auron wrote on 2025-02-09, 18:29:

amadeus777999 wrote on 2025-02-06, 21:05:

I installed a SCSI card in the PackardBell P60 I got today and checked if this would yield some performance gains compared to the IDE controller - it did but it was a "µ-improvement".
In DOS, DOOMII / demo3, the SCSI's fastest time to the IDE's slowest time was 4522 to 4525 realtics - so less than 1/10 of a percent.

doom won't really show much unless you test with only 4 MB RAM. while the game doesn't make perfect use of additional RAM to preload everything, the amount of data it loads on the fly will be small. i would test with some game running at just the minimum required RAM, so there will be some swapping and you can judge whether the drops are worse on SCSI than IDE. but this tends to be a bit of a wash because the drives themselves will be different.

alternatively, hdtach 2.60/2.61 or adaptec threadmark. these show best how poorly the RZ1000 performs after all the bug patching done to it.

Thanks - that clears things up a bit.
I have dis-assembled the setup but I, at least, had speedsys test the SCSI hd and it blew away the ide setup - former was also way newer.
Years back, when these things were en vogue, and foreign to me, I remember a friend of mine(who had the dough) showing me his SCSI based PC and it was shocking how fast the hard drive and the cd rom were... it really did make a difference.

Working on getting my DOS system flushed out.

Installed GeoClock 6 and Screen Thief. Got my "quick launch menu" setup . Added more games to LaunchBox. 😀

I had completely forgot about GeoClock, used to love that app! Still do.

StriderTR wrote on 2025-02-09, 19:19:

Working on getting my DOS system flushed out.

Installed GeoClock 6 and Screen Thief. Got my "quick launch menu" setup . Added more games to LaunchBox. 😀

I had completely forgot about GeoClock, used to love that app! Still do.

Launchbox has a game limit, be careful not to hit it as it will start to corrupt your save file after that. I forget what the limit is, but I hit it. I also contacted the guy who wrote it and he is aware of the limitations but as expected hes not focused on this niche program right now.

I reflowed and soldered the TMU legs of the V2 I was repairing.

Mojo detects both TMUs now and Quake 2 @800x600 yields 689 frames, 21.1 seconds: 32.7 fps.
Bench was launched from within W98.

Are those good for a P-233 MMX?

Thermalwrong wrote on 2025-01-15, 03:48:

*sigh* The other day I found that the repair I did on my Toshiba Tecra 730CDT was not working. The keys that are broken are quit […]
Show full quote

Thermalwrong wrote on 2021-08-15, 23:47:

I didn't just have one laptop that had a bad keyboard. I bought a Toshiba Tecra 730CDT recently to get hold of a Toshiba DeskSta […]
Show full quote

I didn't just have one laptop that had a bad keyboard. I bought a Toshiba Tecra 730CDT recently to get hold of a Toshiba DeskStation V - both were horribly rusted and corroded, but both mostly work.

The dock likes powering off after a few minutes of use, I'm still wondering where to start diagnostics, could it be heat since the fan by the PSU doesn't appear to be doing anything, or maybe capacitors somewhere, either on the board or hte integrated power supply?
I've so far dismantled it, put it back together and removed the pins in the dock's lock so it can be put between regular and emergency eject mode without a key.

The Tecra 730CDT was so badly corroded that I had to fully dismantle it and carefully remove the electrolyte corrosion from the magnesium frame. It's been working but the keyboard's Up key and maybe some others weren't working. A replacement keyboard would cost £30 and I'm not interested in paying that, especially having fixed one keyboard, so I've worked to fix this one too.

These '90s laptops have nice big keyboard frames that use plastic stakes (sometimes screws) to hold in the membrane to the keyboard's backplate.
I pulled the keyboard apart, this time using the soldering iron at 200C on each of these pegs so that they could deform without breaking, for later use.
Took apart the keyboard and started following the traces on both sides but doing it just by eye is more work than it's worth. I taped it to the desk to get it flat then took a picture of both sides and drew a layer over the top with lines to visualise it. Nothing wrong with the bottom side:

The attachment keyboard-repair-underside.jpg is no longer available

The topside is much more complex to follow - notice those little black traces connecting each pad, each one is about a 10k resistor. Following it, the discolouration was visible from my picture, one of the top traces at the upper edge of the keyboard has discoloured and the connection has gone from 20 ohms to 60k ohms:

The attachment keyboard-repair-topside.jpg is no longer available

I tried using the conductive paint again, but it didn't do much good. I found a better way this time - the self adhesive copper tape is sticky on one side and conductive on the other. Stick the conductive tape over the part that needs replacing and a few mm longer, fold over the ends so the conductive side is touching the membrane trace at each end. That bridges the connection and provides a trace that's held in place, but maybe a fresh conductive ink pen would be better.

