Matrox: Graphics for Professionals(abortretry.fail) |
Matrox: Graphics for Professionals(abortretry.fail) |
Before I was a Linux type I was a Windows expert, and when I went freelance
in 1996, I built myself a fairly serious NT 4 rig: Pentium II 450, 192MB of
RAM (I think), and dual Matrox Millennium cards so I could drive dual 14-15
inch CRT (glass) screens as a single big display.
I think I kept those cards, as a pair, in my next 3 or 4 PCs – I know I was
still using them about a decade later, because when I switched from SUSE to
Ubuntu in 2004, Ubuntu's very screen-config tools couldn't configure 2 GPUs
and so I had to keep SUSE around for several years in dual boot. Every time
I made some screen config change or whatever, I booted into SUSE, set up my
Matrox cards and then copied `/etc/X11/xorg.conf` into my Ubuntu partition,
then I could reboot into Ubuntu and use them.
Ubuntu finally caught up around release 6.06 or something.
I didn't stay on two 14" displays for long... then two 15", then a 17" plus
a 15", then the widescreen luxury of dual 17" Idek Iiyama displays. :-)
I stopped playing video games much decades ago, so I never cared about what
3D performance was like. Nor did 99% of my consultancy clients. As 3D cards
gradually became the norm, I used to install the free Microsoft-advertising
version of the Computer Artworks "Organic Art" screensaver:
https://cuttlefish.org/archive/earlysites/artworks/organic-art/index.html
... just so the 3D card had something to do. It looked impressive, and when
Windows 2000 was new, watching how smoothly the screensaver ran was quite a
treat. When the app first came out it was like a slideshow.
What was good news for me was when LCD flat screens started to appear: lots
of people and companies started throwing out even high-end high-quality CRT
screens, and big ones too. For years I ran a pair of 21" Trinitron displays
which cost a fortune new. I filled my garage with good quality CRTs just as
spares. I had a small and rather sickly cat and she vomited a lot. There is
a nice big warm flat space on the back of a big CRT ideal for a cat to take
naps on. She killed dozens of monitors, by vomiting into them, but I had so
many spares that for several years I simply didn't care. I sent them to the
recycling and grabbed another one from the spares.Just to get the timeline right -- the PII 450 wasn't released until 98, so I'm assuming this is about when you mean. I always thought it would be awesome to get a Matrox card but settled for the ATI cards (I can't remember when I got my first dual display but I remember that Matrox made quad head cards pretty early.
> For years I ran a pair of 21" Trinitron displays which cost a fortune new.
These things were so heavy.
I’ve used a big monitor setup ever since (part of my sign-up incentive for my current job was a 30” monitor) and currently have two 16:10 30” monitors.
Ordinary relatively cheap consumer Matrox cards could be paired up. I couldn't afford a fancy dual-output card, but I could afford 2 second-hand Matrox Millennium cards, and that did the job fine. At that time, Matrox was about the only brand of (parallel PCI) graphics card where the drivers could handle multiple cards. You couldn't do it with ATI or S3 or any of the other commodity GPUs back then, but any old Matrox with any other of around the same generation worked. They could have dissimilar amounts of memory and things and it still worked.
About 20 years later, I was working at SUSE in Prague, and we had lots of old nVidia cards available in the office spares pile, and lots of spare LCDs around the 22" - 24" size range.
Two screens was easy: most nVidia pro-range cards can drive 2 outputs.
I had the desk space and available screens, so I decided to try 3.
Then the nightmare began.
Nouveau doesn't care: if it's nVidia, it'll drive it.
But nVidia's own drivers are tied to specific GPU generations, but those generations are not marked on the card or in the easily-accessible specs anywhere. It's mostly a case of installing the card, booting Linux, seeing what the driver describes the GPU family generation as, and then Googling to try to find which of the random pile of cards to hand is a matching GPU family and generation.
I never got 3 screens working together adequately, and the combinations crashed my OS often.
For emergencies I kept a Windows partition and its behaviour with mismatched cards is quite funny in an annoying way.
1. Boot
2. Identify first card
3. Work out matching driver for card 1, download it, install it
4. Reboot
5. Find second card
6. Work out matching driver for card 2, download it, install it
7. Card 1 now does not work. So...
8. Work out matching driver for card 1, download it, install it
9. Reboot
10. Second card now does not work.
11. Work out matching driver for card 2, download it, install it...
It would sit there taking 10min or so per loop until you pulled the plug.
