VOGONS


First post, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie

Hi guys. Today I finally found some time do test all the PCI video cards I own in DOS using the regular battery of tests and games, to see how different brands / models behave in games and how fast they are.

Initially I started testing the cards in one of my 586 machines, but after seeing lots of cards producing identical results in several tests, I decided to move to a faster machine to avoid a CPU bottleneck. Here's the test setup:

I0Vwg7Xh.jpg

SJtEZpIh.jpg

233MHz Pentium MMX
Tekram Impression 533TX (Intel 430TX Chipset, 512KB L2 cache)
64MB EDO DRAM
Creative AWE64 Value
MS-DOS 7.1 (Win95osr2)

The tests were performed under pure DOS, with all drivers loaded, including sound. Games were tested and benchmarked with sound to simulate real world gaming conditions. Here are the results:

iolsTjYh.jpg

In 3DBench2 slowest of the bunch was the Alliance AT15, followed by the Cirrus Logic card. These two are 50% slower then most of the other cards tested. A big surprise here is the low score the SiS 6326 got - I was expecting better since it's a newer card and uses 8MB of SDRAM. The Virge DX is smack in the middle, being outpaced by the Trio64, but not by much. Looking over the numbers, it seems the benchmark is rather picky about how well it will run on certain chipsets - since some of the slower cards (particularly the Cirrus) perform very well in other tests. The fastest card was the Voodoo Banshee PCI, trailed by the 3D Labs Parmedia 2.

3juUtJ2h.jpg

PC-Player benchmark at 320x200 shuffles things up a bit. At this resolution the slowest cards are the Trident 9440 and again the SiS 6326. 3D Labs Parmedia 2 takes the lead, trailed by the two Matrox cards. The S3 Virge is right in the middle of the lot. The Alliance AT15 refused to run the benchmark either at 320x200 or 640x400. The Trio64 and Virge DX are very close in performance at this resolution, but things change at 640x400

ZLr2Jpyh.jpg

Here the Virge distances itself from the Trio64, witch is near the bottom of the charts - second to last, right next the the Cirrus Logic card. The Virge however is number 4, right behind the Millennium. The fastest card here is again the Banshee, this time trailed by the MX based Rush witch comes second - a bit of a surprise. The SiS card performed better, as we will see it tends to do at higher resolutions.

j1l3fGth.jpg

C3D Bench is a simple geometry and lighting benchmark. Here it seems memory speed matters little - as we see from the SiS card's results. The Banshee is again in the lead, trailed by the Parmedia 2. The Trio64 and Virge DX are right in the middle, as is the AT25.

5tYG77Fh.jpg

Increasing the resolution moves the Millenium 2 right next to the Banshee, both scoring 47.9 FPS. The Parmedia 2 and MX based Rush are neck and neck on second place, while in 3rd we see the Virge DX and Millennium 1. Again the Virge DX shows it's superiority to the Trio64, the latter being in the middle, this time scoring a little lower then the Trident 9440. The trident card consistently scores well in geometry and high-res benchmarks - witch leads me to conclude that my sample is crippled by the slow 1MB of FPM ram installed. Cards with fast FPM or EDO only - like the 9440-3 (fast FPM) or 9680 (EDO vram) would perform quite a bit better. The Matrox Mystique refused to run C3D bench SVGA - the program would display "benchmark starting in 5 seconds", then black screen, then "benchmark starting in 5 seconds" again, and it would loop like that until I rebooted the machine.

Frankly I was expecting the Matrox cards to be faster then the Virge DX in most benchmarks, but only the Millennium 2 manages to truly overtake it. At 320x200 the matrox cards are faster, but at 640x400 The Millennium 1 and Mystique are either on par, or slightly slower then the DX witch I find surprising. In 3DBench the Trio64 equals or overtakes all 3 matrox cards, but in PCP and C3D it falls quite a bit behind, witch leads me to think 3DBench employs older rendering techniques witch favor the older Trio64 - then again, the Parmedia 2 and Voodoo Banshee - a much newer cards - come out on top by quite a margin... curious.

Now for the games! I only tested Quake 1.06 (timedemo demo1) and Doom (custom timedemo I recorded while playing trough E1M1). I would have loved to include Descent, but the DOS version does not have a FPS counter like the 3dfx version does.

