Robin4 wrote:

I dont see any Geforce III in this test.

I don't have one but i don't think it will make a difference. Geforce2 Ti and Geforce4 Ti4200 perform very close on these slow platforms, a Geforce 3 would be between them but score very similar.

A TNT2 ultra would be interesting as the TNT cards perform well in this case, atleast in 800x600.

Though. It is a P3-500 and only 800x600. These are my 3d-mark99 results with an TNT2 Ultra.
Just to give a hint, to get one or not for benchmarking. Different drivers benchmarked.
The idea of these results were to test different drivers out, not compare cards.

https://dl.dropboxusercontent.com/u/13450020/ … MarkResults.doc

brostenen wrote:

Though. It is a P3-500 and only 800x600. These are my 3d-mark99 results with an TNT2 Ultra. Just to give a hint, to get one or n […]
Show full quote

Though. It is a P3-500 and only 800x600. These are my 3d-mark99 results with an TNT2 Ultra.
Just to give a hint, to get one or not for benchmarking. Different drivers benchmarked.
The idea of these results were to test different drivers out, not compare cards.

https://dl.dropboxusercontent.com/u/13450020/ … MarkResults.doc

That is quite interesting, I hadnt looked for benchmarks yet for comparison but I have built up an AMD Athlon 500MHz and the main card is a TNT2 Ultra and it scores around 4900 marks, I didnt realise the performance difference between the athlon and p3 was that large.

Think that the first Athlons are just better than the P3's when you use certain systems.
Or the chipset are just better for certain cards.

The main problem with early Athlons is poor motherboards and chipsets. The usual story with AMD of the time. Some AGP cards will be problematic on them, and they can be sensitive to PSU quality because of the poor motherboard designs.

The Matrox G400 was neck-to-neck in performance with proper TNT2s and Voodoo3s, but had poor quality drivers at launch (no OpenGL support). After the drivers matured, you're left looking at a clear winner in terms of features and image quality. Most of the time in performance as well. Last time Matrox had a real entrant in the race.

TNT2 and TNT2 Ultra had the muscle, but depending on manufacturer the analog display quality could be ok to bad. Also had to be careful about picking the drivers. On Windows 9x, I remember Detonator 3.6x gave the best performance, especially in 32-bit. 5.xx series also had some good drivers. After that, the scores and performance seemed to generally decrease as the unified driver was optimized for the Geforce lineup.

Voodoo3 had great analog quality (desktop/2D), but lacked in certain features such as 3D rendering in 32-bit color. Unless you had a top of the line rig, you wouldn't care for that tho. People wanted high frame rates and to get those most of use needed to drop to 16-bit color anyway. Voodoo3 also had Glide, which neither TNT2 or the G400 or G400MAX could give us... and the drivers aren't as fickle as the nVidia Detonators.

Not sure if a sub 1 GHz P2/3 system really benefits from having GeForce and newer class video card as the games that could use those generally required more oomph on the CPU-side as well.

No wonder I felt better performance on a celeron 300 66fsb, running G-400 compared to TNT2.
It felt that way. Never got a chance to actually benchmark the machine.
So.... I can not back this up, only saying that UT99 ran at 800x600 and NFS-2K ran at 640x480,
only viewdistance set to max. All other low and only 2 opponents.

Still have to make the chart for the k6-2 450 but i am now benching most of the cards on a p3 1ghz (at 1ghz and 750mhz) and see how they scale. This system i will only do 1024x768.

Grabbed a board that i haven't used before so i could give it a good testrun too. It is the Asus Tuv4x and it behaves very good so far. It is maybe a few percent slower as the intel boards but it has native tualatin support and can be setup with 1,5gb ram where the intel 815 is limited to 512mb. It also has an agp pro slot. It is a VIA board so hopefully the agp slot doesn't give me problems with the cards.

I found it interesting how the Banshee outperforms the Voodoo 3 and 4 by ~10% with the P3-300 at 800x600; by 5% with the P3-300 at 1024x768; and by 3-10% with the P3-450 at 800x600. However, the Banshee falls 20 - 30% behind the Voodoo 3 & 4 with the P3-450 at 1024x768.

What can be inferred from this - that the Banshee is the superior Direct3D 6 card for lower resolutions and for under powered CPUs compared to the Voodoo 3 and Voodoo 4?

Note that the Banshee is AGP while the V3's are PCI. Maybe AGP saves the CPU some cycles since it can transfer the data to the video card faster?
V4 is a new architecture (several limitations fell like 16-bit only and 256x256 textures max), maybe driver overhead is higher with this architecture?

Yes, banshee is surpisingly fast on slower cpu. I don't understand it either as the banshee is only clocked 10% higher than v2 but is way faster here.

Funny thing also is that v4 is slow while it is the fastest on a k6-2 450. De rest of the voodoo's are not so fast but the v4 is, even beating v5. Exactly the same drivers used.

After seeing these results, I think I made a wise decision to place the Voodoo Banshee into a Cyrix MediaGXm-266 system. Although there does seem to be a limit as to how slow of a CPU you can go with for the Banshee to be faster than the Voodoo3. In a 486, the Voodoo3 is faster than the Banshee.

Went thrift store hunting today and found a diamond tnt2 ultra and a medion geforce3 ti200. Will benchmark them too and add the results.

For the record, I have been fiddeling and testing an P5A board with a K6-2-500mhz cpu and 128 megabyte of ram.
For the 3DMark99, I get aprox these results... (I can not remember the actual results, only roughly)
The driver used for nVidia was some version 30 driver. All cards were AGP.
The driver for the AGP, is the ALI-AGP-Version 1.9

Geforce2-MX400 64mb: Somewhere between 2500 and 2600.
TNT2 M64 Pro 32mb: Somewhere between 2600 and 2700.
Matrox G400 32mb: Somewhere between 3100 and 3200.

The Matrox really surprised me big time, as I thought that nVidia cards would beat the hell out of it.
Boy, was I in for a surprise... A really good one, as the matrox is the only one wich is passive cooled.
If I can sort that boot issue that I have out, the board is headed for a new build.

I think you are seeing Matrox's driver optimization of 3DMark99. I suggest testing some games you play.

Is there a benchmark program similar to 3DMark99/2000 with the exception that it uses OpenGL instead of Direct3D?

feipoa wrote:

Is there a benchmark program similar to 3DMark99/2000 with the exception that it uses OpenGL instead of Direct3D?

GL Excess is the only thing that comes to mind.
http://www.glexcess.com

swaaye wrote:

I think you are seeing Matrox's driver optimization of 3DMark99. I suggest testing some games you play.

Don't know about that. Yet games run a little better on G400.
NFS 2000 run smore smooth, and UT99 are running without any lag, wich TNT2 did on the same settings.

TNT2 M64 should be slower than a G400 32MB. GF2 MX400 is another story. But it's hard to say how a K6-2 500 will perform because it's so slow. You really should be using Glide with a Voodoo3.

swaaye wrote:

TNT2 M64 should be slower than a G400 32MB. GF2 MX400 is another story. But it's hard to say how a K6-2 500 will perform because it's so slow. You really should be using Glide with a Voodoo3.

G400 is good enough as of now. 😉