vvbee wrote:

Had a brief look at the 128 pro v the matrox g400 with newest drivers and a fast cpu in 640 x 480 x 16 in some number of games from 199x to 200x. Tossup for speed, saw several graphics glitches with the rage though and the image quality is coarser. Matrox has a more mature driver whatever the case, not sure that there's a technical reason to prefer the rage. Maybe subjectively aesthetic or some corner case.

Rage was always a lost cause, but R200 and up actually got decent drivers towards the end 🤣

swaaye wrote:

The anisotropic filtering is very fast. But a Radeon 9500 or newer has much improved texture filtering overall.

It is only fast because they are cheating! And they still were with the R3x0. Here's an interesting article about that from the German 3DCenter (translated into English): https://alt.3dcenter.org/artikel/2003/11-21_a_english.php

Unfortunately, the "I only care for fps and don't know what image quality is" consumer accepted products like these. With the GeForce 6 and 7 series nVidia joined the club of AF cheaters (GeForce 8 and newer is back to close-to-perfect AF).

That's why I always have a really hard time picking an optimal GPU for that DX8 to early DX9 time frame. ATI cheats with AF throughout all their cards, so no ATI. GeForce4 is good, but only for earlier DX8 stuff. GeForce FX is good, but not very good for heavier DX9 load due to the quirkiness of its GPU architecture. GeForce 6 & 7 cheat with AF like ATI does, so no GeForce 6 or 7. That leaves GeForce 8 or newer for heavier DX9 loads the GeForce FX can't handle anymore... But GeForce 8 is PCIe already...

If someone doesn't care for image quality than this point of course is irrelevant.

ATI does not 'cheat', their R200 architecture simply does not support Trilinear filtering and Anisotropic filtering together. The result of this is a slight loss in filtering quality but nothing major. They did 'cheat' with texture quality overall in some drivers (the Quack debacle) but those 'optimizations' were quickly rolled back.

That article is full of shit wrt the R300, by the way. I was around at the time of R300 anisotropic filtering optimizations, and those optimizations used adaptive levels of filtering on vertical surfaces, the resulting IQ difference was negiligble. Regardless, they quickly added an optimization toggle switch to the drivers to disable that! The website's lack of any meaningful screenshot comparisons is testament to that.

Saying ATI 'cheated' with those adaptive filtering optimizations is like saying EVERYONE cheats today with adaptive anti aliasing algorithms, tesselation optimizations etc. Did nVidia cheat with their shitty anti aliasing as well when the 4x nvidia AA was equivalent to the 2x ATI AA? Especially when you had no 'Quality' setting at all?

https://www.pcper.com/reviews/Graphics-Cards/ … liasing-Quality

We can see that at 2x anti-aliasing on the Radeon 9700 is as good as the 4x anti-aliasing on the NVIDIA cards.

No, they simply optimized differently. No card has PERFECT IQ.

Before you post clickbait articles from Mr. Nobody websites, here's a link from Anandtech:

https://www.anandtech.com/show/970/14

It's important to note that in most cases (such as the one above), you won't be able to tell any difference between ATI's performance and quality anisotropic filtering settings.

appiah4 wrote:

Before you post clickbait articles from Mr. Nobody websites, here's a link from Anandtech:

https://www.anandtech.com/show/970/14

It's important to note that in most cases (such as the one above), you won't be able to tell any difference between ATI's performance and quality anisotropic filtering settings.

Well, this "Mr. Nobody" website and one of their forum members actually created the D3D AF Tester which is used throughout many AF implementation analysis attempts, even at Anandtech. So, just because you don't know this specific website doesn't necessarily mean they aren't known by the rest of the world or that they have to be considered "Mr. Nobody" websites.

AnandTech themselves say (in their review of the HD5870) that ATI finally has angle-independent AF with the HD5000 (using the D3D AF Tester hosted at 3DCenter.org...): https://www.anandtech.com/show/2841/13

I will not argue with what you said (as I'm no GPU architect and as I don't want to start some sort of dispute), I just brought up an article I remembered from back then (I bought an FX 5900 Ultra in mid-2003 after reading many reviews of then current cards and did not regret it, although it was replaced after a shorter than usual while with a GeForce 6800GT which also has a subpar AF implementation).

PS: I don't really know why you use such rude language. You could have just said you think I'm wrong for whatever reason without some of those vulgarities etc. In no way did I offend you personally. I think we should leave emotions out of this. This is a civilized community after all, isn't it?

