Is this a good graphics card for a retro win95/DOS build?

[Eudimorphodon]

Now, what's this nonsense you're on about "no more than 10-20%"?

LOOK AT THE GIGANTIC SPREADSHEET OF BENCHMARK RESULTS LINKED ON PHIL’S WEBSITE, and start with the Pentium-and-up systems with PCI and AGP video cards BECAUSE THAT IS WHAT THIS THREAD IS ABOUT.

Phil's Ultimate VGA Benchmark Database Project

These are the DOS benchmarks I use for all my videos and articles! Just unpack them on your DOS Retro Gaming PC and type DOSBENCH.BAT DOS Benchmark Pack The menu guides you through the options....

www.philscomputerlab.com

Yes, you’re going to find a lot more variation between video chips this old, because this is exactly where you *will* find, for instance, VGA chips that don’t support linear framebuffers or generate a shed load of wait states. This is not the generation of computers we were talking about here, AT ALL. Chucking this up here as if it’s relevant to this discussion is revealing about your level of good faith.

If you look absolutely hard enough, sure, you can find instances on that spreadsheet that exceed my 10-20% ballpark, like the guy that tested everything from a Geforce4 to a 1MB S3 Trio in a 1400mhz Celeron, but even in that ridiculous situation the spread is only around 40%. That is entirety explained by bus speed/VRAM differences.

I'm not going to go digging out 30 year old datasheets to prove it to you, because the data speaks for itself.

I don’t give a rat’s rear about datasheets. I want you to demonstrate the existence of, hell, I’ll take a single example, of a DOS game that uses these 2D acceleration features. Because they don’t just magically work on any code touching the VGA card, you need a driver that knows about them. VESA BIOSes don’t know about them, CASTLE WOLFENSTEIN 3D doesn’t know about them (I mean, come on, dude), and, critically, THEY ARE NOT STANDARDIZED between video cards from different manufacturers nor between models. You are just pulling stuff out of your ass unless you can prove those features were actually used.

The source code for Wolf3D, DOOM, and Quake are all publicly available. Start digging. I’ll wait.

 

[resman]

You seem to be unfamiliar with the concept of DMA. Video cards don't need to access host resources to do DMA transfers. The example I'm referring to is Blitter, which can do complex block copies of pixel data in memory, without intervention of the host CPU. The ET4000/W32i has a hardware blitter engine and some hardware sprite support, which makes it a very fast ISA video card.

You're doubling down on your ignorance. Know when to stop. Non of those benchmarks would have taken advantage of any of those hardware features. Eudimorphodon pretty much laid it out for you.

 

[Plasma]

So after a little digging around on the 'net I'm toying with getting an nVidia RIVA 128 for this build(when I eventually have money to finish it). I know their graphics support was terrible, but this is supposed to be one of the first AGP cards.

If someone can recommend a better card that can be had for $20-30 I'm all ears. Performance does not have to be top dog, compatibility is slightly more important. But mainly I am looking for that mid-90s timeframe. Once you hit 1998+ I have a better system on hand.

I haven't used the 128, but I had a TNT in the Windows 9x days and it was great.

 

[resman]

So after a little digging around on the 'net I'm toying with getting an nVidia RIVA 128 for this build(when I eventually have money to finish it). I know their graphics support was terrible, but this is supposed to be one of the first AGP cards.

If someone can recommend a better card that can be had for $20-30 I'm all ears. Performance does not have to be top dog, compatibility is slightly more important. But mainly I am looking for that mid-90s timeframe. Once you hit 1998+ I have a better system on hand.

RIVA 128 would be an excellent choice (but I'm biased). I don't know why you think it would have terrible graphics support. It had solid VESA compatibility for the time and was one of the fastest DOS VGAs. Very competitive Windows 2D and Direct3D performance with later drivers.

 

[hunterjwizzard]

RIVA 128 would be an excellent choice (but I'm biased). I don't know why you think it would have terrible graphics support. It had solid VESA compatibility for the time and was one of the fastest DOS VGAs. Very competitive Windows 2D and Direct3D performance with later drivers.

I'll admit my research wasn't thorough, there aren't a lot of (non-video) reviews out there for 28 year old graphics cards out there, but a couple of retrospective articles I did find mentioned the RIVA 128 as not being as good as the PCI Voodoo1s. Still, if its a capable DOS card for that price I think its exactly what I'm looking for. I never planned to spend very much money on this build, all-in it'll be about $48 +parts I found in my closed :D

 

[Eudimorphodon]

Non of those benchmarks would have taken advantage of any of those hardware features.

