PCI Graphics Card Roundup S3 Cirrus Logic Trident Tseng 

i loved my dual m3d cards...was just fun stuff back in the day... a buddy wanted the cards so bad he bought me a tnt2 ultra for them and the 8mb videocard i used with them... hes still got them... in the best board for the combo of parts he used that was made..even though the total parts to do the "dream build of the era" was....more then a new ryzen 5 system....heh...

@@AshenTechDotCom Yeah. Matrox mystique was the way to back in the 90s. I had a multi monitor setup back when it blew peoples minds. The arrow goes from one monitor to the on beside it. The weird thing is I'm still a dev and actively programming but I rarely use two monitors. I have one 1440 curved monitor and i Split stuff. It's all good. My code line lengths have got ridiculous... :-) I never bought into the whole 80 chars a line bullshit. Luv and Peace.

ha ha ha me to

Agreed! They are really good cards.

HPZeta Cirrus Logic cards also tend to have better VGA DACs than most S3 cards.

@@lordmithras47 I know I'm forever behind in this conversation but man you are not kidding. My S3 has some pretty awful palletes

@@amberselectronics : Definitely, there was a lot of variability in quality from OEM to OEM. Generally, Diamond Multimedia's S3 cards were good.

So true :D

I remember I have read the review on it in PCRevue magazine .

Both of those games don't use a 3D accelerator, everything is rendered on the CPU and the graphics card just draws the images.

i use a S3 Trio 64V+ in my Pentium 133 rig and it's pretty compatible, works fine with both Windows 3x and 9x.

True! Good post!

Great bargain.

1000 games? Nice, that certainly puts a lot of weight on your opinion. I tried far less but still formed a good opinion.

Haha, this hobby keeps you young :D

Tell that to my newly discovered grey hairs..

>

My take on DOS game compatibility is that people like to focus on the 0.1% of games that have issues and then declare what is or isn't a good DOS card 😅 For the 99.9% of games it really doesn't matter and they will look and play just fine.

Quite strange hearing the guitar solo in Descent 2 without the pitch bend, isn't it? lol btw, the whole synth core is using pure logic, so no CPU / MCU cores or anything like that. The MIDI files are being played by my external player though (based on the "MrMIDI 2" project online), but the MIDI input goes directly into an IO pin on the FPGA board, and then it just spits out the audio, and that's it. The latency must be VERY low, because it's grabbing the full set of 32 samples in only 66 clock cycles (at 24 MHz), so that only takes around 2.7 microseconds (whereas receiving a MIDI byte takes a whole 320us). The 32 sample outputs are updated faster than the highest note playback, so the output is running at around 93 KHz atm. The main limiting factor is the speed of the Flash chip (70ns), but I could still bump up the main clock to much higher rates, and probably manage to do 88 KHz samples etc. I had an AWE32 card in the 90s, and loved just listening to MIDIs all day, and messing with keyboards etc. I also used to repair and use many different "retro" machines, IBM XT / AT, and newer PCs, and of course fondly remember seeing games like Wolfenstein 3D, Doom, and Quake for the first time.

Oh wow, that sounds really cool! The Descent 2 track reminds of a mix between wavetable and an Amiga! If that makes sense. It does sound a bit raw around the edges, but very unique and fitting for an older retro gaming PC. Heck that's one cool project! I do have the Mist FPGA computer, so I know a bit about FPGA, and I think it's awesome. Man, keep working on it, this is one cool idea. Imagine a SF2 compatible board, maybe a wavetable board itself? Do keep me in the loop please about this one :D

PhilsComputerLab Thanks. ;) Yep, the synth does have a certain charm to it atm. I know it has that slight crackle too, but that's just where the sample update is clashing with the MIDI Note On stuff (if the MIDI state machine is currently accessing the ROM to grab the sample header, the audio update has to wait a few extra clocks for that to finish, hence the clicks). That will all be fixable of course, and I'll be testing a proper Soundfont later. Ahh, if you have a MiST board, then I can definitely get it running on that. ;) I don't have one myself, but I designed a "breakout" board for the BeMicro CV boards before, and I have most of the design done for a board that's essentially more of a MiST "clone". I've been holding off on getting that made though, as I realised that I may have the new-new HDMI board done in a few months, and that will have a similar spec to the MiST, and have the SDRAM etc., but on a tiny board of only around 55mm by 45mm. I also didn't really want to do a "clone" that would seem like somebody else's work tbh. Then again, the MiST is essentially generic (as with many FPGA boards), and at least they have made the sources available for all the cores now. I must try the Archimedes one soon. :p I will keep you updated on the synth, and may be sending you a file to test on the MiST if / when I get SDRAM support working. I may be able to load the Soundfonts into SDRAM without needing a CPU core at all though, as I already wrote a FAT32 / SD card core. Oh, and yes - one of the goals of the synth project will be to get some proper PCBs made for a WaveTable add-on board. ;) I just figured out the basics for the SF2 format last night. It's actually quite a simple format once I understood what the docs was on about when it was talking about the "chunks" in the file. Really, the SF2 is just a collection of simple headers, and then some other entries that relate to the patch info, and then point back to the offsets for the samples etc. Shouldn't take too long to implement. Only issue is that the Flash chip on the DE1 dev board is only being used in the 8-bit wide mode, so I'll have to do two accesses for each of the 16-bit sample values, or just get the SDRAM loading working. I haven't added the velocity, pitch bend, aftertouch, ADSR, nor any effects yet, but just hearing some 22KHz or 44.1KHz samples will be a massive improvement. I haven't had any issues with hanging notes yet, as there's no real possibility that it can miss any MIDI data. It's not even buffering the data in a FIFO as such yet, it just runs so fast that it can always process the MIDI data before the next byte even arrives.