Bjaiuq2h.jpg

In doom, the fastest card is the Voodoo Banshee - witch is shaping up to be the fastest 2D card of the bunch - followed by all 3 matrox cards, the Virge DX, Trio64 and Parmedia 2. While there is quite a bit of a difference between the Banshee and the rest of the pack, there is little difference between places 2 trought 7. The DX performs very close to the Millennium - only 1.2 fps difference - while the Millennium II and Mystique pull ahead 1 more FPS. The AT15 is by far the slowest of the bunch, managing to be over two times slower then the Banshee. The AT25 follows suit.

aegUISah.jpg

In Quake we have the 3D Labs Parmedia 2 and Voodoo Banshee right at the top with over 48 fps, closely fallowed by the matrox cards. The S3 cards are both in the middle, with the SiS 6326 and Cirrus Logic cards at the bottom. There is a ~19% difference in performance between the slowest and fastest cards in this test, so the margin is not as big as it was in doom. Still the game is perfectly playable on all cards tested. The SiS card only redeems itself at 640x400, where it scores close to the Millennium II - unfortunately not all cards (about half) would run quake at 640x400 so I did not include it in the bechmarks.

Now for game compatibility:

2xrBJWOh.jpg

I tested all cars with some of the more troublesome DOS games, witch worked perfectly on all but the Matrox cards, the AT25 on the Voodoo Rush and the 3D Labs Parmedia 2. The rest ran all games flawlessly. The AT15 showed rare minor "jumps" while scrolling but they are only noticeable if you really look for them. The smoothest of the bunch was the Cirrus Logic card, followed by the SiS. Jazz ran particularly smooth on these, as did Supaplex. On the Matrox cards both Jazz and Keen were nausea-inducing - the screen jumped around whenever the image would scroll to the left or right. The Paramedia 2 showed "flickering corruption" in keen - like some sort of artefacting that would pop up now and again while scrolling, on only in parts of the screen. Golden axe displayed tearing while scrolling, like playing a 3D game with v-sync off.

Conclusions:

The S3 Trio64 and ALG2302 seem to have the best ballance of performance and compatibility - usually sitting in the middle of the benchmark charts. All games ran flawlessly on them, and they are solid performers. The Trident card seems to only pick up speed at higher resolutions. Performance-wise it beats the S3 Trio at 640x400, but stays behind it at 320x200. The trident also feels smoother then the S3 in windows at resolutions over 800x600. I'd say it's a good card for retro gaming - if seems to run just right while playing older low-res games, and it seems quite competitive at higher resolutions. The Trident is also one of the few cards able to run quake at 640x400, and it performed a little slower then the S3 Virge DX, while trailing behind both cards at 320x200 - curious huh?

The Matrox cards show their muscle in games and in windows, unfortunately quite a few games refuse to run properly on them. They are great for games like Quake and Duke3D in VESA modes, but older titles - particularly platforms - seem to have issues with these cards. Even newer DOS games like C&C Gold (the DOS version) have issues - the game tears and stutters when scrolling left and right, but not when scrolling up and down.

The ALi ALG2302 surprised me quite a bit. It not only performs well enough, it ran every game I could come up with perfectly. So did the MX86151 equipped voodoo Rush, witch unlike it's AT25 equipped brother, managed to both perform very well and run all games perfectly - no tearing or stuttering whatsoever. I'd say the MX86151 is the best Voodoo Rush card for a DOS / early windows glide PC.

As for the title of BEST OVERALL DOS GAMING PCI CARD - I'd give that to the PCI Voodoo Banshee. Not only does it manage to be the fastest card tested, but it runs all games w/o any glitches or bugs, has excelent image quality in both DOS and Windows regardless of resolution, and is 3D capable! GL_Quake, Quake2, 3DFX Descent and Carmageddon as well as Blood, MW2, MW3 and other early 3D games will run beautifully on this card on a pentium MMX system, or in a K6-III - the latter allowing for slow-downs critical for speed sensitive games. This card can cover a lot of gaming territory, making it the best choice for a late 80's - late 90's gaming rig. Some 2000 games even manage to run pretty well on it + a K6-III - like NFS Porsche (albeit at lower resolutions and detail levels).

Reply 1 of 25, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Thank you so much for all your hard work on putting this together! That's a great selection of cards and this really gives a good idea of what to expect from each one.

I know the Voodoo 3 and Voodoo Banshee are fairly similar, I wonder how they compare for DOS tests. From my experience, the Voodoo 3s are a bit more common. I've also found several Banshees that have run so hot for so long that they're actually discolored, which is concerning... V3s seem like they are cooled a bit better.

