FX series are supported in Amithlon!

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

It is because in most cases, it is the last "good" Windows 9x driver in terms of compatibility with older titles.

Those shader performance "improvements" were nVidia cheating to hide the deficiency the GeForce FX has in popular DX9 titles that uses SM2.0, affecting negatively the image quality.

Gmlb256 wrote on 2023-05-28, 22:14:

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

It is because in most cases, it is the last "good" Windows 9x driver in terms of compatibility with older titles.

Those shader performance "improvements" were nVidia cheating to hide the deficiency the GeForce FX has in popular DX9 titles that uses SM2.0, affecting negatively the image quality.

I think those driver versions came after the cheating because in those reviews they already mentioned about the previous cheating of nvidia. Also the reviews mentioned improved image quality. I dont see it impossible that they would have also done actual driver optimization instead of assuming that it was all just cheating. I suppose it is natural to assume that when they had been caught already before but it could also mean that people who just assume that would miss if there was actual improvement later.

https://web.archive.org/web/20071221202438/ht … e/content/88/21

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

From my tests, 53.04 was the fastest driver for my FX 5900XT in Win9x games (DirectX 7 and below), while having slightly lower scores in various 3D Mark versions.

In terms of compatibility, Thief 2 crashes with a black screen on 5x.xx series drivers, although this can apparently be bypassed by tweaking one of the game's .cfg files. However, 45.23 isn't perfect either, as Need for Speed 4: High Stakes crashes with it, while Shogo displays corrupted text in the menu, at least on FX cards. I think 56.64 solved the NFS4 crash, not sure about Shogo.

Joseph_Joestar wrote on 2023-05-28, 16:28:

2mg wrote on 2023-05-28, 15:25:

Slightly offtopic, buy why not skip FX for GF6 series?

Seem to work in W98, still in P4 era, support pallete/fog...

GeForce 6 series dropped support for paletted textures. It only supports table fog.

Also, as others have mentioned, newer drivers that these cards need have compatibility issues with certain games. More info on that here: NVIDIA GeForce FX driver testing on an Intel 440EX summary and report

Really? They are recommended around these parts as fully compatible and solid performes for W98...

2mg wrote on 2023-05-29, 13:20:

Joseph_Joestar wrote on 2023-05-28, 16:28:

GeForce 6 series dropped support for paletted textures. It only supports table fog.

Also, as others have mentioned, newer drivers that these cards need have compatibility issues with certain games. More info on that here: NVIDIA GeForce FX driver testing on an Intel 440EX summary and report

Really? They are recommended around these parts as fully compatible and solid performes for W98...

The missing paletted texture support on GeForce 6 cards has been known since ages. It can easily be confirmed by running Final Fantasy 8 and seeing the card fail the relevant check in the config menu. The missing paletted texture support is also mentioned by one of Nvidia's engineers in the OpenGL EXT_paletted_texture specification.

As for the driver issues, I thought that was common knowledge on this forum as well, but I did include a link to that thread just in case. Basically, the further you move away from 4x.xx drivers, the worse Win9x game compatibility becomes.

Joseph_Joestar wrote on 2023-05-29, 14:18:

Dang, I was really hoping to somehow avoid FX series, both due to loudness and heat, eBayoux prices for 5700+ models, the "FX xx50 versions fixed something so get only those", and crappy DX9 perf.

OTOH I can always plop GF4MX (there are no Tis for PCI right?) into a PCI slot and FX or GF6 into AGP or PCIE, or in my case since I've found those AGP+PCIE4x (but not those faux AGP ala Asrock AGI) mobos and get GF4Ti into AGP and an FX or GF6 into PCIE 😀

GF4 for max compatibility, FX or GF6 for FPS.

PS: sorry for offtopic.

Anything GeForce 2 and up will have excellent performance in Windows 98. I use the lowest tier GeForce 2 MX as my go-to choice for a lot of builds because it's fast, fanless, and relatively cheap.

One thing that a lot of people forget is that their Windows 98 machine may have a much older CPU compared to their video card, and their games will be bound by the CPU. Even a GeForce FX 5200 is more than enough to saturate many low end Windows 98 machines. I even did a test once and found there to be no difference in performance between a FX 5200 and a FX 5950; I don't remember the CPU, but it was a Slot 1. So before you go out and throw the fastest noisemaker into your case, run some benchmarks to see if it even makes a difference.

The GeForce advantage is the on-chip Transform and Lighting acceleration that moves additional operations off of the CPU. So even the lowest end cards yield significant advantage for any Windows 98 era games compared to older accelerators. Even the 3dfx Voodoo 5 5500 lacked TnL acceleration.