I was hostile towards the contents of the article, not yourself, but even so the most I said was that it was "full of shit" which would hardly count as 'vulgar' language. I did not mean to be rude towards you, if I came across as such you have my apologies.

OK then, everything's fine.

This is what I get for mentioning ati Rage... 🤣

appiah4 wrote:

ATI does not 'cheat', their R200 architecture simply does not support Trilinear filtering and Anisotropic filtering together. T […]
Show full quote

ATI does not 'cheat', their R200 architecture simply does not support Trilinear filtering and Anisotropic filtering together. The result of this is a slight loss in filtering quality but nothing major. They did 'cheat' with texture quality overall in some drivers (the Quack debacle) but those 'optimizations' were quickly rolled back.

That article is full of shit wrt the R300, by the way. I was around at the time of R300 anisotropic filtering optimizations, and those optimizations used adaptive levels of filtering on vertical surfaces, the resulting IQ difference was negiligble. Regardless, they quickly added an optimization toggle switch to the drivers to disable that! The website's lack of any meaningful screenshot comparisons is testament to that.

Saying ATI 'cheated' with those adaptive filtering optimizations is like saying EVERYONE cheats today with adaptive anti aliasing algorithms, tesselation optimizations etc. Did nVidia cheat with their shitty anti aliasing as well when the 4x nvidia AA was equivalent to the 2x ATI AA? Especially when you had no 'Quality' setting at all?

https://www.pcper.com/reviews/Graphics-Cards/ … liasing-Quality

We can see that at 2x anti-aliasing on the Radeon 9700 is as good as the 4x anti-aliasing on the NVIDIA cards.

No, they simply optimized differently. No card has PERFECT IQ.

Before you post clickbait articles from Mr. Nobody websites, here's a link from Anandtech:

https://www.anandtech.com/show/970/14

It's important to note that in most cases (such as the one above), you won't be able to tell any difference between ATI's performance and quality anisotropic filtering settings.

Seems like you didn't really read the article, imo. I read it and it doesn't say ati "cheated". I think they explain fine how there are trade offs when implementing these things (quality vs transistor budget). They also point out that you may not notice it in practice. You may not agree with their conclusions, which is fine but I think their data is fine. You can download the filter tester and try it out yourself

Are we saying here that no 3D game from the Windows 9x era (arguably 1997-2000) benefited more visually, in features or peformance, from an ATI card?

I'm particularly interested in the Rage Fury MAXX, which appears to have taken the briefiest lead over the GeForce SDR before the Geforce DDR was released, at least with raw FPS.

W Gruffydd wrote:

Are we saying here that no 3D game from the Windows 9x era (arguably 1997-2000) benefited more visually, in features or peformance, from an ATI card?

I'm particularly interested in the Rage Fury MAXX, which appears to have taken the briefiest lead over the GeForce SDR before the Geforce DDR was released, at least with raw FPS.

As I mentioned before -- Tomb Raider 1's Windows Version is ATICIF only and may be the best looking one. It also can run at a locked 30FPS at 800x600 on a ATI Rage Pro Turbo and while the PowerVR versions can do 1024x768 they probably run below 30FPS, I imagine.

For my part, I remember fondly the ATI Rage Pro that I had in my first PII build back in 1997-8 for its overall crapiness, poor performance in games and especially poor support in OpenGL/Linux (which I was heavily into at the time). I got rid of it when I afforded a TNT2, that made a big difference. However not long ago I acquired an identical one (ATI Xpert@Play 98) to use in a nostalgic build at some point. 😀

That bad early experience with ATI didn't deter me from reverting back from Nvidia at the time of Radeon HD6000 series, and to this day my main desktop has an ATI/AMD video card.

the resulting IQ difference was negiligble.

It is very noticeable in some examples, but overall it was a nice trade-off (quality/performance wise). Nvidia adopted it too (NV40).

Regardless, they quickly added an optimization toggle switch to the drivers to disable that!

They didn't. Angle free filtering was added only with R520 and with substantial performance loss.

stamasd wrote:

For my part, I remember fondly the ATI Rage Pro that I had in my first PII build back in 1997-8 for its overall crapiness, poor performance in games and especially poor support in OpenGL/Linux (which I was heavily into at the time). I got rid of it when I afforded a TNT2, that made a big difference. However not long ago I acquired an identical one (ATI Xpert@Play 98) to use in a nostalgic build at some point. 😀

That bad early experience with ATI didn't deter me from reverting back from Nvidia at the time of Radeon HD6000 series, and to this day my main desktop has an ATI/AMD video card.