On that same note, a TNT or TNT2 would be very interesting to see compared for a couple tests. 😀

My most recent Windows 98\DOS time machine (K6-2 500) is using a TNT 2 Pro, but I'd be interested to know how its compatibility and DOS speed compare to the rest.

Now for some blitting from the back buffer.

Reply 2 of 25, by clueless1

User metadata
Rank l33t
Rank
l33t

Excellent roundup. Thanks, that was really cool. The Banshee really seems to be a gem! I wonder if it's similar enough to the Voodoo3 2000, a card I've been able to find a couple of compatibility issues with:
-in Wing Commander 3 it needs to use the alternate video mode to play the movie cut scenes. Check the setup program video tests.
-in Duke 3D, my V3 refuses to run in 1024x768. It crashes shortly after launching the game. I tested 10 graphics cards awhile back and found the V3 and Rage 128 Pro both did this, but the other 8 cards could run that res:
10 Way DOS Graphic Card Benchmarks on Celeron 333

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 3 of 25, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
Ozzuneoj wrote:
Thank you so much for all your hard work on putting this together! That's a great selection of cards and this really gives a goo […]
Show full quote

Thank you so much for all your hard work on putting this together! That's a great selection of cards and this really gives a good idea of what to expect from each one.

I know the Voodoo 3 and Voodoo Banshee are fairly similar, I wonder how they compare for DOS tests. From my experience, the Voodoo 3s are a bit more common. I've also found several Banshees that have run so hot for so long that they're actually discolored, which is concerning... V3s seem like they are cooled a bit better.

On that same note, a TNT or TNT2 would be very interesting to see compared for a couple tests. 😀

My most recent Windows 98\DOS time machine (K6-2 500) is using a TNT 2 Pro, but I'd be interested to know how its compatibility and DOS speed compare to the rest.

Thank you 🤣

Unfortunately I don't have PCI versions of the TNT / TNT2 or Voodoo 3 to test. This review actually contains all the PCI cards I own. I will make another review testing AGP cards - from the Riva 128ZX, to the Vanta, TNT2, Geforce and so on.

As far as I remember, up to the TNT2 nvidia cards have pretty good DOS compatibility and the Voodoo 3 is also a great DOS card.

clueless1 wrote:
Excellent roundup. Thanks, that was really cool. The Banshee really seems to be a gem! I wonder if it's similar enough to the […]
Show full quote

Excellent roundup. Thanks, that was really cool. The Banshee really seems to be a gem! I wonder if it's similar enough to the Voodoo3 2000, a card I've been able to find a couple of compatibility issues with:
-in Wing Commander 3 it needs to use the alternate video mode to play the movie cut scenes. Check the setup program video tests.
-in Duke 3D, my V3 refuses to run in 1024x768. It crashes shortly after launching the game. I tested 10 graphics cards awhile back and found the V3 and Rage 128 Pro both did this, but the other 8 cards could run that res:
10 Way DOS Graphic Card Benchmarks on Celeron 333

Thanks a lot. I'll test Wing Commander 3 on the banshee. As for Duke3D, it runs fine at 1024x768 on the banshee (tested it a while ago on my K6-III rig).

Reply 4 of 25, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t

Whoops... yeah, I forgot about the cards all being PCI. There aren't many TNT\TNT2 PCI cards out there, and the PCI V3s tend to be quite pricey.

Its great to know where all these cards stand though! I'll probably try some similar tests on my K6-2 500 system (ISA, PCI and AGP) eventually. I have more 1995-2001 graphics cards than I care to admit, so I'd better use them for something.

Will be interesting to see how they compare. 😀

Now for some blitting from the back buffer.

Reply 5 of 25, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
Ozzuneoj wrote:

Whoops... yeah, I forgot about the cards all being PCI. There aren't many TNT\TNT2 PCI cards out there, and the PCI V3s tend to be quite pricey.

Its great to know where all these cards stand though! I'll probably try some similar tests on my K6-2 500 system (ISA, PCI and AGP) eventually. I have more 1995-2001 graphics cards than I care to admit, so I'd better use them for something.

Will be interesting to see how they compare. 😀

Glad to be of help 😀 - I'm just sorry I don't have a Cirrus Logic CL-GD5436 or GD5462 card - they miles above the GD54M30 in this review and I kind of feel like the only cirrus card here sets a bad example.