Is there a list of early TnL supported games....is it REALLY such a big thing for a retro-gamer?

dr.zeissler wrote on 2023-05-29, 18:40:

is it REALLY such a big thing for a retro-gamer?

For many titles released prior 2000, it isn't a thing since support for hardware T&L debuted with Direct3D 7.

FWIW, nVidia managed to take advantage of that feature (regardless of whether is optimal or not) in their OpenGL ICD due to the high-level nature of the API.

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

It's said a lot, but I haven't seen good evidence for 45.23 being the "best" driver.

There isn't a lot of literature on the topic (that I am aware of). From what I have seen, later drivers have a lot of problems but 56.63 (the driver that supports all FX cards) seems to be just as good (and bad) as any driver that came before it.

Reason people said 45.23 is the best compatibility with games (especially!) and speed, reliable. But this lacks support for later video cards and also using later drivers breaks the game compatibility as well.

Cheers,

mothergoose729 wrote on 2023-05-29, 19:35:

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

It's said a lot, but I haven't seen good evidence for 45.23 being the "best" driver.

There isn't a lot of literature on the topic (that I am aware of). From what I have seen, later drivers have a lot of problems but 56.63 (the driver that supports all FX cards) seems to be just as good (and bad) as any driver that came before it.

I found 56.64 to be the best driver for the FX series in my experience, with the exception of the Thief games. It's also measurably faster on a Pentium 4, most likely due to SSE optimizations that exist for that processor.

Kahenraz wrote on 2023-05-29, 21:20:

mothergoose729 wrote on 2023-05-29, 19:35:

Baoran wrote on 2023-05-28, 21:57:

This thread made me wonder. Is there a good reason why people prefer using detonator 45.23? If I remember correctly nvidia increased the shader performance of FX cards in later driver versions. In reviews detonator 52.16 showed clear shader performance improvement over 4X.XX drivers in FX series cards and it is possible that they continued to improve that even after that version.

It's said a lot, but I haven't seen good evidence for 45.23 being the "best" driver.

There isn't a lot of literature on the topic (that I am aware of). From what I have seen, later drivers have a lot of problems but 56.63 (the driver that supports all FX cards) seems to be just as good (and bad) as any driver that came before it.

I found 56.64 to be the best driver for the FX series in my experience, with the exception of the Thief games. It's also measurably faster on a Pentium 4, most likely due to SSE optimizations that exist for that processor.

Isn't there a patch for that?

http://www.thief-thecircle.com/media/patches/

I had an fx quadro card for a while and consistently saw better performance with 56.64.

pentiumspeed wrote on 2023-05-29, 19:47:

Reason people said 45.23 is the best compatibility with games (especially!) and speed, reliable. But this lacks support for later video cards and also using later drivers breaks the game compatibility as well.

Cheers,

I have heard this line repeated a bunch of times, but never with any details. Which cards and which games?

mothergoose729 wrote on 2023-05-30, 01:05:

I have heard this line repeated a bunch of times, but never with any details. Which cards and which games?

I've asked the same question but never found any convincing evidence of this, with the exception of Thief, which seems to have a workaround.

With a FX5200 in windows XP, 45.23 crashes in 3dmark2001 on game 4, but 66.93 completes them all without issue. Haven't yet tried the same versions on 9x to see if the results are the same

maxtherabbit wrote on 2023-05-30, 03:45:

With a FX5200 in windows XP, 45.23 crashes in 3dmark2001 on game 4

Do you mean the Advanced Pixel Shader test? If that fails, it usually indicates some kind of system stability problem.

For what it's worth, all 3DMark 2001SE tests run fine with 45.23 on my FX 5900XT.

Joseph_Joestar wrote on 2023-05-30, 06:48:

maxtherabbit wrote on 2023-05-30, 03:45:

With a FX5200 in windows XP, 45.23 crashes in 3dmark2001 on game 4

Do you mean the Advanced Pixel Shader test? If that fails, it usually indicates some kind of system stability problem.

For what it's worth, all 3DMark 2001SE tests run fine with 45.23 on my FX 5900XT.

No, the "nature" game. There is no stability issue, I assure you

maxtherabbit wrote on 2023-05-30, 03:45:

With a FX5200 in windows XP, 45.23 crashes in 3dmark2001 on game 4, but 66.93 completes them all without issue. Haven't yet tried the same versions on 9x to see if the results are the same

You cannot compare any if the NVIDIA Windows 9x drivers to Windows 2000/XP drivers. They are very different.