I'm starting the countdown to when we get some vague handwaving about how VGA *DOES* have a "blitter" (which is a not-really-accurate description of some of the tricks you can do with the EGA/VGA page latches) and how anything that uses that counts as "accelerated" and somehow a card that has a Windows accelerator chip (I mean, that's what the "W" is for in "ET4000/W32i") magically accelerates everything because it's all connected...

Speaking of EGA/VGA latches, I'm going to throw out a single bone to chew on: I do remember reading an allegation that DOOM could suffer unusually lousy performance on some VGA cards that otherwise perform well because the chipsets in those cards had slow I/O to the latches; this wouldn't matter on software that used the normal VGA mode 13h for 256 colors, but software using "Mode X" techniques (which DOS DOOM apparently did; it doesn't run in mode 13h, it runs in a 320x200 "Mode Y" to enable page flipping) suffer. DOS games didn't use acceleration, but it wasn't *that* rare for them to use Mode X. On a sufficiently fast machine this is probably going to be completely lost in the noise but, heck, remember that reference to the 1.4Ghz Celeron tested with a bunch of different video cards? The slowest result for DOOM on its list was actually the GeforceMX440, slower than an S3 Virge and only half as fast as the fastest card which was a... GeForce2. This result is mostly anomalous compared to the rest of its test results, so... sure, there you go, maybe if you're really super unlucky when trying to pick out video cards for your late-90's DOS gamer build you might pick a card that kinda sucks compared to theoretical on Mode X games. But lets be frank, it still probably runs them way faster than you need them to run.

Another little nugget that I'm vaguely recalling is by the mid-90's there were some cards that started having somewhat imperfect EGA mode compatibility? (I kind of want to point a finger at S3 cards again for a minor quirk I actually ran into with a Pentium 90-ish machine with an S3-something in it, either a Virge or Trio?) This might be a ghost lurking in the background if you wanted to build the "perfect" PCI DOS gaming machine and ensure it's not going to have problems running older titles like Commander Keen, et al, but.. this isn't a "performance" issue, per se..

 

[resman]

...
The slowest result for DOOM on its list was actually the GeforceMX440, slower than an S3 Virge and only half as fast as the fastest card which was a... GeForce2. This result is mostly anomalous compared to the rest of its test results, so... sure, there you go, maybe if you're really super unlucky when trying to pick out video cards for your late-90's DOS gamer build you might pick a card that kinda sucks compared to theoretical on Mode X games. But lets be frank, it still probably runs them way faster than you need them to run.
...

Yeah, that would be odd. Sans NV1, all the NVIDIA chips should have the same VGA core. But, you can't rule out some botched bus interface code that wasn't caught until later. VGA performance wouldn't have been a priority by then.

 

[Eudimorphodon]

RIVA 128 would be an excellent choice (but I'm biased). I don't know why you think it would have terrible graphics support. It had solid VESA compatibility for the time and was one of the fastest DOS VGAs. Very competitive Windows 2D and Direct3D performance with later drivers.

I'll admit my research wasn't thorough, there aren't a lot of (non-video) reviews out there for 28 year old graphics cards out there, but a couple of retrospective articles I did find mentioned the RIVA 128 as not being as good as the PCI Voodoo1s.

So far as I'm concerned the *only* thing a Voodoo has over a Riva128 is GLIDE support. If you're willing to live without that and just want a solid Windows 9x card the Riva128 is probably the best card from that era; it was especially great you came to it from a hideous joke like an S3 Virge or Trio (my experience with it). The performance is similar to and it's technically more advanced than a Voodoo, in that it can actually use AGP DMA and supports *huge* textures (2048x2048!, vs. Voodoo's 256x256) for a 1997-vintage card.

My only "concern" with it is if you're going with a Pentium II or higher CPU you might find yourself wishing you'd held out for a TNT/TNT2-level card to support higher resolutions/more VRAM, but, hey, if you want that just throw that ATI Rage Maxx in. Gotta be honest, the card you had already *is* probably better than the Riva128 unless you manage to find some weird edge case where those bad rumors about problems with the dual-core-ed-ness of it bite you.

Yeah, that would be odd. Sans NV1, all the NVIDIA chips should have the same VGA core. But, you can't rule out some botched bus interface code that wasn't caught until later. VGA performance wouldn't have been a priority by then.