Reply 6 of 25, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t
kanecvr wrote:
Ozzuneoj wrote:

Whoops... yeah, I forgot about the cards all being PCI. There aren't many TNT\TNT2 PCI cards out there, and the PCI V3s tend to be quite pricey.

Its great to know where all these cards stand though! I'll probably try some similar tests on my K6-2 500 system (ISA, PCI and AGP) eventually. I have more 1995-2001 graphics cards than I care to admit, so I'd better use them for something.

Will be interesting to see how they compare. 😀

Glad to be of help 😀 - I'm just sorry I don't have a Cirrus Logic CL-GD5436 or GD5462 card - they miles above the GD54M30 in this review and I kind of feel like the only cirrus card here sets a bad example.

I recently discovered that I had packed away in a box (of cards I probably won't use, but I couldn't just get rid of) a GD5440. I will be sure to test it out when I finally get around to testing these things. 😀

Now for some blitting from the back buffer.

Reply 7 of 25, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie

Ok I tried Wing Commander 3 (it was a pain in the ass to get working) and it seems it displays correctly and videos are fine on the Banshee.

Reply 8 of 25, by clueless1

User metadata
Rank l33t
Rank
l33t

Cool. Seems it's a better DOS card than the V3 😀 Thanks for testing!

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 9 of 25, by James-F

User metadata
Rank Oldbie
Rank
Oldbie

Excellent benchmark write-up, Thanks!
I have the exact same setup as you kanecvr so I'll post some benchmarks with the VGA cards I have.

Pentium MMX 233MHz.
Intel 430TX Chipset, 512KB L2 cache.
64MB EDO DRAM.
SD to IDE adapter with 32GB UHS-1 (Class10) SD card.
Yamaha YMF719E-S.
MS-DOS 7.1 (Win98SE).

TNT2 M64 32MB:

3DBench2: 176.1
PC Player 320x200: 58.5
PC Player 640x400: 27.3
Chris 3D 320x200: 165.0
Chris 3D 640x480: 48.2
Doom: 93.7
Quake: 52.4

CL5446 2MB

3DBench2: 159.1
PC Player 320x200: 55.9
PC Player 640x400: 25.7
Chris 3D 320x200: 134.5
Chris 3D 640x480: 43.5
Doom: 85.0
Quake: 50.6

The TNT M64 is very compatible card with VBE 3.0, special resolutions and special text modes, but it lacks UniVBE support or 3DFX.
The CL5446 is also a great VGA card for DOS gaming, but it is VBE 1.2 and lacks some of the special text modes like 132x43 and 4x18 fonts and special resolutions most noticeable in demoscenes.
Both work with special game resolutions and highly compatible for DOS gaming according to THIS infamous page, and my testings.

The difference in benchmark can be considered negligible, as for DOS gaming the CPU does the heavy lifting unless you use a very old VGA card which retards even the 320x200 resolution.

Last edited by James-F on 2016-10-17, 07:14. Edited 8 times in total.


my important / useful posts are here

Reply 10 of 25, by brostenen

User metadata
Rank l33t++
Rank
l33t++

Cool test resultat. I kind of had the idea, that both CL and S3's might have been bottle necked by a 5x86-133. As they were so close in my tests. I will keep these results in mind, when building a Pentium1 system in the future.

Don't eat stuff off a 15 year old never cleaned cpu cooler.
Those cakes make you sick....

My blog: http://to9xct.blogspot.dk
My YouTube: https://www.youtube.com/user/brostenen

001100 010010 011110 100001 101101 110011

Reply 11 of 25, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Very well done kanecvr!

And I can remember the SiS6326 being exceptionally good at being exceptionally slow, even for generic Windows stuff. But we did often use XP at lower resolutions, having tweaked the GUI to not be so bloated which didn't help it too much, even Virge seemed to be faster.

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 12 of 25, by clueless1

User metadata
Rank l33t
Rank
l33t
James-F wrote:

The TNT M64 is very compatible card with VBE 3.0, special resolutions and special text modes, but it lacks UniVBE support or 3DFX.

I noticed the other night on my Pentium/TNT2 M64 PC that Tie Fighter Collector's CD would not load, giving a VESA error. I don't remember having issues with that game, but I did switch to the TNT2 relatively recently. If in fact it's a compatibility issue with the card, this would be the first DOS game I've come across that has a problem with the TNT2 M64. If you have this game, would you give it a go?