The attachment IMG_11589.jpg is no longer available

Then put Kapton / polyamide tape over the top which keeps it connected. Then I put some extra Kapton tape over the copper just in case it could short on anything.

The attachment IMG_11579.jpg is no longer available

Tested it one last time and the up key now works, no random key presses so it's all melted back together, again using the soldering iron to press on them but with a bigger round tip to squish it nicely:

The attachment IMG_11599.jpg is no longer available

Yay! now the Tecra 730CDT is back to 100% working. I've got quite a full set now - Tecra 730CDT, Tecra 740CDT and Tecra 750CDT 😁

*sigh* The other day I found that the repair I did on my Toshiba Tecra 730CDT was not working. The keys that are broken are quite important so I had to take the keyboard apart again, using the soldering iron to undo all the heat stakes so I could check out what was wrong.
In hindsight, the repair I did here may not have been the fault, I think, because there was a break further up the line where corrosion damage eventually killed some more of the trace. Quite possible the conductive adhesive for the copper wasn't making good enough contact any more. Conductive paint is much better but harder to apply.
I pulled off my copper tape repair which resulted in ripping the membrane on the edge. 🙁 So I had to clean off and re-route three keyboard traces through the smaller area.

The attachment IMG_5093 (Custom).JPG is no longer available

At least now I've learned some very efficient methods for doing keyboard membrane trace repair. Now I use the cheap silver conductive paint that you get from china in a syringe, but to apply it I take off the needle part and use a rubber thermal paste spreader nozzle thing as a brush to get a nice even coat. To lay down the new traces I use some cheap polyimide tape (which has poor adhesion) as masking tape to get the lines, then paint on the new trace.
This conductive paint has to be heat cured though, so I use my USB-PD hotplate with a thermal pad on top to bring it up to 100c and that then starts conducting and I can then check for crossover with other traces, scraping off paint where that occurs.
This method works really well, having the thermal pad and hotplate keep the membrane in place makes it easier to work on.

Got another laptop in with a bad keyboard, the Armada 1700 - it was missing enter key and a bunch of others so was basically un-usable. UK keyboards for the Armada 1700 / 1750 series are unavailable so of course I had to take it apart and do repairs 😁 This keyboard uses a mix of screws and plastic stakes to hold the keyboard together, which makes it do-able. Most modern keyboards don't come apart like this and require removing every key part until the sheets can be removed, which is not worth it imo.

This laptop keyboard membrane repair process is getting simpler each time I do it:

1. Take keyboard apart and find the bad keys. *easy*
The plastic heat stakes can be undone with a soldering iron with a big hoof tip at ~220C, just hold it on the heatstake and twist a pry tool in between to pry up that segment. Do that for a while until its all apart. I had to clean corrosion off the aluminium plate so that it was flat again since anything uneven in there could cause keys to be bad or press randomly.
Trace the lines with pen on the clear side. Use a multimeter in resistance mode to check for connections. This keyboard has a green non-conductive layer to mask & protect traces, that can be cleaned with a cotton bud soaked in IPA with some rubbing to get to the silver conductive layer to test. Check the traces against the key pads (darker conductive paint)

2. Use low-tack kapton tape to mask areas. Use a small rubber spreader to dab a small amount of the conductive paint from the syringe and spread a thin layer of conductive paint over the masked area. It works best if it's kept fairly thin & even. Then using USB-C hotplate with a thermal pad attached, heat the area to ~110C for a bit and check resistance with a multimeter to see in realtime whether the connection repair has worked. Mostly it cures in a minute or so and I generally see less than 10 ohms on the repaired segments.
In this trace repair, the bridge that connects one trace to another seems to have lost its connection. It goes over another trace so I have to be careful to keep within the dark green line area that's properly masked off from the trace beneath.

