Ltning

@ltning@weirdr.net

He/him. This is one of my alter egos in the retro world. Read about this instance on front page. My goal is to be able to post here from a 286* running DOS. Might be a while..

and enthusiast with a craving for retro (mostly PC) hardware. Four kids and a wonderful patchwork family.

*Speaking of 286es: http://floppy.museum/

29 following, 77 followers

4 ★ 3 ↺
/brúno antúnɛs/ boosted

[?]Ltning »
@ltning@weirdr.net

Welcome to my mini ISA VGA shootout!
TL;DR: ISA Matrox cards are really, really slow in DOS.

I recently built an original Pentium 60MHz system, built on an ECS motherboard. Around the same time I received a "mystery" VGA card: A Matrox MGA Impression ISA card. And since most of my builds are "open builds" and therefore easily accessible, that machine got the pleasure of becoming the test bench for the Matrox.

As already revealed, the Matrox performs atrociously bad. So bad, in fact, that I had to test a couple other ISA cards to make sure it wasn't a system issue. I used my go-to benchmarking tool from Phil's DOS Benchmark Pack. I really don't want to experience Doom with this card..

And without further ado, the contestants and their results in this spur-of-the-moment benchmark run:
- Baseline: A 32-bit PCI S3 Virge/DX based card with 4MB RAM: A perfectly workable 48.2
- The low-end Trident TVGA9000C with 512KB RAM (this is a real garbage card): A pretty shitty 14.2
- The mid-range Cirrus Logic CL-GD-5422 with 1MB RAM (this is a decent card, know for compatibility but not necessarily speed): A barely bearable 24.7
- And finally, the "star" of the show, the Matrox: A whopping 10.9!

I said it was atrocious, didn't I? But hey, I'm gonna use this one with anyway, so who cares about DOS performance, right? ;)


Montage: Close-up of the S3 card installed in the system, next to picture of the 3DBench result

Alt...Montage: Close-up of the S3 card installed in the system, next to picture of the 3DBench result

Montage: Picture of the Trident (a small ISA card) next to picture of the 3DBench result

Alt...Montage: Picture of the Trident (a small ISA card) next to picture of the 3DBench result

Montage: Picture of the Cirrus Logic (small ISA card) next to picture of the 3DBench result

Alt...Montage: Picture of the Cirrus Logic (small ISA card) next to picture of the 3DBench result

Montage: Picture of the Matrox (a very large full-length ISA card) next to picture of the 3DBench result

Alt...Montage: Picture of the Matrox (a very large full-length ISA card) next to picture of the 3DBench result

    ...

    [?]derSammler »
    @derSammler@oldbytes.space

    @ltning The Trident is not a garbage card. It's just that people can not do benchmarks correctly or do them in a way that creates the result they have already in mind anyway.

    First mistake: testing ISA graphics cards in a PCI system - which seems to be a very common mistake these days. On a PCI system, the ISA bus is bridged. You never get really good speed from ISA graphics cards in a bridged bus.

    Second, the Trident cards were aimed for 286/386 and lower-spec 486 systems. When used in these, you won't see much difference between a TVGA9000 and a GD-5422.

    Not long ago I tested a Tseng Labs ET3000 and a Realtek VGA card, the latter according to badly done benchmarks is one of the slowest VGA cards. Guess what? In a 386DX-33, they both performed almost exactly the same.

    But yeah, put it into a Pentium system and it becomes crap. But that's not the card's fault...

      ...
      0 ★ 0 ↺

      [?]Ltning »
      @ltning@weirdr.net

      Hey there :) Trust me, I know testing ISA cards in PCI machines is "wrong". And that there are many reasons for this, but the bridges you mention are not, in themselves, the problem. I'd like to say "not all bridges are created alike", and also that the direct cpu-to-ISA thing has not been true since the 286 era. From anything 386 onwards there has always been a bridge of some kind involved. So yes, it's "wrong", but only in the sense that "there's a PCI bus here, wtf am I punishing myself with an ISA VGA?!?".

      ISA performance can be "as good as it gets" on a PCI machine, and most BIOSes will allow you to configure it to a point that it can easily saturate the bus and even break stuff - just like in the good old days. Wait states and bus speed tuning being the most important things.

      I also know that Tridents are not all garbage, but it is fair to say that the 9000-series (which is essentially a 8900C with some additional integration) isn't going to win any performance prizes no matter what machine you put it in. The 8900D on the other hand is quite impressive in DOS, keeping up with all but a small handful of much more expensive cards. It is, of course, useless for a GUI system since it has no acceleration functions, but for plain VGA it's pretty good.

      All that said, I have not made any attempt at tuning for speed in these tests, and as I'm sure came across in my post - this is a highly un-scientific test that is only meant to gauge the relative difference between those cards (and with room for failure even at that). Fact of the matter, and what I wanted to confirm, is that the Matrox is unbelievably slow; I'm fairly sure the original VGA implementation with discrete chips (rather than a "VGA chip") would've performed better. Hell, even UniVBE warns me when configuring it that the card is unbearably slow in DOS, and doesn't even expose more than 1MB in VGA mode, so I should not expect much from it. Well, I guess I confirmed that, at least. :D

      The reason is, of course, that the Matrox is made for GUI applications. And in those, compared to its peers, it absolutely shines according to reviews at the time. I'm looking forward to testing it in a GUI environment, and I'll surely post about that somewhere as well. And I'll use more period-correct hardware, I promise! :)

        History

        about this site - powered by snac