The Kyro had some unfixed hard locks with anything running UnrealEngine2 so its role as a gamer card was already dwindling

The games it works best with happen to be very british 😀 *COUGH*Anything by Computer Artworks, Lionhead and Rage*COUGH*

My GeForce 256 and GeForce 3 Ti200 are both dead... 😢

So here are my results:

9200 and 9200 SE tested with Catalyst 6.11
9550, 9700 Pro, 9800XT tested with Catalyst 9.3
FX5200, FX5600 tested with Detonator 43.51

I was messing around with different drivers before I took all the test scores, so some of the GPU-Z screenies show different driver versions:

Rage 128 Ultra

9200

9200SE

9550

9700 Pro

9800XT

FX5200

FX5600

Card photos to come!

GeForces & Radeons, GeForces & Radeons everywhere. Is anyone interested in Kyro II vs Voodoo5 vs Ati Rage Fury MAXX benchmarks? Suggestions are welcome.

batracio wrote:

GeForces & Radeons, GeForces & Radeons everywhere.

Terribly sorry buddy, but as awesome as I am, I just can't benchmark a card I don't own!!

🙁

But we like nineties cards, not noughties cards 😀

We like Tualatins too, don't we?

SquallStrife wrote:

We like Tualatins too, don't we?

We most certainly do!!! 😁

They are, imo, the best cpu's available at their particular speed and "processing power" 😉

They produce little heat, support SSE and are cheap (though somewhat it's harder to get the boards). Only drawback is it's memory bandwidth, Tualatins only work with SDRAM (or on the odd VIA DDR s370 boards, and we sure love VIA, don't we!! 😜)

my good lord gaming on the riva 128 is painful. sure the compatibility was awesome for the time, the 16-bit dither on those will really get on your tits like no other, it makes me cry for even voodoo dithering or geforce waffle dither. swaaye probably said everything that needs to be said about it

shots to come later. noticed nvidia snuck a '1.337' in the latest driver's version 😉

hey, on the bright side, it's adequate for a spot on full speed representation of Mario 64! (in CornH)

What I didn't realize until recently is that there are two Rage 128 chips. There's the original "Rage 128" and there's the "Rage 128 Pro" refresh. The latter one has decent dithering whereas the original is a dithering artifact extravaganza only matched by a DX10 card running 16-bit color games. 😉

swaaye wrote:

there are two Rage 128 chips. There's the original "Rage 128" and there's the "Rage 128 Pro" refresh. The latter one has decent dithering whereas the original is a dithering artifact extravaganza only matched by a DX10 card running 16-bit color games. 😉

Ow, that's new to me. I'll check which cards I have later, interesting.

Btw, is the Rage 128 (any!) good for retro gaming in any practical way?

no the R128 is not really a good choice for anything. I think it does have better DVD acceleration than TNT2 and Voodoo3 but that is it.

swaaye wrote:

What I didn't realize until recently is that there are two Rage 128 chips. There's the original "Rage 128" and there's the "Rage 128 Pro" refresh.

I wonder which one my ex-Dell "Rage 128 Ultra" is?

i should mention i didn't mention rage

leileilol wrote:

i should mention i didn't mention rage

🤣 I'm hallucinating Rages now!!

Well Riva 128 has its own problems including that strange noisy texture filtering. And the ugly texture seams.. And the stutter on texture loads in some games. And. .... 😉

I think all of the 3D chips up until 2001 or so had flaws. It wasn't until a little while after DirectX caught on thus forcing everyone to code around one standard that things really got nice, I think.

I wonder which one my ex-Dell "Rage 128 Ultra" is?

It'd be more related to the Rage 128 PRO, just different clock/memory speeds perhaps.

The differences between Rage 128 and 128 PRO/Ultra:

http://www.xbitlabs.com/articles/video/displa … ti-furypro.html

Very few and far in between. Most reasonable change is the addition of DXTC support. Also I should note that by the time TNT2/Voodoo3 cards were out, there were variants of original Rage 128-based cards with AGP 4x support.

I think all of the 3D chips up until 2001 or so had flaws.

Rage 128/128 Pro - too many
Radeon 7x00/8500/92x0 - weird anisotropic filtering implementation (cannot be used simultaneously with trilinear filtering), various bugs and issues with older games
KYRO/KYRO II - anisotropic filtering takes a HUGE performance hit when enabled (though its quality is only up to the level of the GF256/GF2/GF4 MX's), no hardware T&L, works great in some games, crappy in many others. Never tried one of these on any Pentium 4 motherboards but I keep hearing of lack of AGP 4x support. Not recommended for running games released after 2002.
TNT/TNT2 - trilinear approximation makes things at long distances look ugly when using trilinear mode
Voodoo3 - inability to render in 32-bit color (although its 22-bit postfilter is close enough if you ask me), 256x256 texture resolution limitation, no AGP 4x support - the most compatible chip for pre-2000 games however.
Voodoo4/5 - lacks a lot of features that the GeForce 256 already had 8 months earlier. only major difference with its AA implementation is rotated grid SSAA (as opposed to the ordered-grid SSAA on the GF256/GF2). difference is not really noticeable when playing games if you ask me.
SiS 315 - bilinear approximation (again, most noticeable with slightly distant objects), no anisotropic filtering support, some games crash, lots of games don't look right

makes you feel grateful for what we have now huh

It wasn't until the GeForce 256 (or the Radeon R300 line on ATI's side) that things started "looking right" IMO. The GeForce FX line is one big exception to that rule though - way more flaws (for DX9-level games and apps) than I could imagine.

One weird thing no one ever mentions about the KYRO is the imprecise texture mapping

try something with HUGE polygons (like say... old games) and check out that wobble! It'll blocky up too

What specs do not show are efficiency improvements, very similar to what Nvidia did with second TNT. Rage 128 Pro was ATI's first chip with every feature working correctly. While 90's accelerators had serious hardware bugs, flaws mentioned by Pippy are just conscious choices, tweaks or driver incapability.

What specs also don't show are anything resembling practical results in a game. 😉 I remember asking a Raven Software developer what he thought of T&L and if it solved any bottlenecks. He said that texturing was the #1 bottleneck with graphics, not geometry. He was working on the SOF2 team and JK2 was in development I recall.

leileilol wrote:

One weird thing no one ever mentions about the KYRO is the imprecise texture mapping

Which reminds me that PowerVR SGX has approximated bilinear filtering. I've been playing with a modded Nook Color and an original Motorola Droid and have noticed this myself. It's somewhat blocky. I suppose when your 3D chip has to use mere milliwatts, things have to be trimmed. 😉

It's interesting to play Android Quake 2 on it and look at the image quality. It renders at 1024x600 and it runs very smooth. Looks good too other than the very slightly ugly filtering.

I haven't tried Quake3 yet because it's just too hard to play these games on a touch screen.

GXL750 wrote:

I think all of the 3D chips up until 2001 or so had flaws. It wasn't until a little while after DirectX caught on thus forcing everyone to code around one standard that things really got nice, I think.

Coding around a single standard was the worst thing that could have happened because what was there to distinguish one card from another if they all ran the same games equally well? What was there to encourage innovation when innovation was no longer needed as long as your card was compatible with the latest version of DirectX? Why develop your own API's like Glide to make the games look better if nobody would program for it because DirectX took over?