It could just be a quirk of that particular card/PC combination, or, yeah, maybe they just stopped really testing obscure features like that and didn't notice a regression that wasn't likely to matter in the real world anyway. DOS was pretty dead by the time the 440MX came out and considering a card like that was meant to be paired with a 2-ish-ghz CPU a game from the mid-1990's was going to run "fast enough" regardless of if the card's VGA backwards compatibility had been left in the corner to grow spiderwebs.

This is just based on a quick eyeballing, not a scientific assessment (I'm frankly getting sick of looking at that spreadsheet), but I would hazard the observation that the Doom FPS test seems like it may be the one that has the most "outliers". (IE, you test X-many VGA cards in the same system, which for the most part cluster fairly closely, but one of them has an anomalously bad result not only compared to the other cards but compared to how well it did on the other tests.) So, sure, I'm willing to grant that maybe Mode X is a possible achilles heel for DOS performance? Of course, I'm also going to state that my faith in the reliabilty/reproducability of these results is tempered by a pretty large cube of rock salt. And, again, let's be serious: a Pentium II will run any DOS game just fine. Full stop.

 

[hunterjwizzard]

My only "concern" with it is if you're going with a Pentium II or higher CPU you might find yourself wishing you'd held out for a TNT/TNT2-level card to support higher resolutions/more VRAM, but, hey, if you want that just throw that ATI Rage Maxx in. Gotta be honest, the card you had already *is* probably better than the Riva128 unless you manage to find some weird edge case where those bad rumors about problems with the dual-core-ed-ness of it bite you.

Except that I am specifically building a Windows 95 box, I already have a nice win98 system I may put the Fury MAXX in for giggles.

TNT2s look to be not much more expensive(I see one right now I could have for $20), but is it going to have the same compatibility with 95/DOS?

 

[Eudimorphodon]

TNT2s look to be not much more expensive(I see one right now I could have for $20), but is it going to have the same compatibility with 95/DOS?

So far as I'm aware the TNT2 had Windows 95 driver support; a google seems to confirm that. And regarding DOS, per @resman's comment apparently all Nvidia cards since the Riva128 have essentially implemented the same VGA-backwards-compatibility core. It looks like the only Nvidia chip that had *any* DOS programs that could specifically leverage acceleration was the original (and apparently pretty disasterous) NV1 (which offered a DOS API roughly comparable to GLIDE), and you can almost count the software programs that used it on one hand.

Seriously, get it out of your head that DOS is likely to matter much here unless you specifically seek out one of the accelerators listed in that link I just re-pasted above *or* a Voodoo card to run software that specifically supports them. If you don't buy one of those cards *specifically* to run the accelerated versions of those particular games then you're going to be using software rendering, and then all that matters is your card has "decent" VGA backwards compatibility and performance, which is most late-90's cards.

... I will probably regret asking this, but is there really a reason to set up a Windows 95 computer specifically when you already have a Windows 98 one? I mean... Windows 98 is Windows 95 but slightly less broken, where's the list of games that actually, for reals, only run on Windows 95 and not later versions?

 

[Agent Orange]

Except that I am specifically building a Windows 95 box, I already have a nice win98 system I may put the Fury MAXX in for giggles.

TNT2s look to be not much more expensive(I see one right now I could have for $20), but is it going to have the same compatibility with 95/DOS?

Hunter:
If you haven't already been there, I would suggest that you visit VOGON's as they have some really good tips for a W95 system.

 

[hunterjwizzard]

... I will probably regret asking this, but is there really a reason to set up a Windows 95 computer specifically when you already have a Windows 98 one?

The short answer is "why not?"

The long answer is because I bought this PII motherboard by accident and I like building computers. There's a longer version of this story but it isn't very interesting.

 

[Eudimorphodon]

I mean, hey, if you want to do it "just 'cause", well, go for it. That's pretty much the only reason to do most of what we talk about on these forums.

Again, I was just wondering if there was an actual technical reason why you'd "need" to have a Windows 95 computer to provide coverage for some particular class of software. I was under the impression that if Windows 9x is really a thing you want in your life a Windows 98se machine would probably have you covered.

(The closest thing I have to a working machine from this era is a (I forget, 800mhz?) Cyrix Mini-ITX box that I set up as a DOS-only computer just for laughs; it's essentially pretending to be a very fast 486. I've *considered* installing Windows 9x on it because doing so would potentially actually *improve* its ability to run some DOS games because it could use Win9x virtualized drivers instead of the limited and pretty cruddy options that bare DOS offers on that hardware, but quarter-century-old scars left over with dealing with Windows 9x BS have so far prevented me from doing so... yeah, if I *were* personally going to do it I'd only want to do it once.)