3. Once that's done, put the membrane sheets onto the aluminium plate of the keyboard so that they're loosely reassembled. Now plug that into the laptop and test whether all the keys are fixed. To use this you'll have to find some way to press the sheets together if the keyboard has just the one cable - there should be an area with a bunch of thin vertical pads which is how the two membranes connect together. To test the keyboard while it's loose, these have to be pressed together for the keys to work. I was able to use 4 fingers on one hand to hold them together then the other hand to type. These connections don't have to be held together all the time and can be left un-pressed and re-pressed safely while the computer is on.
If it's still not working properly then re-check connections and check the other sheet for any faults.

Here you can see an example of how damaged membrane traces look, they're discoloured compared to the more silver / green traces that still work. Also some secret graffiti inside the membrane layers with a heart shape and a company logo (not pictured):

This repair was frustrating but rewarding, repaired one trace and then found some more keys that don't work. Rinse and repeat until in total 3 traces are repaired and they all overlap.
It seems that I got away with using a thin application of UV soldermask to stop the trace repairs contacting each other even though they overlap:

Re-test and once everything is working then the keyboard can be reassembled and re-heat-staked back together. To prep the heatstakes I usually use pliers to squish the plastic stakes back into a shape that will fit through the aluminium holes and usually they'll 'click' back into place, once all the stakes are back in place I used the soldering iron at ~220c to press them for a couple of seconds to make a nice flat shape again.

Once that was fixed I had to clean coffee stains out of the LCD panel... But the sound system on this Compaq Armada 1700 is fantastic, ESFM never sounded so good 😀

Shagittarius wrote on 2025-02-09, 20:52:

Launchbox has a game limit, be careful not to hit it as it will start to corrupt your save file after that. I forget what the limit is, but I hit it. I also contacted the guy who wrote it and he is aware of the limitations but as expected hes not focused on this niche program right now.

Yeah, I remember reading somewhere that it had problems with hitting a memory limit if you added too many games, but I think that was a pretty high number. If I recall correctly, it was in the 300's. I would be lucky to even get anywhere near 100, so I'm not all that concerned about it. I only have a total capacity of 4GB (2x2) in my DOS system, and I only use LB for games, nothing else. I would likely run out of drive space long before I hit LB's limit on games. 😜

It might be 127 or 255, if it's an overflow bug.

If you can call it a retro activity .. i voted for my favourite games in the GOG dreamlist https://www.gog.com/dreamlist/

I tried the search function but surprisingly haven't come across a thread about this yet?

If anyone wants to help vote NFSU2 (and Most wanted) up please do :p

I do however doubt that it will actually make much difference as the licensing for the music and cars has always been an issue ... hence why it never made it to GOG in the first place. I am interested to see if they can actually get the games that are most voted onto the platform.

No one lives forever and Freelancer are in the top 5 so would be cool to see those come back.

G-X wrote on 2025-02-09, 22:49:

If you can call it a retro activity .. i voted for my favourite games in the GOG dreamlist https://www.gog.com/dreamlist/ […]
Show full quote

If you can call it a retro activity .. i voted for my favourite games in the GOG dreamlist https://www.gog.com/dreamlist/

I tried the search function but surprisingly haven't come across a thread about this yet?

If anyone wants to help vote NFSU2 (and Most wanted) up please do :p

I do however doubt that it will actually make much difference as the licensing for the music and cars has always been an issue ... hence why it never made it to GOG in the first place. I am interested to see if they can actually get the games that are most voted onto the platform.

No one lives forever and Freelancer are in the top 5 so would be cool to see those come back.

It would be super cool if they put out the games and got a bunch of unsigned bands to fill out the soundtrack.

I dug out my Socket 479 system and found that with some different memory, I was able to push the front side bus 160 MHz! Everything is working fine with 53 MHz PCI speed. I've played both Doom 3 and Half-Life 2 for hours this way with no issues.

What pci cards are installed?
Softfsb and the datasheet might show more fsb settings. You're remarkably close to 166 which should drop the pci back down.

myne wrote on 2025-02-10, 01:39:

What pci cards are installed?
Softfsb and the datasheet might show more fsb settings. You're remarkably close to 166 which should drop the pci back down.

