aitotat wrote on 2023-04-28, 04:34:

8x speed is the fastest drive I accept. Faster are too noisy. I'd like these 8x drives would be a bit more quiet but these I can tolerate. Perhaps 6x drive is ideal but I've never had one of those.

I think 4x drives are ideal for DOS gaming. They are nearly silent, while providing enough speed for even the most demanding titles. For speed/noise testing, start the original Command & Conquer and play its lengthy intro movie. If it stutters, your drive is too slow, and if your drive sounds like a vacuum cleaner, it's too fast. Of course, there are also tools like CDBeQuiet! which can slow down faster drives to the desirable 4x speed.

BTW, another aspect that I found to be important with CD drives was whether they cause the game to freeze when changing between CD audio tracks (CDDA). Not sure what this depends on, maybe the buffer of the drive itself? Anyway, it can be tested by playing the first level of Tomb Raider (Caves). If you're moving Lara around at the exact moment when the background music track expires and restarts, you will notice that the game freezes for a second or so on lower quality drives. With better drives, this does not occur, and the track looping is seamless.

Joseph_Joestar wrote on 2023-04-28, 05:46:

I think 4x drives are ideal for DOS gaming. They are nearly silent, while providing enough speed for even the most demanding titles.

Well, not all of them. Mitsumi says HIIIIIIIIIIII WIRRRRRRRRRR HIIIIIIIII WIRRRRRRRRRRRRRRRRRR . and being period correct within a period correct strict budget means that. 😀

leileilol wrote on 2023-04-28, 06:02:

Well, not all of them. Mitsumi says HIIIIIIIIIIII WIRRRRRRRRRR HIIIIIIIII WIRRRRRRRRRRRRRRRRRR . and being period correct within a period correct strict budget means that. 😀

This is one of those comments that I could litterally *hear* loud and clear 🤣

Joseph_Joestar wrote on 2023-04-28, 05:46:

BTW, another aspect that I found to be important with CD drives was whether they cause the game to freeze when changing between CD audio tracks (CDDA). Not sure what this depends on, maybe the buffer of the drive itself? Anyway, it can be tested by playing the first level of Tomb Raider (Caves). If you're moving Lara around at the exact moment when the background music track expires and restarts, you will notice that the game freezes for a second or so on lower quality drives. With better drives, this does not occur, and the track looping is seamless.

That is a great test. I didn't even know there were differences like that. Anyway, I tested Tomb Raider (original CD) with 8x Sony, 8x Creative and 4x4 NEC and also with 8x Mitsumi that is on my DX4 system. Well, the only one that passes is the 8x Mitsumi (but I tested it on a DX4 system in case something else matters also). That was a surprise for me. I would have thought that at least the Sony is good. Both Sony and Nec had small pause but the Creative had most noticeable pause. It was easily the worst. It was also louder than Sony so I won't be using that. The Nec was really quiet and it is a nice drive.

Now what to do? I hope I can find 4x Panasonic drive for the DX4 system that originally had one. So I'll use the 4x4 NEC on this Pentium system and keep the Mitsumi on a DX4 system.

Not much progress lately since I haven't had time to do much anything. I did install the Noctua fan. It is very quiet but blows enough to keep the CPU cool. That small air flow makes a huge difference. I don't feel comfortable using a 1200rpm Noctua on a power supply. 1800 rpm would have been better there. I'll keep the Nexus in the PSU but I halved the resistance from before. I can finally start testing the graphics cards.

I just wanted quickly to do all the tests since I really want to get this system complete.

And the results are here.

What can I say? Both of the tested Expertcolor cards were the only S3 cards without any image quality issues. My quality tests are mostly on how compatible signal they produce for LCD displays. For proper tests I should have installed Windows and test with 1024x768 using refresh rates from 60 to 85. That would certainly produce much more differences among cards. So basically my tests are good for DOS use only.

Second observation is that for 2D performance it really does not matter what version of Trio64 or Virge you have. As long as you have at least 2MB memory (=64bit bus width that seems to be useful for VESA modes), it is the clock rate that matters. And there sure were differences!

