I also prefer FX one, especially since FX gives good nglide support so we also cover glide games 😀

Don’t know any game that doesn’t work on FX but works on GF4

The main reason over using a GF4 over the FX series are that the latter one are terrible for DX9 stuff except the Aero theme for Windows Vista and 7, and requires later drivers that comes with more CPU overhead (especially on old computers).

Any place where I can read more about this?

You can read the articles about the GeForce FX regarding the performance and image quality on DX9 titles such as Half-Life 2. nVidia also cheated a bit with the drivers doing shader replacement to improve performance at the expense of the visual quality prior replacing them with the GeForce 6 series.

AnandTech did wrote one article related to this in 2004. The article was about the comparison between DX8 and DX9 used in Half-Life 2 and on benchmarks you will see that the performance of the GeForce FX series is lower compared to the Radeons of that era (R300) when DX9 feature set is used. Many people back then were in denial about this and nVidia had to do damage control.

My experience with the FX series is limited. I've only used an FX5600 Ultra but even it was weak compared to a true Geforce 4 ti (4400 in my case). Games from 2000 or older ran pretty much the same, but stuff from 2002 like NOLF2 ran at half the frame rate, dipping into the low 20s on the first level. And it completely chokes in DX9 like Gmlb256 said. The higher tier FX series cards are priced stupidly high on eBay, even higher than the top Geforce 4 cards.

Kahenraz wrote on 2021-08-30, 15:44:

Is there any reason to use a GeForce 4 over an FX series? Having DirectX 9 on tap is convenient.

The reason is cost. Good, top of the line GeForce FX cards can be quite expensive.
A GeForce 4 Ti 4200, on the other hand, can still be purchased for a reasonable price, if you wait and hunt a bit (just bought another one for $20). And most GeForce 4 Ti 4200 cards I've tried can be easily overclocked to Ti 4400 clock speeds, some even beyond that.
Just make sure to buy the AGP 4x version, because the 8X versions are generally less overclockable (and they also need newer drivers). I myself prefer driver version 30.82 with the GeForce 4 Ti AGP 4X cards.

Desomondo wrote on 2021-08-30, 17:02:

My experience with the FX series is limited. I've only used an FX5600 Ultra but even it was weak compared to a true Geforce 4 ti (4400 in my case). Games from 2000 or older ran pretty much the same, but stuff from 2002 like NOLF2 ran at half the frame rate, dipping into the low 20s on the first level. And it completely chokes in DX9 like Gmlb256 said. The higher tier FX series cards are priced stupidly high on eBay, even higher than the top Geforce 4 cards.

I'd love to see some benchmark comparisons with these games.

I haven't done any serious benchmarking before. I've only used Fraps to monitor a few games as I switched out parts until I got the performance I wanted. I can certainly run some benchmarks though. It just might take me a few days to set up everything and run them.

Mods, maybe split off the GeForce4 vs. FX discussion into a separate thread?

Might be better if we kept the focus on paletted textures, table fog and their impact on Win9x gaming here.

Joseph_Joestar wrote on 2021-08-31, 03:15:

Mods, maybe split off the GeForce4 vs. FX discussion into a separate thread?

Might be better if we kept the focus on paletted textures, table fog and their impact on Win9x gaming here.

I agree with this.

Gmlb256 wrote on 2021-08-31, 13:13:

Joseph_Joestar wrote on 2021-08-31, 03:15:

Mods, maybe split off the GeForce4 vs. FX discussion into a separate thread?

Might be better if we kept the focus on paletted textures, table fog and their impact on Win9x gaming here.

I agree with this.

Split off to new thread.

Kahenraz wrote on 2021-08-30, 15:44:

Is there any reason to use a GeForce 4 over an FX series? Having DirectX 9 on tap is convenient.

• The GF4 Ti runs cooler.
• The GF4 Ti are just as fast if not faster than any FX below the 5700.
• The FX gets very expensive above the 5700.
• The FX is crap for DX9 gaming. In many games they cheat(ed) with DX9 effects and they sort of still ran them in DX8 to compete with ATI in framerate. You get the benefit of nGlide though.
• You can use older drivers for the GF4 = less overhead on weaker PCs.
• In my experience the faster (and hotter) FX cards are not particularly reliable. I only had bad luck with them.

The main reason I have FX 5600 is cost and availability. They are a lot cheaper than GF4 Ti. Given that an expensive GF4 Ti could die any time and many sellers offer no warranty at all I prefer my cheap solution. FX 5900 is also very expensive and often sold with no warranty.

When did Nvidia introduce GPU throttling, so that a dead fan didn't also kill the GPU?

Throttling can help with the lifespan of the hardware but also reduces performance if poorly cooled, especially when doing something intensive with it.

I say GeForce FX but only with the higher models.

I've been able to get Quadro and GeForce FX cards in the higher models for less than $100 except for the what I am pretty sure is an engineering sample FX5950 and a PCIe FX5900. Both of those cards were $150.

The first FX5950 I got was off of eBay for a whopping $50, but both of the fans were bad so I swapped a much better aftermarket cooler onto it.

As for GeForce 4 Ti cards, I have gotten multiple of the higher end ones for not more than $50 each, including the way overprice Quadro 4 980XGL

The normal eBay prices for the GeForce 4 and FX cards are just silly.

bloodem wrote on 2021-08-30, 17:06:

Just make sure to buy the AGP 4x version, because the 8X versions are generally less overclockable (and they also need newer drivers). I myself prefer driver version 30.82 with the GeForce 4 Ti AGP 4X cards.

I haven't seen any qualitative benchmarks between NV25 and NV28... NV28 seems like a bait and switch by nVidia... I have seen NV25 trouncing NV28 with antialiasing enabled, but there has to be a benchmark out there that can demonstrate the performance drop better.

You say 30.82 is better than 45.23, why exactly?

Also, are you certain that NV28 will not work with older drivers (even with a hacked .INF ?)

I could do some nv25 vs nv28 benchmarks. Would just need to make sure both cards are clocked the same.

Memory timings also paly a factor. So even if the memory is clocked higher is could end up being slower if the timings are slower.

cyclone3d wrote on 2021-09-01, 01:16:

I could do some nv25 vs nv28 benchmarks. Would just need to make sure both cards are clocked the same.

Memory timings also paly a factor. So even if the memory is clocked higher is could end up being slower if the timings are slower.

There's definitely a difference between the cards, the NV25 has double the transistor count... But nVidia managed to keep mostly the same performance between the two designs.

As far as I'm concerned, different series between NV25 and NV28 are all the same (4200, 4400, 4600), like you say, they're just clocked differently.

One thing I would do to keep it an apples to apples comparison is limit the AGP to 2x or 4x or whatever the NV25 is capable of on the NV28, since NV28 is 8x.

I've also got both cards. I might try a Quake3 Demo001, that usually is able to show differences between different cards pretty well.