I'm using an Audigy 2 ZS, no other cards installed. It never occurred to me that this chipset might have a 1/4 divider. I've gone as high as 163, but it crashed in 3D 03. Unfortunately, this board has no voltage options (other than increasing CPU voltage).

It's not the chipset. It's the clockgen.

Usually a small rectangular chip with 48 pins and a metal lid oval oscillator nearby.

Brands are usually ICS, realtek, Cyprus
Find that, find the datasheet, find the list of clockspeeds

Nexxen wrote on 2025-02-09, 21:06:

Mojo detects both TMUs now and Quake 2 @800x600 yields 689 frames, 21.1 seconds: 32.7 fps.
Bench was launched from within W98.

Are those good for a P-233 MMX?

I'm getting 41,5 fps in 800x600with high quality sound, 42,7 fps with low quality sound (and max performance sound) in map1.dm2 on a P-233 MMX with 64MB SDRAM on a 430TX board with 512KB cache using fastvoodoo2 4.6 drivers. Do you have vsync disabled ingame and in drivers (under "Voodoo 2 OC" settings)? Did you use the MiniGL driver for Quake2? You should be getting around 40fps, see Phil's Voodoo 2 scaling project.

Siran wrote on 2025-02-10, 07:14:

Nexxen wrote on 2025-02-09, 21:06:

Mojo detects both TMUs now and Quake 2 @800x600 yields 689 frames, 21.1 seconds: 32.7 fps.
Bench was launched from within W98.

Are those good for a P-233 MMX?

I'm getting 41,5 fps in 800x600with high quality sound, 42,7 fps with low quality sound (and max performance sound) in map1.dm2 on a P-233 MMX with 64MB SDRAM on a 430TX board with 512KB cache using fastvoodoo2 4.6 drivers. Do you have vsync disabled ingame and in drivers (under "Voodoo 2 OC" settings)? Did you use the MiniGL driver for Quake2? You should be getting around 40fps, see Phil's Voodoo 2 scaling project.

Make sure that you're not using too much memory in your 430TX as well.

Intel 430TX performance comparison with 64MB and 128MB memory

Kahenraz wrote on 2025-02-10, 07:25:

Make sure that you're not using too much memory in your 430TX as well.

Intel 430TX performance comparison with 64MB and 128MB memory

Very true, that's why I'm using 64MB on mine. Not sure if OP also has a 430TX board though. Tested 256MB vs 64MB on my board (also a AN430TX btw) and 3dmark score with the V2 jumped almost 14% (from 1198 to 1361) with 64MB due to the cacheable area limit.

Siran wrote on 2025-02-10, 07:14:

Nexxen wrote on 2025-02-09, 21:06:

Mojo detects both TMUs now and Quake 2 @800x600 yields 689 frames, 21.1 seconds: 32.7 fps.
Bench was launched from within W98.

Are those good for a P-233 MMX?

I'm getting 41,5 fps in 800x600with high quality sound, 42,7 fps with low quality sound (and max performance sound) in map1.dm2 on a P-233 MMX with 64MB SDRAM on a 430TX board with 512KB cache using fastvoodoo2 4.6 drivers. Do you have vsync disabled ingame and in drivers (under "Voodoo 2 OC" settings)? Did you use the MiniGL driver for Quake2? You should be getting around 40fps, see Phil's Voodoo 2 scaling project.

Kahenraz wrote on 2025-02-10, 07:25:

Siran wrote on 2025-02-10, 07:14:

Nexxen wrote on 2025-02-09, 21:06:

Mojo detects both TMUs now and Quake 2 @800x600 yields 689 frames, 21.1 seconds: 32.7 fps.
Bench was launched from within W98.

Are those good for a P-233 MMX?

I'm getting 41,5 fps in 800x600with high quality sound, 42,7 fps with low quality sound (and max performance sound) in map1.dm2 on a P-233 MMX with 64MB SDRAM on a 430TX board with 512KB cache using fastvoodoo2 4.6 drivers. Do you have vsync disabled ingame and in drivers (under "Voodoo 2 OC" settings)? Did you use the MiniGL driver for Quake2? You should be getting around 40fps, see Phil's Voodoo 2 scaling project.