Slowest S3 cards were the oldest Diamond Vision968 (I don't know the clock. MCLK did not support it) and noname Virge DX. That Virge DX only runs with 43 MHz while the fastest S3 was clocked to 75 MHz!!! What a difference! I tried to update Virge DX to reference BIOS but that got only 2 MHz more so I didn't run tests with it. I did test by overclocking it to 65 MHz. I was able to run tests with lots of graphics errors. It has only 60ns memory chips!!!

Both of the Trio64V2/DX cards did have issues with black levels (and the other card was otherwise good quality Expertcolor). I already updated some time ago the noname cards BIOS. It seemed to have 60 MHz clock. For the Expertcolor I used reference BIOS linked in this compatibility chart. That had 50 MHz clock like the original BIOS. I don't remember where I got the other one but it was better. For the Virge DX I used 2.01.16 BIOS linked in that chart. It wasn't much better but it did have VESA 2.0 support at least. But I restored the original BIOS because it was write-once-type of chip and I don't want to waste a good EEPROM for that Virge DX. It was a bad card.

So when selecting S3 card make sure:
* that is is a good quality brand (Expertcolor sure seems to be)
* It has fast memory chips
* you can update BIOS if necessary (because of the black level issue)

What about the other cards? Millennium is fast and quality card but it has compatibility issues with games like Jazz Jackrabbit. That is too bad because otherwise it would have been great for this system. Ati has some VGA compatibility issues as well and it was surprisingly slow on some tests but fast on Doom. Tseng is a old 1994 card and fast for a period correct build. But it also has compatibility issues. I didn't test them here but take a look at that compatibility chart.

TNT was the fastest and Voodoo 3 almost as fast. Both are free from issues both both are too new for this system. So I'll use the Expertcolor Virge.

aitotat wrote on 2023-06-17, 13:11:

So when selecting S3 card make sure: * that is is a good quality brand (Expertcolor sure seems to be) * It has fast memory chips […]
Show full quote

So when selecting S3 card make sure:
* that is is a good quality brand (Expertcolor sure seems to be)
* It has fast memory chips
* you can update BIOS if necessary (because of the black level issue)

Yeah, the image quality of S3 cards can vary greatly depending on the board manufacturer. Hercules, ELSA, STB and Diamond branded cards tend to be among the better ones.

I use a Hercules Terminator 64/Video (S3 Trio64V+ 2MB) which has excellent image quality and can even deliver a 120 Hz refresh rate at 640x480. This is without any third party tools, just using the included drivers and utilities from the official CD. Very useful if you're running SVGA DOS games (e.g. WarCraft 2) on a CRT monitor. Some screenshots can be seen here.

Hi there! It's perhaps not the most reliable source, but the '96 era MovieCD format mentions a Pentium 75 as a minimum requirement (or 486DX2-66, respectively)..

Re: Any codec to make MovieCD (MVI2) videos?
https://en.wikipedia.org/wiki/MovieCD

On the other hand, I had gotten a Pentium 133 a few years ago that
had a 1997 date code printed on the bottom of the AT chassis.

So I'd say that a ~100 MHz Pentium wasn't wrong for 1996.
Though Wikipedia says that 133 to 200 MHz models were already available in 1996.

In practice, the first Pentiums were released just two years earlier anyway (in '93, they were slow and not readily available yet).

Edit: I' m using an S3 ViRGE 325 in that Pentium 133.
Some older videos can be seen in my channel

There seems to be always some unexpected problems with my builds. I didn't find any 5.25" covers that would fit. I got this computer without any. I had one that could be made to fit with glue or something. But it was a bit too gray even after retrobrighting it. Remember, this case is more white than usual.

I could try to hunt down one but that could take forever to find correct one. Then I could use 5.25" floppy drive but that would have made this system look a bit too old. And I'm saving it for a 386 build. I do have another drive but that is too gray and not so good looking anyway.

I could add modern DVD-RW-drive so I could read everything. That would work and make sense but I don't like having two optical drives in this system and I want to keep this as 1996 as possible. I'll use DVD-RW-drive on a super socket 7 build.

Last option is to use one of those combo floppy drives that I have on my DX4 system. They look much more modern than typical 5.25" floppy drive. Too bad the other one had broken 3,5" drive. Maybe I can fix it? I have already cleaned and lubricated it but that did not change anything. I have some tested Teac 3,5" drives that had gone bad while they were in storage. That sounds like a capacitor problem so maybe this one is capacitor related as well?