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 13 of 25, by Ozzuneoj

User metadata
Rank l33t
Rank
l33t
James-F wrote:
Excellent benchmark write-up, Thanks! I have the exact same setup as you kanecvr so I'll post some benchmarks with the VGA card […]
Show full quote

Excellent benchmark write-up, Thanks!
I have the exact same setup as you kanecvr so I'll post some benchmarks with the VGA cards I have.

Pentium MMX 233MHz.
Intel 430TX Chipset, 512KB L2 cache.
64MB EDO DRAM.
SD to IDE adapter with 32GB UHS-1 (Class10) SD card.
Yamaha YMF719E-S.
MS-DOS 7.1 (Win98SE).

TNT2 M64 32MB:

3DBench2: 176.1
PC Player 320x200: 58.5
PC Player 640x400: 27.3
Chris 3D 320x200: 165.0
Chris 3D 640x480: 48.2
Doom: 93.7
Quake: 52.4

CL5446 2MB

3DBench2: 159.1
PC Player 320x200: 55.9
PC Player 640x400: 25.7
Chris 3D 320x200: 134.5
Chris 3D 640x480: 43.5
Doom: 85.0
Quake: 50.6

The TNT M64 is very compatible card with VBE 3.0, special resolutions and special text modes, but it lacks UniVBE support or 3DFX.
The CL5446 is also a great VGA card for DOS gaming, but it is VBE 1.2 and lacks some of the special text modes like 132x43 and 4x18 fonts and special resolutions most noticeable in demoscenes.
Both work with special game resolutions and highly compatible for DOS gaming according to THIS infamous page, and my testings.

The difference in benchmark can be considered negligible, as for DOS gaming the CPU does the heavy lifting unless you use a very old VGA card which retards even the 320x200 resolution.

Do the TNT2 M64 and TNT2 Pro have a similar level of compatibility and performance in DOS? According to wikipedia the TNT2 (standard, Pro and Ultra) are "NV5" chips while the M64 and Vanta chips are actually "NV6"... but I can't seem to find any detailed explanation of the differences, if there are any. I have a Pro in my main 9x\DOS system but haven't really done enough testing to determine how compatible it is. I haven't had any problems with it yet though.

Now for some blitting from the back buffer.

Reply 14 of 25, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote:
James-F wrote:

The TNT M64 is very compatible card with VBE 3.0, special resolutions and special text modes, but it lacks UniVBE support or 3DFX.

I noticed the other night on my Pentium/TNT2 M64 PC that Tie Fighter Collector's CD would not load, giving a VESA error. I don't remember having issues with that game, but I did switch to the TNT2 relatively recently. If in fact it's a compatibility issue with the card, this would be the first DOS game I've come across that has a problem with the TNT2 M64. If you have this game, would you give it a go?

I have an "arrrgh!" version of Tie Fighter (if you catch my drift) and I remember it works perfectly on the Banshee, the cirrus, trident and the S3 cards. Have not tested it on other cards.

The TNT2 has a few compatibility issues with some dos games - tearing and stuttering in a few older dos games - but I haven't tested it extensively since I only have AGP TNT2 cards and I saw no point in testing most of these games on an AGP machine. It would kind of make sense on a K6-III build tough.

Reply 15 of 25, by clueless1

User metadata
Rank l33t
Rank
l33t

I nailed down the issue with Tie Fighter, and it has nothing to do with the TNT2 M64. Bizarrely, TIECD gives the VESA error if it's installed to any partition other than C:

edit: man, Banshees are expensive! AGP versions are very common, but going for $30-$60. I see only one PCI version on ebay, for $99.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 16 of 25, by Imperious

User metadata
Rank Oldbie
Rank
Oldbie

I got a PCI banshee as well as a 2Mb S3 Virge DX for $45 AUD shipped. The Banshee only had TGA-BS written on it, no 3dfx logo or other obvious identification.

After a clean up from what looked like smoke damage from a fire It worked beautifully and has got the best Image quality I've seen in quite a while (compared with other PCI cards in my 486).
Massive overkill for my 486 but I might just leave it in there.

Atari 2600, TI994a, Vic20, c64, ZX Spectrum 128, Amstrad CPC464, Atari 65XE, Commodore Plus/4, Amiga 500
PC's from XT 8088, 486, Pentium MMX, K6, Athlon, P3, P4, 775, to current Ryzen 5600x.