Well if you ever release this and need support / a review let me know :D FPGA is cool, though SOC board computers like the Raspberry Pi / Orange Pi are getting more powerful also. The community already has FluidSynth as well as Munt running on the P2 and Pi 3 through a MIDI in. Something to keep in mind. I'm also playing around with that, but it's more to learn about the Pi, how to program the GPIO pins and basically just to learn something new :D

That's the thing - the reward for me is doing as much as possible in raw logic rather than using a Pi etc. It's not quite like "real" hardware on a Rasp Pi or SOC, because it's running a lot of stuff in software of course, and there's usually that underlying OS to get in the way. I do enjoy messing with the Rasp Pi and other SOC boards, and I expect the latency is just fine for most things, but writing the FPGA code from scratch means that I'll have complete control over every aspect (for the first time for a MIDI project). It will also have ultra-low latency, and can still (eventually) use large sample sets with high sample rates / bit depths etc. without loading down nor requiring a CPU. It's also easier to hook up things like other classic synth chips - the very proto board I have the MIDI signal hooked up to atm also has a socket for a YM2612 chip, for example. ;) I do a little bit of C coding, but Verilog and raw logic is somehow easier for me, and it's cool just to be making "real" hardware once the FPGA is configured. :p I'm thinking about building a full MIDI synth with a keyboard action etc. It's quite hard to find the decent actions at a reasonable price though. I've just been looking for some, but I want the type that has the proper switch bars under the keys, instead of the horrible rubber pads used on my current keyboards. I had a Cheetah MK7 MIDI keyboard in the 90s, and it was brilliant. It did have a few keys missing though, and a fault with the EPROM meant that it only worked about 10% of the time you powered it up. It's still one of the best I've tried though.

I think so. In the VOGONS thread someone mentions a Virge.

Not something I've tested (yet). Thing is, I find UniVBE a bit annoying with registration and all of that. But yea, something to look into.

You answered your own question by ranking them in order :)

The S3 VIRGE DX Is an S3 64V+ update with a Little 3d function

If you wanted 3D it was OpenGL or nothing with NT4 or 2000, and none of these chipsets supported OpenGL. IIRC you could get some limited Direct3D or Glide support in 2000 but most driver support was poor for gaming class 3D chipsets - drivers focused on Win98 at that time. Most people doing OpenGL work in NT4/2000 were dropping serious cash for things like FireGL boards.

Actually, in 2000, you had pretty much all the Direct3D options you have in XP. But yeah, it was OpenGL or nothing in NT, mostly.

I don't see why not?

I love learning of such details, thank you for sharing! 🙂

Yes it turns out that some BIOS have this issue! You can update the BIOS to solve this. Seeing that the S3 cards also need a driver to unlock full speed, I prefer Cirrus Logic cards now, they are better all rounders. But if you have a S3 card with fixed BIOS and don't mind running the driver, they are still very good.

@@philscomputerlab Is the driver the univibe thing one can load to boost performance on some cards? I'm not familiar with it. Thanks.

Yes. The game is capped at 35 FPS, but the benchmark goes beyond that.

Those are the main cards I got :)

The Rage is a bit to "powerful" for this project IMO. I also don't have anything with PCI.

PhilsComputerLab : Fair enough, I'd still like too how a Rage XL performs on this platform since you can buy them brand new. Yes, they do come in PCI.

Or Matrix Millennium I 😋

Ati 3d Rage 1 and 2 too

S3 also had crap drivers (I had a Trio 64, followed by a S3 Virge). The S3 Virge replaced the Trio family and was one of the 1st graphics cards to support 3D acceleration, sadly it didn't have the grunt to run the features at a reasonable framerate, making the feature rather redundant (if you used the feature in most games, the framerate fell through the floor, although for the time it did look rather pretty!) 😁

Can't go wrong with S3 but checkout the VBE S3 driver as well as the speed up patch. Matrox has clearer VGA image, so if your games all work on Matrox, it's a good choice. I see no reason to use ATI over S3 or Matrox...