With Linux, it's just a case of:
Human IDs GPU; manually installs driver; sees if both cards work.
The answer is, generally, "no", so:
Swear, remove one card, substitute, try again.
Eventually I went to the IT department who gave me an elderly, huge, AMD card. 2 full length slots, 2 cooling fans, and a decade old so rubbish performance by 20-teens standards. But, 4 display outputs, and use any combination you want -- and native FOSS drivers, so no need to manually install anything.
Joy, peace, and calm descended.
Then a couple of months later, COVID... and working off a little Dell laptop with only an Intel onboard GPU and one external screen. I did in the end get 2 external screens and the built in screen working, but only using an X11 desktop that didn't do much fancy 3D rendering (Xfce) because it didn't have enough VRAM for triple-head and compositing.
At the time I worked in broadcasting and we had a lot of respect for Matrox, however, I can't recall exactly how they lost relevance for us, in the broadcast-adjacent market, when we would still be running 3D Studio in MS-DOS. Windows 95 came along with OpenGL, and, if I remember, Matrox and OpenGL wasn't the combination you needed.
For the office PCs it was embedded graphics that ended them in my neck of the woods, with graphics being whatever Dell/AST/Compaq supplied.
first real Matrox OpenGL driver shipped with G400 in 2000
In the broadcast world they're still very common. X.mio/DSX cards are widely used for professional SDI I/O, video servers and channel-in-a-box systems, and Matrox also has dedicated M264 hardware accelerators for H.264/AVC encoding and decoding.
The current flagship X.mio5 family is basically high-density professional video I/O: 12G-SDI variants for UHD workflows, plus newer 25GbE versions for SMPTE ST 2110/IP environments. Very different market from the old Millennium/G400 days, but Matrox is definitely still around
I was just trying to remember what was available -- that was an incredible machine for its time, I'm not surprised you ran into issues getting it configured.
Back then it felt a lot more like certain configs were just never going to work even if they were "compatible". Today it seems like with sufficient effort you can make just about anything work.
No matter! A transistor, a couple of diodes, and half a dozen or so resistors soldered into a DE15 plug shell made the "sync plunger" that cured all your woes.
I probably still have a couple lying around. I could still just about draw the circuit from memory, not much good in these HDMI days.
I got it working, but the one I had was pretty limited - it did a specific resolution at 60 Hz (I may have the details wrong - it was a long time ago). I never ended up using it, as I had a newer 17" DEC Trinitron from an old Alpha 3000/300LX that I also got super cheap, and it was a much better screen.
That was an awesome card for the time, and an incredibly stable analog signal, which Matrox was known for - other cards at the time would get fuzzy, shimmer, or have ringing / ghosting.
At the time, for good support you needed to *buy* an X11 server for accelerated graphics support. I forget which one I paid money for, but within a few years it wasn't necessary as open source had caught up.
Later we got to demo early builds on pre-release TNT2 cards and it caused a panic for us because our game had always been graphics limited, and the graphics pipeline was built to be very steady. Incredibly smart programmers wrote our renderer from scratch in those days. Suddenly the frame rate was really fast, and choppy, because the hit from physics and AI between frames caused stutters. After that NVIDIA dominated.
The pre 3DFX era was kinda interesting. The first hardware accellerated game I saw was VQuake on the Rendition Verite 1000 or whatever the branding was. It blew me away. When the Voodoo1 came out I was shocked by the framerate, but thought it looked ugly compared to the Rendition version. Technically the Verite had worse texture filtering, but I was used to the aliasing and thought the Voodoo looked overly smoothed.
Back when 2d acceleration mattered Matrox was king. And you could always trust it'd drive even the most picky CRT at whatever refresh rate.
You also could get the capture card with the Matrox Millenium and turn a Win95 pc into a legit video editing station, all be it hilariously limited vs today. I worked with a non profit briefly that did stop motion animation production with this setup.
I was the first with a TNT2 at the big LAN party group in my area. The Voodoo2 SLI setups were still faster, but with the TNT2 you could do 16 bit color with a tweaked color map for id engine games, and it looked absolutely killer.
That's pretty remarkable, isn't it? How many tech companies that had a period of being a household name have lasted 50 years and are still majority-owned by all or some of the founders?