Make sure that you're not using too much memory in your 430TX as well.

Intel 430TX performance comparison with 64MB and 128MB memory

Thank you both for the help!!

This is the board I have. I was sloppy and just kept it with the V2 as it worked. I put 128 MB for W98 purposes only.
https://theretroweb.com/motherboards/s/pcchips-m530

At this point I think it's good to upgrade the system with a better board.

V2 drivers are Fastvoodoo 4.6, no patches applied. I'll try it with my P-III 1000.
In Quake2 it gave 3 options for the renderer, software, openGl (iirc) and another Glide something.

Last thing, do you know where to find a SLI cable how-to DIY guide? It'd be fun to try the 3DFX SLI 😀

Nexxen wrote on 2025-02-10, 11:52:

Thank you both for the help!! […]
Show full quote

Thank you both for the help!!

This is the board I have. I was sloppy and just kept it with the V2 as it worked. I put 128 MB for W98 purposes only.
https://theretroweb.com/motherboards/s/pcchips-m530

At this point I think it's good to upgrade the system with a better board.

V2 drivers are Fastvoodoo 4.6, no patches applied. I'll try it with my P-III 1000.
In Quake2 it gave 3 options for the renderer, software, openGl (iirc) and another Glide something.

Last thing, do you know where to find a SLI cable how-to DIY guide? It'd be fun to try the 3DFX SLI 😀

You're very welcome!

The Voodoo 2 does scale pretty well as you can see in Phil's documentation, so a PIII-1000 will make great use of it, especially in SLI. I recommend using the linked MiniGL driver (it comes with an installer) so you get the option to choose the 3dfx renderer in Quake II which is faster, don't forget to disable vsync in the Voodoo 2 OC settings of your drivers, otherwise your PIII-1000 will most likely not produce more than 60fps even in SLI.

The PCChips board you use has a 430VX chipset that also has the 64MB cacheable memory limitation, so reducing your RAM (as counter-intuitive as that sounds) to no more than 64MB will improve your performance under Windows.

As for the SLI cables, Vogonswiki has a DIY guide for how to mod a FDD cable to use as SLI

Siran wrote on 2025-02-10, 12:47:

You're very welcome! […]
Show full quote

Nexxen wrote on 2025-02-10, 11:52:

Thank you both for the help!! […]
Show full quote

Thank you both for the help!!

This is the board I have. I was sloppy and just kept it with the V2 as it worked. I put 128 MB for W98 purposes only.
https://theretroweb.com/motherboards/s/pcchips-m530

At this point I think it's good to upgrade the system with a better board.

V2 drivers are Fastvoodoo 4.6, no patches applied. I'll try it with my P-III 1000.
In Quake2 it gave 3 options for the renderer, software, openGl (iirc) and another Glide something.

Last thing, do you know where to find a SLI cable how-to DIY guide? It'd be fun to try the 3DFX SLI 😀

You're very welcome!

The Voodoo 2 does scale pretty well as you can see in Phil's documentation, so a PIII-1000 will make great use of it, especially in SLI. I recommend using the linked MiniGL driver (it comes with an installer) so you get the option to chose the 3dfx renderer in Quake II which is faster, don't forget to disable vsync in the Voodoo 2 OC settings of your drivers, otherwise your PIII-1000 will most likely not produce more than 60fps even in SLI.

The PCChips board you use has a 430VX chipset that also has the 64MB cacheable memory limitation, so reducing your RAM (as counter-intuitive as that sounds) to no more than 64MB will improve your performance under Windows.

As for the SLI cables, Vogonswiki has a DIY guide for how to mod a FDD cable to use as SLI

Thanks again!! I'll follow instructions.
I had Vsync On. I checked the pics I took. COAST is 256KB L2. Yes, 128MB is too much.

I'll also add a fan, as in a recent post someone pointed out that faster cpu = higher temps on the V2 cards.
When they came out I could only read about it in magazines... now I have 2.

Nexxen wrote on 2025-02-10, 13:13:

I'll also add a fan, as in a recent post someone pointed out that faster cpu = higher temps on the V2 cards.

Good choice, I used a 80mm fan that's attached to my case's side panel to cool the V2 without needing to modify the card itself.