That might be something for another video.

They can often be slower, as I recall (still have one about) the S3 Trio64-V2 and S3 ViRGE (3D/DX) will typically provide the best bang for your buck in terms of 2D Acceleration. VESA Vibe 1.3b or above is necessary for DOS Gaming, but as long as you have > 8MB System RAM and you load it into High Mem you won't run into any 'Edge' Cases where you don't have enough Base Mem. Specifically I recall this affecting games like X-Wing, F1GP, etc. still in terms of Windows 3.x (Win32GS) or Windows 95a/b/c (DirectX 5); you will see the S3 absolutely dominating performance wise even against 3D Accelerator Cards unless Memory is an issue. As I recall the All-in-Wonder was ATIs Solution to this, with arguable results. Personally I'd generally recommend an S3 ViRGE DX/2 + 3DFX Banshee. Now don't get me wrong the ViRGE was certainly a decent Budget 3D Card, but 3DFX was by far the best option available at the time in terms of Performance. This would usually be closely followed by Matrox Mystique or Power VR, assuming games actually supported them... which was the biggest problem for most competitors to 3DFX in the Mid-Late 90s. The same would be true for the RIVA TNT and NV-1. It is better to have consistent performance and support (IMO) than the odd gem that was designed specifically for said Card that provides Great FPS. There also was very little stock (at the time) placed on Benchmarks as it was quite well known that Developers would optimise the hell out of getting good Scores on common benchmark programs to give the illusion they were more awesome than they really were. Not much has really changed really.

Leyvin (StudioRaven) Riva TNT and NV-1: You could not get any more different in graphics chips if you tried. The NV-1 used Quadratic rendering as opposed to the triangles that every other graphics chip used. It really was only applicable in full strength to Sega Saturn and Sega arcade board-derived games, and in fact, the only PC card to use that chip provided ports to use Saturn controllers with. While they did make an OpenGL compatible driver, it was really hampered trying to translate between the two. The Riva TNT threw that out and went with the same triangle-based rendering all the other cards used (and still use), and was actually pretty well supported in the day before DirectX was released as middleware to both make a lot of that moot and put a nail in the Voodoo's coffin as Glide was not compatible with DirectX. With the increasing focus on Windows for gaming, specialized APIs like Glide became less important, but 3Dfx was slow to move on from Glide and suffered for it in the long run. Really, the window of Glide dominance was actually somewhat small in the grand scheme of things.

Leyvin (StudioRaven) the Tseng ET6300 was rebranded as the ATi Rage 128 after ATi bought them. It may surprise you.

JohnnyNismo I think both of you are forgetting the Riva 128 and confusing it with the Riva TNT.

If you want one, yes, it's suppose to be the normal price for those cards. People give them usually (that's how I got my cirus logic to). They are not worth 45$. If you pay that price, you are juste helping the prices to go higher and helping greedy people that scavenge the parts for next to nothing.

OK, thx for the reply. I didn't get it. I do already own a S3 Virge / DX 4MB just in case I want to build something retro. :)

i have found some socket 3 and socket 7 boards on ebay for reasonable prices, but many are really expensive like $120 all the way to $500, sound card and video cards of that era are getting pricey too, i guess out of greed or the parts getting rarer and rarer from recycling.

Yes it's part of the BIOS. I put a link down below, people have modified the BIOS as well as written a small driver to fix it.

For DOS, which is mostly 320x200, shouldn't matter much. Windows is a different story, here you could see differences. Matrox are the way to go for this...

Thanks for the fast reply, Phil. In my experience, DOS is also problematic. I have video cards that have exceptional 2D quality in 720 x 400 / 640 x 480 (Voodoo Banshee, Voodoo 3, Elsa GeForce 2 MX / GTS), others that were horrible but fixed them by removing the RFI filter (Asus V7700 GeForce 2 GTS, Inno3D GeForce 3 Ti 200), and some that seem to be unfixable (basically, all of my Riva TNT / TNT2 cards, Medion GeForce 4 Ti 4200 AGP 8x, etc). Some are so blurry that even the text on the DOS command prompt is a pain to read.

@@ruxandy I see. I haven't noticed much, but admit that I haven't done much DOS stuff with Nvidia cards. I have old ISA cards and here I can see sort of stripes on LCD monitors.

@Dalle Smalhals never said that these were all the GPUs I owned. I also have ISA, VLB and PCI video cards from the DOS era. The cards I listed above were just as an example that even newer cards can suffer from poor image quality.