Reply 17 of 25, by kanecvr

User metadata
Rank Oldbie
Rank
Oldbie

I got some PCI cards in trade today and benchmarked them using the same PC and test methodology:

nhWx6C3h.jpg?1

- ATi 254VT4 2MB PCI
- OAK Spitfire OTI641111 2MB PCI
- Cirrus Logic CL-GD5446BV-HC-A 1MB PCI

I'm posting the preliminary results here, and I will add them to the first thread later on.

Cirrus Logic CL-GD5446BV-HC-A 1MB PCI

Quake 1.09 SW Timedemo 1 - 49.0 FPS - yes, you read right. This thing tops the Quake performance charts. Brostenen was right, these babies are FAST! I took a photo for all non-believers and CL haters out there.
Doom E1M1 speedrun timedemo - 584 realticks
PC Player Benchmark 320x200 - 57 fps - best score in PCP at this resolution. It tops the matrox and 3dlabs cards
PC Player Benchmark 640x400 - 24.9 FPS - this puts it on par with the Matrox Millennium II and the Mystique.
3DBench2: 163.1 fps
Chris's 3D Benchmark vga - 149.4
Chris's 3D Benchmark svga - 43.4

Compatibility-wise this card runs every game I tested perfectly. No tearing or stuttering whatsoever.

Image quality is great. Unfortunately I couldn't test high resolutions in windows since the extra vram chips I have are 70ns and the card doesn't like them (artifacts) and I couldn't be bothered to pull some 60ns chips out of my virge card, but what I have managed to test looks great. At low resolutions the card seems to display warmer colors - again, like Brostenen mentioned in his thread. It looks great on an LCD. 800x600 is crisp and clear, in windows and in Duke Nukem 3D.

OAK Spitfire OTI641111 2MB PCI

Quake 1.09 SW Timedemo 1 - 43.1 FPS
Doom E1M1 speedrun timedemo - 776 realticks
PC Player Benchmark 320x200 - 49 fps
PC Player Benchmark 640x400 - 19 FPS
3DBench2: 110.4 fps
Chris's 3D Benchmark vga - 99.7 fps
Chris's 3D Benchmark svga - 20.9 fps

This is one of the slower cards in the tests, but still offers decent performance. Nothing impressive about it.

Compatibility - this card ran all games flawlessly.

Image quality - good. I didn't install windows drivers, so I tested in DOS with Duke3D and Quake. Both games looked really good, especially duke. Too bad the card isn't a bit faster.

-ATi 254VT4 2MB PCI

Quake 1.09 SW Timedemo 1 - 43.1 FPS No, I didn't mix up oak spitfire results with the ATI results. Regartless of how many times I re-run the timedemo, it scored 43.1 FPS.
Doom E1M1 speedrun timedemo - 584 realticks - same performace as the Cirrus cards, and a few others in this test.
PC Player Benchmark 320x200 - 49.3 fps
PC Player Benchmark 640x400 - 25.8 FPS
3DBench2: 136.3 fps
Chris's 3D Benchmark vga - 101 fps
Chris's 3D Benchmark svga - 44.2 fps

This cards seems to perform similarly to the Trident tested before. It does well at 640x400, but has mediocre performance at 320x200 - except for Doom, witch puts it in the upper part of the charts.

Compatibility - Jazz was OK, as well as Supaplex and Prehistorik. Golden Axe seems OK as well. The image sometimes "jitters" when scrolling, but nothing too noticeable unless you look for it. I did see a scrolling problem in keen. It's like a slight stuttering - not as bad as on the matrox cards, but noticeable. The game is playable, but the stuttering can get annoying after a while.

Image quality - this is where the ATi card really shines. Output is on par with matrox cards.

Reply 18 of 25, by brostenen

User metadata
Rank l33t++
Rank
l33t++

Yeah.... 5446 for the win. 😁
I actually have two 5446's with 2mb 60ns each. 😜
Looking at eBay that usually are extreme in prices, I guess the 5446 gives lot of Bang for the bucks. They are like somewhat between 17 to 20 US Dollars each, plus shipping. Great value for the money.

Don't eat stuff off a 15 year old never cleaned cpu cooler.
Those cakes make you sick....

My blog: http://to9xct.blogspot.dk
My YouTube: https://www.youtube.com/user/brostenen

001100 010010 011110 100001 101101 110011

Reply 19 of 25, by clueless1

User metadata
Rank l33t
Rank
l33t

Now we need a TNT2 M64 vs Banshee vs CL-GD5446 grudge match.

BTW, are you running the standard Doom test? If so, then 74690/584 realticks = 127.89 fps, which is much faster than the Banshee.

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks