File->Save Local OK, on systems similar to this sometimes the benchmark sections trigger issues, please will you try it and post the save file?

Do you plan to install any other OSs? I suspect Windows NT V4.00 SP6a will work quite well, how much memory does it now have? I think it's possible to run W2K with 64MB, but feel 128MB is needed for it to run at a half sensible speed. There are also i486 systems running Windows XP!

i assume this board clocks PCI frequency 1:1 with FSB so with your PCI running at 50MHz you are pushing a lot of the other parts of the system beyond their limits.

I am currently in something of a similar situation - I have a socket 7 motherboard that should not have 100MHz FSB - but it does. this allows me to run a K6-3+ CPU at 600MHz. BUT! as a result I am running PCI at 50MHz because there is no 1:3 divider on intel chipset boards. I had problems with the onboard IDE controller. To solve the problem I swapped to a promise Ultra100 TX2 PCI IDE card which is rated to run at 66MHz PCI frequency. My GPU is a voodoo3 PCI which is supposedly identical to the AGP version, hence it can also run 50MHz PCI completely fine.

If you still cannot get windows to install, and you want to try other ways to find stability, then I would say the IDE card is your next step. They are not expensive. I paid £15 for mine a few weeks ago.

red-ray wrote on 2026-09-17, 11:27:

I guess installing @ 200 MHz was OK, was it?

-Yes, I now know it's stable at 200MHz. Only relaxed memory timings and installation was successful.

red-ray wrote on 2026-09-17, 11:27:

I wonder if my SIV32L utility will do a Menu->File->Save Local OK, on systems similar to this sometimes the benchmark sections trigger issues, please will you try it and post the save file?

- I will try this out in the coming hours and reply.

red-ray wrote on 2026-09-17, 11:27:

Do you plan to install any other OSs?

-If I find 128MB RAM that run fine I want to install Windows XP on this one. Proper OS to "The Monser!" configuration. 😀

FullYes wrote on 2026-09-17, 14:29:

i assume this board clocks PCI frequency 1:1 with FSB so with your PCI running at 50MHz you are pushing a lot of the other parts of the system beyond their limits.

Correct, S3 Virge card had visible corruption. Anything older (ATi Mach32, 64) could not run with PCI@50MHz. 3DRage IIC and 3DRage II Pro PCI cards run fine with PCI@50.
Then I have two Radeon 7000 PCI cards, identical. One is not POSTing. The other one works almost perfectly. One down side is that with "soft" restart and reset (Ctrl+Alt+Del or restart from Start menu), I get no image.
I think the PC is working I get a POST beep and starts loading the OS. - No picture.
I must do a reset button reset to get picture.

FullYes wrote on 2026-09-17, 14:29:

If you still cannot get windows to install, and you want to try other ways to find stability, then I would say the IDE card is your next step. They are not expensive. I paid £15 for mine a few weeks ago.

Win98SE is running "fine".
I say "fine" because I fu*ked up. 😁 During installation I went custom installation and un-selected Internet Explorer...
I was like: I don't need that!
Result: Whenever I open My Computer or Document or anything Windows Explorer... I get an error:
WebView and missing Internet Explorer.
Windows 98 is using WebView by default and since app is crashing I can't go to settings and stop WebView.
Probably could do with registry or something.

However, now that I see Win98 working I will try to adjust memory timings and do a new installation.

What is the current memory? I guess 40 MB = 2 x 16 MB + 2 x 4MB.

JAKra85 wrote on 2026-09-17, 15:00:

If I find 128MB RAM that run fine I want to install Windows XP on this one. Proper OS to "The Monser!" configuration. 😀

Good luck with the search. I found the SIV save file from an i486 running XP and it only had 64 MB! SIV took 32 seconds to start.

I just noticed it's RTM and suspect with SPs more memory will be needed.

JAKra85 wrote on 2026-09-17, 15:22:

I say "fine" because I fu*ked up. 😁 During installation I went custom installation and un-selected Internet Explorer...

I suspect if you install en_ie60_sp1.exe it could fix this. In general it wise to install this as it installs COMCTL32 V5.50.4916.400 which many programs benefit from.

red-ray wrote on 2026-09-17, 11:27:

I wonder if my SIV32L utility will do a Menu->File->Save Local OK, on systems similar to this sometimes the benchmark sections trigger issues, please will you try it and post the save file?

I will get you the file soon. 😀
Can copy easily to the monster, but not from the monster... Monster->Floppy->WorkHorse PC(486)->USB drive->Main PC.

I am facing multiple issues right now.

1)
I am(was?) getting random page fault errors. I though it's VGA driver related, but it was not. Same issue with 3DRage II.
Now I believe the Internet Explorer error was also caused by this.
2)
Changed the cache back to the Winbond ones: 256K. Seems good now. No more page faults, so far. I could run SIV32L.
(next time I run SIV32L, remind me not to do it directly on the CD... it was reading, skipping so much and so loud I though the optical drive died.)
3)
After installing the Radeon 7000 driver for some reason it sets an out of range frequency for the monitor.

EDIT:
It reports the CPU as 3.45V, I am quite sure it's 3.6V.
Also, the memory is 2x16MBs currently.

JAKra85 wrote on 2026-09-17, 17:29:

Can copy easily to the monster, but not from the monster.

Thank you, in general SIV did OK, the only issue is the drive temperature is missing, what drive is it and did a SIV popup say you need to reboot?
If it's not there the next time please do SIV32L -DBGSDM -EXIT=20 > SIV_DBGOUT.log and post the .log file.

I had half assumed you would have it connected to your network, it's easy enough to setup Windows 10 to make this possible. Can't you use a USB Flash Drive? Just copy the storage driver files from a Windows Me system.