Only one electrolytic capacitor in that 3,5" drive and it was leaking. I desoldered it and used some white vinegar because there seemed to be some minor corrosion. I didn't have any of those surface mounted electrolytics so I used normal through hole one. It is a very tight fit when you put the PCB back to the drive.

If you have fixed floppy drives, then you might already guess what will be the next problem. It did not work. No discs were detected. But since the capacitor clearly was broken and mechanics sound correct, then maybe heads are out of alignment? I tried to format a disc and it went perfectly. I was able to write it full and read it back. But it did not detect discs made in other drives.

But this was a problem I had caused. There is a track 0 sensor on that PCB so the PCB needs to be put back to the exact location where it was. And I didn't even mark it in any way. After couple of tries I was able to get it working. I had to push back the PCB as much as possible (I think the new capacitor is on the way) but that did the trick. I was able to successfully read all 9 original Windows 3.11 discs. So this is the drive I'm going to use. I'm very happy I got this fixed.

But I'm not done soldering just yet.

needs this mod

But it was hard to get that tiny chip installed. Key is to properly straighten the pins. They are too short. I did't get them straight enough and it was very hard to solder with gaps between pins. But eventually I managed. I had to test with multimeter few times that everything is connected and nothing is shorted.

The other mod to remove capacitors and shorting inductors was easy. Those capacitors (C232, C233 and C234) were not even installed on this card.

I would have liked to use Voodoo 1 made in 1996. This card is made in 1998. It doesn't matter really anything and besides, that Voodoo would have been a good ugrade for this computer in 1998. So no real complains. I installed those heatsinks last year. Originally it had no heatsinks and that is typical for Voodoo 1 cards.

So the Voodoo works but the 5,25" part of the combo disk drive did not! I was afraid I would have to remove it and fix that one too. But luckily head cleaning disk did the trick and made it work again. By the way, to make the 3.5" drive A and 5.25" drive B you need to connect the drive to the middle of the cable (before the twist). If you connect it after the twist, like A drive is normally connected, then 5,25" drive will be A and 3,5" drive will be B. But there is an option in the BIOS to swap the drives. I connected it to the middle since that made the cable fit better.

Edit:
But why Voodoo in the first place? Because this system is mainly for DOS games and I want some system to properly cover 3D accelerated DOS games. And Voodoo 1 is simply perfect for that as nothing else will be as compatible with 3D accelerated DOS games and Voodoo 1 was released in 1996.

This system is not so simple looking as I first intended. Nothing wrong with these drives but I would have preferred normal 3,5" disk drive and the simple looking Sony CD-ROM-drive (and that self made CF-HDD-drive). I tried to get the cables connected in such way that they would affect air flow as little as possible. This case certainly isn't the best for air flow. But it is good enough.

Finally time for sound cards. I already installed Orpheus II with X2GS so no surprises here. But actually I'm going to install another sound card as well.

Oldest cards are Sound Blaster AWE32 CT3900 and GUS ACE. CT3900 might not have been available anymore in 1996 but GUS ACE certainly was. I know this since I bought my rev 1.0 in 1997. The one in picture is rev 1.1 and it is a good thing I took that picture since

Those are Nichicon branded so they should be good but the two big black ones are leaking and there looks to be some minor corrosion already. I need to take care of that fast. So if anyone has one of these cards (or any rare card), do take look at the capacitors.

Everybody knows GUS by now and ACE is the card that is meant to be used with some other (SB compatible) card. That makes GUS ACE easiest to setup and use. It is a great card. But there are GUS PnP pro and ARGUS in that picture as well. PnP Ultrasounds used to be hardest to configure but Unisound made that very easy. I don't recommend to use those Gravis initialization utilities anymore. But you'll want to install GUS PnP drivers anyway to get sound patches etc. I don't need a separate GUS on this system because Orpheus II already has one (GUS PnP) included.

Sound Blaster AWE 32 CT3900 is the oldest card I selected. It is very much possible that this card was not available in 1996 since this is the last full non-PnP AWE32. I used to prefer non-PnP cards before Unisound. Now I prefer PnP cards mostly because you can disable FM synthesis and use another card for that without fear of any conflicts. And CT3900 does have the DMA clicking bug and that is the biggest fault with this card.

