Joseph_Joestar wrote on 2022-07-07, 14:10:

BEEN_Nath_58 wrote on 2022-07-07, 14:03:

To make sure my GPU is not just failing to render fog in Thief 2, I want to test them in these two games. For these one, mentioning the graphical settings and map should be sufficient.

For NFS High Stakes I used the demo version which only has one track. Not sure what it's called, but it does have fog. I think @bloodem has the full version of that game an is probably more acquainted with it. He did mention Celtic Ruins as a track which has fog.

As for Rogue Spear, I'm completely unfamiliar with that game. Never played it and don't own a copy.

I just discovered an open-source project on GitHub, DirectDrawCompat, brings back fog support. I don't know which fog, but it brought back whatever the fog is in Shadows of the Empire.

BEEN_Nath_58 wrote on 2022-07-07, 14:15:

I just discovered an open-source project on GitHub, DirectDrawCompat, brings back fog support. I don't know which fog, but it brought back whatever the fog is in Shadows of the Empire.

Does it work on Win98 or just on modern operating systems?

Joseph_Joestar wrote on 2022-07-07, 14:20:

BEEN_Nath_58 wrote on 2022-07-07, 14:15:

I just discovered an open-source project on GitHub, DirectDrawCompat, brings back fog support. I don't know which fog, but it brought back whatever the fog is in Shadows of the Empire.

Does it work on Win98 or just on modern operating systems?

It is for modern OS, starting Windows XP, although XP-7 are no longer properly supported.

I can confirm NFS High Stakes has fog, both with DDrawCompat and dgVoodoo2. Unfortunately, without these wrappers, the game just renders a black screen so DxWnd has to fix things for this game to again check out if Windows can really support this fog (and rule if this is table fog, since it hasn't worked in 3 earlier confirmed games).

Joseph_Joestar wrote on 2022-07-07, 14:10:

For NFS High Stakes I used the demo version which only has one track. Not sure what it's called, but it does have fog. I think @bloodem has the full version of that game an is probably more acquainted with it. He did mention Celtic Ruins as a track which has fog.

As for Rogue Spear, I'm completely unfamiliar with that game. Never played it and don't own a copy.

I played today with one of the overkill Win 98 systems in my hometown, which has a Radeon X850XT (Catalyst version 6.2, default installation, no registry tweaks).
I remembered that NFS High Stakes had issues with Catalyst 6.2 (it crashed to desktop immediately - just like with newer nVIDIA drivers), so I've decided to pursue this further to see if I can find a solution.
Finally, I found a patch (the unofficial NFS High Stakes XP 2.1 patch), which, among other things, also "improves the effects" and the game's compatibility with WinXP.

So I installed it and, sure enough, the game works perfectly on Win98 with it. But, more importantly, what surprised me is that it fully restores the fog effect on the Radeon card, and it looks as it should (just like with any nVIDIA card, at least the ones from that era).
It's not clear what this patch does exactly when it comes to the fog effect, but it's definitely not table fog anymore.

bloodem wrote on 2022-07-21, 10:49:

Finally, I found a patch (the unofficial NFS High Stakes XP 2.1 patch), which, among other things, also "improves the effects" and the game's compatibility with WinXP.

It's not surprising that the unofficial patch fixes the missing fog on ATi cards. This is similar to how the unofficial Thief 2 patch restores the missing stars in the night sky on Nvidia cards.

I avoid such third-party patches (mods would be a better word) while testing older games for the presence of these features, and only use the latest official patch released by the original developers. That said, these mods are very useful for people who just want to play the games without having to worry about graphical issues. 😀

Joseph_Joestar wrote on 2022-07-21, 11:06:

I avoid such third-party patches (mods would be a better word) while testing older games for the presence of these features, and only use the latest official patch released by the original developers. That said, these mods are very useful for people who just want to play the games without having to worry about graphical issues. 😀

Yeah, of course, I know.
I only mentioned it here because, not only does it fix the fog issue on the Radeon, but it actually fixes the startup crash with these newer drivers. So maybe someone will find it useful. 😀

bloodem wrote on 2022-07-21, 11:18:

I only mentioned it here because, not only does it fix the fog issue on the Radeon, but it actually fixes the startup crash with these newer drivers.

On that note, I had no problems with Catalyst 6.2 drivers when using the NFS: High Stakes demo on my Radeon 9250.

So that crash is either specific to the full version of the game or to the X850 card.

Joseph_Joestar wrote on 2022-07-21, 11:24:

On that note, I had no problems with Catalyst 6.2 drivers when using the NFS: High Stakes demo on my Radeon 9250.

So that crash is either specific to the full version of the game or to the X850 card.