For a time in the 1980s, I was writing drivers for graphics boards, and later doing 3D animation software. I tried a lot of graphics boards from now-forgotten vendors. Texas Instruments had the first PC RGB monitor and display that could display white text without color smear. The Wang Professional Image Computer was incompatible with everything else in PCs, with a different card form factor, a different monitor interface, and a different memory/device address layout that allowed 720KB of memory instead of 640KB. Plus their own version of DOS. Fujitsu lent me a prototype Fujitsu Sapphire, which never worked right. When I went to return it to their location in Sunnyvale, they'd moved out, shutting down their 3D graphics business. Dynamic Pictures had a decent 3D board, the Oxygen 202, that worked with NT 3.51. It took a while to get that working, but eventually it worked and I got 3D work done.
Th code is still online, in full early 2000s web styling. https://64kib.com/old_foft_site/
> On the 17th of December in 2004, the company announced the QIP LP PCIe where the QIP meant Quad Information Display. This was the industry’s first PCIe x16 graphics card, and it was a derivative of the Parhelia. [...] The card could drive four displays with a maximum resolution of 1600 by 1200 per display,
I think we had 4 of these in one system and it was the only card on the market that allowed for this. We ran this on Linux and getting the bios set up correctly to support this was a nightmare. My memory is fuzzy though
Matrox était connu pour faire d’excellentes cartes pour le médical. Malheureusement, 3dfx était trop en avance pour les jeux.
no textures, alpha blending, transparency, bilinear filtering
>Mystique ... transparency lookup table
no alpha blending, transparency, bilinear filtering. "Transparency Lookup Table" was a corpo speak code for color keying and a checkerboard mask that would stipple half the texture on top of whatever was there before
> G200. These cards supported ... OpenGL ... use of an OpenGL to Direct3D wrapper which hurt performance
Hurt compatibility first, performance second. Proper OpenGL driver shipped long after 2000 when G200 was already obsolete
>G200 ... performance was good
Performance was good for last place in benchmarks https://web.archive.org/web/20240907223521/https://www.anand...
> it still couldn’t beat the Voodoo 2, but it was fairly competitive with the RIVA and S3 Savage 3D.
How could it be competitive with cards half its price? or cards 2x as fast at same price point?
MATROX G100 8MB $99 April 1998
MATROX MILLENNIUM G200 8MB $230 August 1998 https://assets.hardwarezone.com/2009/reviews/video/g200/g200.html
optional DVD Upgrade Module (Hardware MPEG2 decoder) for G100/G200 $79 April 1998
STB Velocity 128 4MB $129 June 1998
S3 Savage 3D ~$150
Creative Labs Graphics Blaster TNT 16MB $169 September 1998, $125 November 1998
Diamond Viper V550 16MB $200 September 1998
Creative Labs 3D Blaster VooDoo2 8MB $229 June 1998 and dropping fast after TNT release
Diamond Monster 3D II 8MB $249 June 1998 and dropping fast after TNT release
I think the remnants are here now[0]:
Even though the card was expensive, it didn't have Direct3D acceleration like my classmate's Riva TNT (or 2?). So gaming wasn't impressive at all. At the time we were playing Need for Speed 2, I remember the image wasn't noticeably different than a 3DFX Voodoo. However, when Need for Speed 3 Hot Pursuit came out, Riva TNT could do realistic shading and reflection-like environment mapping on cars which was indeed a leap.
One other thing I remember is that the graphics card had memory slots on it, like a socket for a chip. I think it was called 'WRAM', and its price was much more expensive than the card itself.
maybe because they had limited support for many features?
>Millennium
no textures, alpha blending, transparency, bilinear filtering
>G400
slow but ok, but its predecessor G200 only got OpenGL driver 2 years later with release of G400
Gimmicks that are even today way overused. The world would be better off without them.
Note that I am not into 3d games, and while someday I mean to learn CAD that day is far in the future. If I was into either of those things the features you name would be important. However since I'm not those things are useless gimmicks - YMMV.
And later I got one as pretty much the most solid Linux graphics card, which even was capable of going dual screen. And while I don't miss XFree86 modeline configs, it was nice when there wasn't just a GPU triopoly.
This feels intuitively wrong to me now and I doubt that's how it worked. In any case I dutifully copies the Matrox provided X11 driver to OpenBSD and enjoyed multi-monitor support.