Perhaps the best Creative card in the list is AWE64 Gold. But is it really the best? It has CQM and no wavetable header and you can't use SIMMs to expand memory (but it does have 4MB RAM included). So basically you would want to use this with some SB Pro clone but with Orpheus II there is really no need to use this at all.

Then there is SB 16 CT2940. This is my favorite SB16. Good sound quality, this card has real Yamaha OPL3, you do have (buggy) wavetable header but not other bugs because it has Vibra pro. This + GUS ACE would be an excellent period correct pair.

Then there are two Sound Blaster 32 cards. CT3600 is basically like the CT2940 but without wavetable header but with EMU8000 instead. But the card in picture has CQM. These cards can be fitted with real Yamaha OPL3 and the next step is to de-CQM this one (actually I have three and I'm going to de-CQM them all). So this is the card I'm going to use with Orpheus II. But Why? To get second real Yamaha OPL3 so I can use dual OPL2 emulation with MIDIto for the few games that supports it. And the AWE sounds might be nice for the few games made for it.

Then there is Sound Blaster 32 CT3670. This is the AWE64 with SIMMs. So it fixes the least significant issue that the AWE64 Gold has. And it has IDE as well but since it is behind a PnP it is kind of useless (just like with CT3600 and CT2940). You already have two IDEs on PnP systems so you don't need it and for the older systems the IDE won't be available before driver initialization. If you would have bought Sound Blaster 32 for this system in late 1996, then maybe this card is what you would have got as the CT3600 is the older of the two.

Then there is MK8330. New card just like Orpheus II. And also with audio chip from 1997. So neither are really period correct for this system. But both are very good cards.

There is Yamaha 71x based card in the picture. That card is from 1997 but 71x was available in 1996 (at least I think it was) so this is a good period correct card for this system. It has SB Pro and WSS compatibility with real Yamaha OPL3. One good thing is that it can support SB and WSS at the same time unlike some other SB Pro clones.

Finally there is Terratec Gold 16/96. A very good SB Pro clone. No real Yamaha OPL3 but with a very good ESFM instead. This card would be a really good looking pair for AWE64 Gold.

They are tested and working. One of them gave me some trouble. It wasn't detected at all after I had done the mod. Soldering looked to be just fine. I assumed that most likely the OPL3 chip is broken since surely the card would have been detected if just the YMF DAC would have been broken. So I desoldered the OPL3 and tested the card again. It was now detected properly. Then I installed last of my four OPL chips to the card and that was a working one. Looks like a bad OPL3 can make the card completely dead. So three out of four China OPL3 chips worked. Could have been worse. I was going to use that last one to mod one of my CT2940 cards but not this time it seems.

Finally I recapped the problematic card. I'm going to use that one. I didn't have new 2.2µF capacitors so there are two old ones on the card.

Nothing special really, just a PCI version of those common 3Com ISA cards. I wanted a 1996 PCI network card but this one is made in 1998 but at least this model was available in 1996. I could have used ISA version since both are 10Mbps. Maybe I'll replace this later to something. But I don't need it much so it does not matter.

There is one free shared ISA/PCI slot. But it is a tight slot. X2GS on the other side and Voodoo 1 heatsinks on the other side. Hard disk is 120GB 5400rpm Samsung and then there is the CF-HDD-drive with 60 GB Toshiba 1.8" HDD. Not exactly 1996 drives but I want this to be quiet. If I would have wanted this to really be a 1996 high-end system, then I should have installed SCSI controller to the last remaining slot and used perhaps 3 GB SCSI HDD. But I don't even have those.

Orpheus II has 16 MB RAM and SB32 CT3600 has 8 MB.

Now we need a monitor, keyboard and mouse. And speakers too.

That is the same model that I had on my 1996 system back then. It is a bit yellow but I could retrobright it. It also needs some minor adjustments (Focus and G2 voltage a little bit). But I like that monitor much and it is a good looking one. And it is made in 1996, just perfect for this system, although high-end system from 1996 would have surely used more expensive monitor.

Actually that is made in 1997. I don't know was it released in 1996 or 1997 but close enough. I've long wanted to find a 15" monitor with Trinitron tube and here it is. Perfect timing for this system.