Don't have the demo, but for me the startup crash happens with the full game (+ official patch 4.50) on anything newer than nVIDIA Detonator 3x.xx or ATI Catalyst 4.x (yes, it also crashes with the highly recommended Detonator 45.23 drivers) 😀.

bloodem wrote on 2022-07-21, 12:58:

Don't have the demo, but for me the startup crash happens with the full game (+ official patch 4.50) on anything newer than nVIDIA Detonator 3x.xx or ATI Catalyst 4.x (yes, it also crashes with the highly recommended Detonator 45.23 drivers) 😀.

You can download the demo from here.

Try it on the X850 and see if it crashes or not. I will test it with a GeForce4 Ti4200 and 45.23 later today.

Joseph_Joestar wrote on 2022-07-21, 13:24:

You can download the demo from here.

Try it on the X850 and see if it crashes or not. I will test it with a GeForce4 Ti4200 and 45.23 later today.

I've tested the demo just now on the Radeon X850XT + Catalyst 6.2. It does crash, but in a different manner compared to the full version.
The full version goes to a black screen and immediately crashes to the desktop (no errors, no warnings). This one does show an exception message.

It's interesting that for you the same driver worked (albeit, with a different card). I only tested cards like the 9800 PRO and the X800/X850 and they all exhibit the same issue (at least with the full game).
EDIT: make sure to run the 3DSetup program and switch to Direct3D - by default the demo (just like the full game) starts on Software mode and it does work! 😀

bloodem wrote on 2022-07-21, 13:51:

I've tested the demo just now on the Radeon X850XT + Catalyst 6.2. It does crash, but in a different manner compared to the full version.
The full version goes to a black screen and immediately crashes to the desktop (no errors, no warnings). This one does show an exception message.

I just tried the demo with a GeForce4 Ti4200 + 45.23 and it crashed with the same exception error as in your screenshot.

It's interesting that for you the same driver worked (albeit, with a different card). I only tested cards like the 9800 PRO and the X800/X850 and they all exhibit the same issue (at least with the full game).

Not sure if this affects anything, but my Radeon 9250 + Catalyst 6.2 was in a period-correct Celeron 466 machine at the time of testing. No issues with running the demo in D3D mode there, other than the missing fog.

Joseph_Joestar wrote on 2022-07-21, 14:56:

I just tried the demo with a GeForce4 Ti4200 + 45.23 and it crashed with the same exception error as in your screenshot.

Bingo! 😀 If you switch to Detonator 30.82 it should work (maybe some other drivers in between 30.82 and 45.23 also work, but never tried it).

Joseph_Joestar wrote on 2022-07-21, 14:56:

Not sure if this affects anything, but my Radeon 9250 + Catalyst 6.2 was in a period-correct Celeron 466 machine at the time of testing. No issues with running the demo in D3D mode there, other than the missing fog.

I don't think it's a platform issue. A much more plausible explanation for this different behavior is the actual difference in GPU architecture. The Radeon 9250 is an RV280 card (which is based on the R200 of the Radeon 8500 family).

So, it's quite possible that, when it comes to the older cards/architectures, the newer 6.2 driver actually continues to reuse a big portion of the old code from previous drivers, and only bringing significant changes for the newer R3xx & R4xx series (well, for the R4xx series the code is without a doubt totally new, since these cards are only supported by Catalyst 6.2 on Win98 😀 ).
This would also explain why, in contrast to the nVIDIA cards, older ATI cards don't typically suffer from performance issues when going with newer drivers (they simply reused more of the older code that they considered "stable" for those cards). Of course, this is just a guess, don't have any actual proof. 😁

bloodem wrote on 2022-07-21, 15:32:

If you switch to Detonator 30.82 it should work (maybe some other drivers in between 30.82 and 45.23 also work, but never tried it).

I just tried the NFS: High Stakes demo with a few different Nvidia driver versions on my GeForce4 Ti4200:

• 30.82 - works fine
• 40.72 - crash
• 45.23 - crash
• 56.64 - works fine

So it might not be the case of "driver is too new" but rather of "this driver version is incompatible". For reference, fog displayed correctly on all driver versions that didn't crash.

Despite what others have said, I still believe version 56.64 to be the best driver release for the FX series. I would like to see an example where an older driver is more compatible for any particular game.

Joseph_Joestar wrote on 2022-07-21, 17:36:

I just tried the NFS: High Stakes demo with a few different Nvidia driver versions on my GeForce4 Ti4200: […]
Show full quote

I just tried the NFS: High Stakes demo with a few different Nvidia driver versions on my GeForce4 Ti4200:

• 30.82 - works fine
• 40.72 - crash
• 45.23 - crash
• 56.64 - works fine

So it might not be the case of "driver is too new" but rather of "this driver version is incompatible". For reference, fog displayed correctly on all driver versions that didn't crash.

