mwdmeyer wrote:

Yeah not sure why. Rage 128 was used in Dell P4s a lot. I don't think the output quality is as good, but 32bit performance doesn't drop as much as the TNT2. Been a while, I might not be correct about that.

What is annoying is the names are all over the place. The chip is Rage 128 Pro, but cards go under weird names such as Rage Fury Pro. XPert 2000 Pro and so on 🤣

EDIT: Turns out my Rage 128 Pro is 64 bit memory version, so I'm going with Matrox for now 😀

philscomputerlab wrote:

Newer drivers do not work with the TNT cards, produce a black screen in D3D games.

Wrong, that is another reason to suspect issues with the setup.

philscomputerlab wrote:

Can you test with the same driver and see what results you're getting? Ideally a similar P4 machine.

I am busy, don't expect much. So I tried P4S533-MX, P4 2,4 GHz, 128 megs SDR, TNT2(m64) 32MB 125/150, Expendable 640x480x16, all sw/hw at default settings.
My Tornado TNT2 cannot do AGP 4x and like so many others, SBA seems to be disabled in bios. At this situation it runs bit faster with AGP disabled: +2 fps. M64 does not suffer from this, outdid yours by only 1 fps. So how was your AGP configuration?

Wrong, that is another reason to suspect issues with the setup.

Fine, be like that 😐

Found some Expendable results from Anandtech. This game seems to just run better on the Voodoo3 card. Nothing suspicious...

Source: http://www.anandtech.com/show/391/19

TNT really should work with different (newer) drivers. At very least the FW 53.04 were good enough to run all the games and benchmarks in my test without much problems.
Maybe some kind of incompatibility with SiS chipset? AGP driver not working properly? Also P4 3.06 is not officially supported on your MB... this may cause problems.

Now I see that Voodoo3 gains lot more from faster CPU than TNT2.
I was fighting the AGP stuff and I can only get SBA on by modifying bios. I believe TNT2 cards like mine do not even have keys for AGP 4x, they might have been worried so much about compatibility issues.
I wonder how are you benchmarking Earth 2150?

havli wrote:

TNT really should work with different (newer) drivers. At very least the FW 53.04 were good enough to run all the games and benchmarks in my test without much problems.
Maybe some kind of incompatibility with SiS chipset? AGP driver not working properly? Also P4 3.06 is not officially supported on your MB... this may cause problems.

Everything is working correctly, no incompatibility, no AGP issues. It is what it is. In some games the V3 is faster, no big deal. Voodoo was for speed, Nvidia for 32 bit colours and larger textures.

Q3A from Anandtech. V2 2000 in front of TNT2 Ultra:

My results show the same. In Q3A the V2 2000 in front of TNT2 Ultra:

29.42 is from mid 2002, over two years of improvements since Q3A, the most demanding game. Remember this is a Windows 98 benchmark, not XP.

CPU works fine, tried 2.4 and 2.66 before starting this project, all good.

Putas wrote:

I wonder how are you benchmarking Earth 2150?

It has built-in benchmark.

philscomputerlab wrote:

Everything is working correctly, no incompatibility, no AGP issues.

Well, how did you check it? By now I don't expect any significant changes, just my experience is telling me that this better be sorted out from the start.

philscomputerlab wrote:

Putas wrote:

I wonder how are you benchmarking Earth 2150?

It has built-in benchmark.

The one from setup program? OK, but I doubt it has any bearing on in-game performance.

Nothing I will say will change your opinion! I've given other sources, what else do you want? Heck even your own tests got similar results. Do you think Anandtech had a faulty system? Why do I get matching results? Do your tests show that the TNT2 is faster than V3 in these games?

I think whatever I do I'll just get a stupid response like:

Wrong, that is another reason to suspect issues with the setup.

I wanted configuration details. Looks like there is some misunderstanding between us. Never mind, was just trying to help.

Putas wrote:

I wanted configuration details. Looks like there is some misunderstanding between us. Never mind, was just trying to help.

I will be adding many more cards, so a better picture will emerge. It's just a lot of work, so maybe I'm a bit more protective than usual. I will also add some BIOS shots to the website, but it's all BIOS defaults settings, nothing special.

And also, when I finish all cards, I will use a new system. Pentium III for example. That will give more data to compare. What system would you choose next?

Benchmarking the Matrox G400 MAX at the moment. Very good VGA output quality, reference level.

I too think that there might be something wrong with the TNT2, it should be more competitive I think. I find it hard to believe for example that on Expendable you're hitting a GPU bottleneck with a TNT2 Ultra scoring 53FPS (56FPS on Anandtech's PIII 600) and the Voodoo3 keeps scaling all the way up to 80+ FPS. But, I've never benched TNT2 cards extensively, so I could be in the wrong, just something to take note of. Perhaps Vsync was engaged on that game? It would certainly explain a lot, V-sync back in those days would automatically lock to 30FPS if it dropped below 60FPS.

I would like to see a screenshot for the TNT2U showing the powerstrip AGP settings page (what powerstrip reads from the chipset).
Also the SIS AGP driver version you used would be nice to know.

philscomputerlab wrote:

It's just a lot of work, so maybe I'm a bit more protective than usual. I will also add some BIOS shots to the website, but it's all BIOS defaults settings, nothing special.

I see. Rather than showing bios settings check in OS what is actually applied. Newer drivers really work.

I disable V-sync with PowerStrip. Forsaken shows nice FPS with TNT2, beating the other cards. Expendable is one game that supports high resolution textures. Maybe this affects performance? Anandtech benchmarks in line with mine.

I'll put the TNT2 back in when I'm done with the Matrox, but here are Matrox details:

PowerStrip:

AGP:

Any more ideas?

So I've finished with the Matrox and here are the same shots with the TNT2. I also tried with AGP speed set to 2x instead of 4x in the BIOS, it made no difference I'm afraid...

And these are my TNT2 Pro cards. One from MSI, the other from ELSA. I tried both, no difference...

I have to move onto the next card. So let's remember the TNT2 results and we will check out how it does if / when I get to the next platform. Likely a fast Intel Pentium III machine.

But the good news is that the GeForce 256 DDR does extremely well!

Just look at it go! Same driver as with the TNT2. Here is a sneak peak:

How many resolutions do you test in, Phil? I recognise that testing only with 1024 in my own project will produce a very limited picture but, given that each extra res will have an exponential effect on the amount of time required, I'd like to limit it as much as poss. Cheers!

brassicGamer wrote:

How many resolutions do you test in, Phil? I recognise that testing only with 1024 in my own project will produce a very limited picture but, given that each extra res will have an exponential effect on the amount of time required, I'd like to limit it as much as poss. Cheers!

Check the project link in the first post 😀

Looks good, was your vanilla TNT2 different or did you emulate it via the Pro card?

Putas wrote:

Looks good, was your vanilla TNT2 different or did you emulate it via the Pro card?

I have 3x TNT2 Pro cards, so I just changed the clocks. The M64 is a different card.