Except this is a bit too good monitor for DOS gaming. Picture is more what you would expect from a 17" tube. I like 15" monitors becase they are new enough for lots of controls and no black bars. But small enough for not having visible scanlines (I know some like them and they do have their use).

I shouldn't complain. That monitor looks excellent on this system and sure is a high-end 15" monitor.

I wanted to use a Keytronic keyboard and I have one in perfect condition. It is currently on my Copam because that was very picky what keyboard works with it. So I chose that IBM instead. It is not in best condition and needs cleaning. But it is a good keyboard even though it is not mechanic. Mouse is MS mouse with wheel. MS serial mouse from my 1996 system did not have a wheel so that wheel mouse is a bit too new. I still have the mouse from my 1996 system but I'll use it on DX4 system.

I was trying to find when the Nokia was released. Didn't find it but if in 1996, then it must have been very late. More likely it is released in 1997. But I did find something else. A monitor review from 1996. Older Nokia won that one but there was a Philips 105A with very good scores. So...

It doesn't have OSD so I thought it was from 1995 and I was going to use it with DX4 system. But this is likely a 1996 monitor even though it does not have OSD. And it doesn't need one. It is very fast to adjust with those buttons on front. In fact much faster than from most OSD menus. This monitor looks good in pictures but it is very yellow from both sides and behind. I was going to try how hard it is to retrobright a monitor starting with this.

This is what to expect from a 15" monitor. Sure, the scanlines can be seen but they are so thin that they disapper from normal viewing distance. I think the scanlines bother me because they make the scandoubling visible and that is not what I like.

So the Philips is a qood quality CRT (and doesn't require any internal adjustments) but is it more suitable for gaming than the Nokia? Hard to tell. Trinitron has better black and better colors and you can clearly see the difference. Nokia also has color temperature adjustment while the Philips does not. I think games look best with warm colors and Philips has cold color (more suitable for reading text).

You can see different speakers now, they are more period correct. I used to have similar. Different brand, no surround button and different knobs. But same shape and size otherwise. I'm not sure but I think I bought mine after I bought GUS ACE in 1997. That would have made sense since the GUS is so great card that it needed better speakers. But forget about period correct computer speakers. These sound thin and crap compared to those black Creative T40 speakers. I really like both T20 and T40 speakers. You can connect two sound cards at the same time to minimize need of an external mixer and those speakers sound great. And they have been available for so long that you can find reasonable priced used ones (I think I have three sets of T40 speakers and two sets of T20 speakers). I bought my first T40 speakers new in late 2008. All the rest have been used when I bought them. I had to fix those first T40 speakers (and those are the same seen in picture) perhaps a year ago. There was a power mosfet that had shorted.

aitotat wrote on 2023-07-09, 08:53:

So the Philips is a qood quality CRT (and doesn't require any internal adjustments) but is it more suitable for gaming than the Nokia? Hard to tell. Trinitron has better black and better colors and you can clearly see the difference. Nokia also has color temperature adjustment while the Philips does not. I think games look best with warm colors and Philips has cold color (more suitable for reading text).

The monitor that I bought back in 1996 along with my first PC was a Phillips Brilliance 15". Not sure about the exact model, but it had a pair of speakers mounted on the side, which provided surprisingly decent sound. Unfortunately, it broke down after a few years of use. I think I may still have it somewhere in my attic, but it hasn't been working for 20+ years.

The scanline thing is definitively personal preference, and may depend on which type of monitor you grew up with. Trinitron vs. Shadow Mask is an interesting debate, and I remember it being discussed frequently on popular hardware forums back in the day. Nowadays, I'd just be happy with a fully functional CRT monitor, as it seems to be getting more and more difficult to keep them in good working condition.

Joseph_Joestar wrote on 2023-07-09, 09:09:

The monitor that I bought back in 1996 along with my first PC was a Phillips Brilliance 15". Not sure about the exact model, but it had a pair of speakers mounted on the side, which provided surprisingly decent sound.

I forgot to mention that the Philips indeed have stereo speakers build in. I should try if they are better than those old computer speakers.

I compared those computer speakers and Philips build-in speakers and the computer speakers are easily better.