Interesting that it works with 56.64. I only tried that driver once, a long time ago, and it created issues with some of the games that I usually test (don't remember exact details, though).
Yeah, it's never about the driver being new, per se. But newer drivers tend to focus on newer software and legacy software support is gradually... forgotten. 😀

Kahenraz wrote on 2022-07-21, 18:54:

Despite what others have said, I still believe version 56.64 to be the best driver release for the FX series. I would like to see an example where an older driver is more compatible for any particular game.

Don't remember exact details, but I did face problems with it in some games, and I was specifically testing the FX series (GeForce FX5900/5900XT). Also, as it's always the case with nVIDIA, it's slower than older drivers.

I think it really depends on the hardware. I found 56.64 to be faster than earlier drivers on Pentium 4 processors from what I think are SSE2 enhancements.

You're absolutely right that it depends on the card. Pre-FX cards are definitely faster on earlier drivers on some configurations, such as slower SS7 and Pentium 2s.

My opinion is always to try 56.64 first, then work your way backwards if you find that it's slow or incompatible.

Kahenraz wrote on 2022-07-21, 19:19:

I think it really depends on the hardware. I found 56.64 to be faster than earlier drivers on Pentium 4 processors from what I think are SSE2 enhancements.

I've heard about this before (the fact that some drivers are apparently faster on processors that support SSE2), but I consider it more of an urban legend than anything else. Anyway, there might be an advantage on Pentium 4 (I only tested one driver version with my Pentium 4 builds, so don't know for sure), however that advantage is most likely due to other factors (maybe some improvements that specifically target the Netburst "hyper-pipelined" architecture?).

Anyway, when it comes to powerful GeForce FX cards, I test them exclusively on Core 2 Duo era machines (or at least Athlon 64 "Venice"), and all of these CPUs have SSE2 (and even SSE3) instructions. And it's on these specific platforms that 45.23 was consistently faster faster for me (which is why I say that the so called "SSE2 optimizations" are most likely misleading).
Anyway, it sounds like a fun experiment. When I get the chance, I will try and do a comparison again between 56.64 / 45.23 / 30.82 (the latter only for GeForce 4 Ti cards). Even more interesting would be to compare different platforms. Unfortunately, the problem is finding the time to do so. 😅

Kahenraz wrote on 2022-07-21, 19:19:

My opinion is always to try 56.64 first, then work your way backwards if you find that it's slow or incompatible.

That sounds reasonable. I mean the 56.64 does have the advantage that it supports all FX 5xxx cards, so there is no need to force the installation. And if the games you play all run perfectly on it, there isn't much point to downgrade to previous versions. 😀

I use the 30.82 driver with my GeForce4 Ti 4200/128MB (Pentium III 1GHz machine), after watching this video: https://www.youtube.com/watch?v=HRhm4aGNI3o
I have previous used the 45.23 or 44 driver version, also faster in games, because the 45.23/44 has much more CPU overhead and is maybe more SSE2? optimized for A64/P4 CPUs...
A comparison between this drivers and CPUs with SSE/SSE2 instruction support would be nice!

bloodem wrote on 2022-07-22, 11:09:

I've heard about this before (the fact that some drivers are apparently faster on processors that support SSE2), but I consider it more of an urban legend than anything else. Anyway, there might be an advantage on Pentium 4 (I only tested one driver version with my Pentium 4 builds, so don't know for sure), however that advantage is most likely due to other factors (maybe some improvements that specifically target the Netburst "hyper-pipelined" architecture?).

Anyway, when it comes to powerful GeForce FX cards, I test them exclusively on Core 2 Duo era machines (or at least Athlon 64 "Venice"), and all of these CPUs have SSE2 (and even SSE3) instructions. And it's on these specific platforms that 45.23 was consistently faster faster for me (which is why I say that the so called "SSE2 optimizations" are most likely misleading).

It depends of the microarchitecture, improvements aren't guaranteed by simply using these instructions on a CPU that has them.

bloodem wrote on 2022-07-22, 11:09:

I've heard about this before (the fact that some drivers are apparently faster on processors that support SSE2), but I consider it more of an urban legend than anything else.

I did find 45.23 drivers (which do support SSE2) to be faster than both 30.82 and 40.72 on my Athlon64 + GeForce4 Ti4200 system.

I can't say with certainty whether it was due to SSE2 or not, but the game benchmark scores are noticeably better. On a side note, 3D Mark scores were higher with older drivers, but actual game benchmarks (both Direct3D and OpenGL) were better with 45.23.