 

[Agent Orange]

Personally, I love building PC's. In the the case of W95, I put plenty together and on the last one, I wondered why. So, that was converted to W98SE and it stands today.

 

[hunterjwizzard]

Again, I was just wondering if there was an actual technical reason why you'd "need" to have a Windows 95 computer to provide coverage for some particular class of software.

I did find a couple games while moving that specify "windows 95 only" but I suspect they were released in '95/96 and mean "won't work in 3.1 or DOS". I haven't tried them yet.

The longer version of this story is the sound card in my windows 98 PC lacks backwards compatibility for DOS stuff. I can't really afford the vintage sound card I want(Gravis Ultra Sound with the removable memory modules), but while absently shopping one day I found this Slot I PII motherboard with integrated Yamaha OPL sound. I ended up winning the auction for $1 + $17 shipping. Then, a few months later(when I made this thread) I found a couple of old PII processes in a forgotten shoebox. It seemed like fate wanted me to build this machine.

Obviously I could run windows 98 or even pure DOS, but I never got to experience windows 95.

 

[Eudimorphodon]

I found this Slot I PII motherboard with integrated Yamaha OPL sound

That's a lucky find; things as new as Pentium IIs with directly-DOS-compatible sound hardware build in are pretty rare. I mean, with chipsets up to the 440BX at least you *can* still get away with throwing a real ISA Soundblaster in there if you've got the slots, but that has its own problems. That said, what is the specific chip? Is it *actually* hardware compatible with naked DOS, or is it one of those cards that provides good soundblaster emulation *within Windows*?

The Cyrix motherboard I have has *partially* DOS compatible sound. The chipset has a built-in Soundblaster Pro compatible DAC that you can use just by setting a BLASTER variable appropriately, but it doesn't have an OPL2/3 synth chip for music; there's an emulation driver that "works ok" but takes up 40K of lower or UMB memory, which is a PITA. (And of course, being software, doesn't sound quite right.) Running Windows on it and launching DOS programs from within it would potentially let me use better PCI soundcards that have real synth hardware (and free up more memory), I have a few of those lying around that might work, but... bleah.

I never got to experience windows 95.

That's not necessarily a bad thing. It's 98 but doesn't work as well. (And, really, if it's 95c with the Internet Explorer-enhanced UI it's almost impossible to tell from 98 anyway.)

 

[Unknown_K]

I purchased W95 the week it came out, and it wasn't bad (this was before IE was a thing). From what I remember a fast W95a system was Pentium 1 based.

Anyway Windows 95 gaming was mostly 2D anyway for things like Command and Conquer etc. Quake (original was software only) came out in 1996 and was DOS based with drivers for 3dfx voodoo 1 or Rendition Verite card that came out later.

My most used W95 gaming system is a P233MMX with a Riva 128 PCI card and a Voodoo 1.

 

[hunterjwizzard]

That said, what is the specific chip?

It appears to be a completely legit Yamaha OPL3.

One chip says "Yamaha OPL YMF715-S". There is also a "Yamaha General Midi YMF721-S".

 

[Svenska]

Obviously I could run windows 98 or even pure DOS, but I never got to experience windows 95.

There are some very relevant differences between Windows 95 / 95a and 95b / 95c in terms of hardware support. I do not recommend the earlier versions on any hardware past mid-1995, because their driver support is mediocre at best. However, they do perform better on lower-end systems (such as an ISA-only, mid-range 486 with 8 MB RAM), and require much less disk space. Only Windows 95b / 95c support FAT32.

I've used Windows 95b on my Pentium 133 until the mid-2000s (far longer than it had any right to be), and the IE4 shell update immensely improved compatibility with later software. At that point, Windows 95 becomes basically a stripped down Windows 98 with less hardware support and slightly lower system requirements. Do not install the shell upgrade on systems with less than 32 MB RAM, you will regret it.

I do not recommend Windows 95 for any system with USB ports or AGP slots. Windows 95 / 95a versions do not support USB at all, and the support in later versions is borderline useless (it's probably good enough for keyboards, mice and joysticks, but forget about mass storage). Likewise, AGP support only barely exists in Windows 95b / 95c. Stick with PCI cards.

 

[hunterjwizzard]

I do not recommend Windows 95 for any system with USB ports or AGP slots.

But the Intel AL440LX is supposed to be famously the first-ever motherboard with an AGP slot !!