I found a problem. CD-ROM-drive is no longer detected after resetting the computer. It is detected when powering on the system. My guess is that this is some sort of master-slave-compatibility issue. 120GB Samsung is primary master and the much older CD-ROM-drive is primary slave. I could first try 40-conductor cable since now I have 80-conductor cable if I remember correctly. If that does not help, I could remove the 1.8" Toshiba HDD since it is quite useless at the moment. That would make the CD-ROM secondary master and CF-to-IDE-adapter secondary slave. But I don't think the cables are long enough for that.

I could connect CD-ROM-drive to SB32 since it has IDE. But it is a PnP card so I would have to load CD-ROM-drivers with devload on autoexec.bat after initializing the card. That doesn't sound that problematic but I use CALL SNDENV.BAT at the end of autoexec.bat to initialize the sound card. This makes managing multiple sound cards easier and prevents anything messing with sound environment variables since they are not in autoexec.bat. But I can't just place devload in there that easily. And I would have to execute mscdex and perhaps smartdrv after CD-ROM-drivers. Essentially I would have to abandon the SNDENV.BAT and set everything in autoexec.bat. Damn that PnP...

If that is too hard or causes other issues, then as a last option I just have to replace the CD-ROM-drive.

Edit:
Looks like 40-conductor IDE cable was enough to fix the issue. CD-ROM-drive is now detected every time after resetting with CTRL-ALT-DEL or with reset button.

aitotat wrote on 2023-07-09, 16:24:

I could connect CD-ROM-drive to SB32 since it has IDE. But it is a PnP card so I would have to load CD-ROM-drivers with devload […]
Show full quote

I could connect CD-ROM-drive to SB32 since it has IDE. But it is a PnP card so I would have to load CD-ROM-drivers with devload on autoexec.bat after initializing the card. That doesn't sound that problematic but I use CALL SNDENV.BAT at the end of autoexec.bat to initialize the sound card. This makes managing multiple sound cards easier and prevents anything messing with sound environment variables since they are not in autoexec.bat. But I can't just place devload in there that easily. And I would have to execute mscdex and perhaps smartdrv after CD-ROM-drivers. Essentially I would have to abandon the SNDENV.BAT and set everything in autoexec.bat. Damn that PnP...

If that is too hard or causes other issues, then as a last option I just have to replace the CD-ROM-drive.

Edit:
Looks like 40-conductor IDE cable was enough to fix the issue. CD-ROM-drive is now detected every time after resetting with CTRL-ALT-DEL or with reset button.

Could you just make two directories with ctcm.exe and ctpnp.cfg? One for loading via config.sys to initialize the ide port. And then the other when you initialize the sound in your batch script. I think it defaults loading the cfg from where-ever you call ctcm.exe from, unless you set the CTCM environment variable. I guess if your onboard is fine no big deal, but pnp was supposed to be magical right?

It started to rain and the front bezel needs more time in the sun. Of course it had to be that part since I can't start putting the monitor back together without that one.

Since when did retro computing started to be like this? Taking apart a CRT monitor for nice a sunbathing in summer. Soldering iron is required in every build I do nowdays. And things get even more weird. I bought a RAL color fan.

I took apart the other Keytronic keyboard, washed it and retrobrighted it and finally put it back together. And every key works. I was going to take a before and after photos but those before photos look much better that it was. I think my phone though the yellow color was because of bad lightning and compensated it. Keyboard was ugly yellow. Now it looks like a new.

This keyboard is a smaller (read cost reduced) Keytronic. It does have the same feel as the bigger one and this one is very easy to take apart and put back together. I tried to take a photo with the RAL color fan but it does not look correct. The color is RAL 9002 Grey white. And so is the painted metal cover of the computer. So a perfect match. Front bezel is a different color but I did not find a match. I'll have to retrobright it first.

But the IBM is a better keyboard, no question about it. It has taken a lot of beating. Looks like someone has melted it with a soldering iron or thats what I first though. But more closely inspection reveals that it is likely the softeners used in the cord that has melted the keyboard but mostly from under the keyboard. Keys are fine fortunately. So I'll have to see what I can do for that IBM. It is model KB-7953.

Any idea what can be done about the cord? Is it enough to clean it with IPA or do I need to replace it?