ATI Mach64 cards were only ever 3d capable as an afterthought, although the 2d quality was usually a cut above most. I even remember using Michel Daenzer's Linux Mach64 DRM/DRI driver with XFree86 4.x in Linux. To be fair, it was better than software rendering.

dionb wrote:

For significant periods from 2000-2010 (and for that matter 1987-~1991) ATi cards offered the best performance of the day and several times ATi was the first to introduce new features or support new DirectX standards.

It really depends on your timeframe and goals with your retro system.

This

At the time, ATI cards were often the best value for money when getting a brand new graphics card for playing what were at the time modern games. Just like we would do now if we were buying the latest GTX 1080 or whatever

Now, with the luxury of being able to buy any card from any of those eras, nvidia are usually the allround best cards to get when you talk about drivers, a very wide range of games that they need to work with, etc etc

infiniteclouds wrote:

W Gruffydd wrote:

Are we saying here that no 3D game from the Windows 9x era (arguably 1997-2000) benefited more visually, in features or peformance, from an ATI card?

I'm particularly interested in the Rage Fury MAXX, which appears to have taken the briefiest lead over the GeForce SDR before the Geforce DDR was released, at least with raw FPS.

As I mentioned before -- Tomb Raider 1's Windows Version is ATICIF only and may be the best looking one. It also can run at a locked 30FPS at 800x600 on a ATI Rage Pro Turbo and while the PowerVR versions can do 1024x768 they probably run below 30FPS, I imagine.

I missed that. Do you mean it's the best looking version overall, or just the best-looking windows version? And did it have exclusive API features? Was its implementation of common visual features superior? If it was the best looking, the count of visually-superior ATI games of this era stands at one.

W Gruffydd wrote:

infiniteclouds wrote:

W Gruffydd wrote:

Are we saying here that no 3D game from the Windows 9x era (arguably 1997-2000) benefited more visually, in features or peformance, from an ATI card?

I'm particularly interested in the Rage Fury MAXX, which appears to have taken the briefiest lead over the GeForce SDR before the Geforce DDR was released, at least with raw FPS.

As I mentioned before -- Tomb Raider 1's Windows Version is ATICIF only and may be the best looking one. It also can run at a locked 30FPS at 800x600 on a ATI Rage Pro Turbo and while the PowerVR versions can do 1024x768 they probably run below 30FPS, I imagine.

I missed that. Do you mean it's the best looking version overall, or just the best-looking windows version? And did it have exclusive API features? Was its implementation of common visual features superior? If it was the best looking, the count of visually-superior ATI games of this era stands at one.

It is the only Windows version, and the best looking version. There is very little to 'texture rolling' than occurs as you move when compared to the 3dfx version. Again I've never played the PowerVR version but the 1024x768 resolution from videos I've seen doesn't run at 30+ FPS. The ATICIF version also has a patch out for it that you can use on your GOG or Steam install to run TombATI on modern OS (Windows 7) - it uses a GLRage wrapper.

I've played it on this computer with the patch as well as real hardware (Rage Pro Turbo) and it is great.

I replaced a VireDX with a Rage Pro PCI yesterday, and I can tell you that the VASTLY superior 2D image quality is a huge benefit compared to the S3 cards. Yes, it has trouble with high res vesa modes (Duke3D does not work in 800x600, but works in 640x480 for example) but screw it, the blurry Windows desktop was making my eyes hurt.

W Gruffydd wrote on 2018-09-02, 20:45:

infiniteclouds wrote:

W Gruffydd wrote:

Are we saying here that no 3D game from the Windows 9x era (arguably 1997-2000) benefited more visually, in features or peformance, from an ATI card?

I'm particularly interested in the Rage Fury MAXX, which appears to have taken the briefiest lead over the GeForce SDR before the Geforce DDR was released, at least with raw FPS.

As I mentioned before -- Tomb Raider 1's Windows Version is ATICIF only and may be the best looking one. It also can run at a locked 30FPS at 800x600 on a ATI Rage Pro Turbo and while the PowerVR versions can do 1024x768 they probably run below 30FPS, I imagine.

I missed that. Do you mean it's the best looking version overall, or just the best-looking windows version? And did it have exclusive API features? Was its implementation of common visual features superior? If it was the best looking, the count of visually-superior ATI games of this era stands at one.

My personal experience with more modern cards between a 9600 Pro and an FX5500 in using MAME in DOS made me choose ATI. I had already heard that ATI did better in DOS and I could see in MAME that the Nvidia card has a bad sync with the audio, in ATI it is difficult to perceive audio delay.