JAKra85 wrote on 2026-09-17, 17:29:

It reports the CPU as 3.45V, I am quite sure it's 3.6V.
Also, the memory is 2x16MBs currently.

SIV reports the default of 3.45V, if it's @ 200 MHz do you feel SIV should report 3.60V?

From the screen I can't tell why SIV decided 4 x 8MB, I need to check the save file. What get's reported with 2 x 16 + 2 x 4 MB?

JAKra85 wrote on 2026-09-17, 17:29:

remind me not to do it directly on the CD... it was reading, skipping so much and so loud I though the optical drive died.

Consider yourself reminded. I rarely run SIV from a CD, these days use a USB Flash drive. I suspect it was loading the xxxdevs.txt files.

I'm putting down The Monster! for a while.
Can't stabilize WIn98SE, getting page fault errors and BSOD left and right.
(Tried: cache on/off/removed, different VGAs, different RAMs, slowest RAM timings, different IDE controller.)
Waiting for a different BIOS, however I have a feeling that will not solve the problem.
Anything in DOS is fully stable, can't crash the system.

red-ray wrote on 2026-09-17, 17:45:

I wonder if my SIV32L utility will do a Menu->File->Save Local OK, on systems similar to this sometimes the benchmark sections trigger issues, please will you try it and post the save file?

Can't complete. Getting page fault errors at different steps.

red-ray wrote on 2026-09-17, 17:45:

what drive is it and did a SIV popup say you need to reboot?

-First time I ran it: yes. Not afterwards.
-Running on a CF card with IDE2CF adapter.

red-ray wrote on 2026-09-17, 17:45:

SIV reports the default of 3.45V, if it's @ 200 MHz do you feel SIV should report 3.60V?

That would not be good/exact either.
In my opinion when the speed is above 160MHz is should report the Voltage as unknown. It could be anything from 3.6V up to 5V.

JAKra85 wrote on Yesterday, 10:28:

Can't complete. Getting page fault errors at different steps. […]
Show full quote

Can't complete. Getting page fault errors at different steps.

red-ray wrote on 2026-09-17, 17:45:

SIV reports the default of 3.45V, if it's @ 200 MHz do you feel SIV should report 3.60V?

That would not be good/exact either.
In my opinion when the speed is above 160MHz is should report the Voltage as unknown. It could be anything from 3.6V up to 5V.

Thank you, it's a shame the system is unstable when running W98 🙁

How far did the save get? If it's always the same place do Menu->File->Save->Save Selected and de-select any problem sections. Either way please can I have the partial file?

For Beta-11 I have changed SIV to remove the voltage when the speed is above ~160 MHz

Given you got the popup SIV must have done the setup for SMART reporting, I need the .log from SIV32L -DBGSDM -EXIT=20 > SIV_DBGOUT.log to know more.

red-ray wrote on Yesterday, 10:48:

How far did the save get?

The farthest was: [mem-heap].

red-ray wrote on Yesterday, 10:48:

Either way please can I have the partial file?

Sure! Are they cumulative or each run generates a blank new file? The past few runs only went a couple of steps before erroring out.

red-ray wrote on Yesterday, 10:48:

I need the .log from SIV32L -DBGSDM -EXIT=20 > SIV_DBGOUT.log to know more.

Will check and get.

JAKra85 wrote on Yesterday, 12:12:

The farthest was: [mem-heap]. […]
Show full quote

red-ray wrote on Yesterday, 10:48:

How far did the save get?

The farthest was: [mem-heap].

red-ray wrote on Yesterday, 10:48:

Either way please can I have the partial file?

Sure! Are they cumulative or each run generates a blank new file? The past few runs only went a couple of steps before erroring out.

Thank you, I am unclear as to if the issue was with SIV or general system stability as I get the impression other saves had an issue earlier. Maybe saving less would work, try Menu->File->Save->Save Selected, press [ Minimal ] and finally [ Start ].

I think [mem-map] was when it faulted, but maybe it was [mem-heap] and the screen had not been updated. Please do Menu->File->Save->Save Selected, uncheck [mem-map] + [mem-heap] then press [ Start ].

A new file is generated every time a save is done.

Looking at [mem-map] + [mem-heap] they should not even be saved on Windows 9x systems so Beta-11 won't as neither report anyting useful on W9x systems.

I have just released SIV V5.88 Beta-11 which also removes 3.45 volts for > ~160 MHz CPUs. Use this and you don't need to do Save Selected.

"The Monster!" died! (technically it's in coma. 😁 )

I did not understand this instability in Windows that was getting worse and worse. I had a hunch...
For testing purposes I decided to move the CPU to an ABit AB-PB4 v1.5 mainboard that also has 50MHz BUS and also PCI bus halving just to confirm that it is not the PCI@50 creating a cascade issue on the previous mainboard.
The good part was that it was POSTing on this new board.
The bad part was that on this board it needed 4V to POST. 3.6V was not enough. - Here, at this point I knew, the CPU is done, the monster is dead.
The instability was exactly the same, on a different mainboard, cache, RAM.
It was unstable even at 160MHz...

It's working on 133MHz, but requires 3.6V to do so.

What I don't understand is:
-The default Voltage for this CPU is 3.45V. Used it at 3.6V and it still degraded quite fast.
-Temperature wise was running very cool. Even if I removed the fan, the huge cooler only got warm: ~40C.

Oh well, it was a fun project! 😁
(will not give up, need to find a few Am5x86 CPUs)

red-ray wrote on Yesterday, 13:44:

A new file is generated every time a save is done.

I am really sorry I did not copy the first try that almost went to the end yesterday. The last few tries only covered 2-3steps before erroring out.
Then same at 160. All I have is normal 133MHz that could complete.
By the time I understood what is going on it was too late, the runs and tries overwrote the files already.