vetz wrote on 2024-01-03, 12:01:

fragmentfi wrote on 2024-01-03, 11:42:

It was the same guy 😁

But good news, he found a floppy which most likely has the drivers. He does not have a floppy drive but he will mail it to me.

With tracking? Would be quite a loss if it goes missing. I also hope you'll upload them to archive.org/vogonsdrivers 😀

Would it not be easier to "drop ship" a USB floppy drive to the guy ?

Assuming this is 1.44MB floppy.

Some great news guys, Petri Nordlund, who wrote the original drivers came through! I now have GLQuake running in NT4! Temper your expectations though as performance doesn't appear to be optimized, Timedemo1 is showing ~10fps on a 1400Mhz Tualatin. It looks fantastic and even supports 32bit modes, takes a hit on performance as expected. I've not been able to get the bus master version of the driver to boot, which I would expect to provide a bit of a boost. It's just really awesome to see this hardware running an actual game. If anyone has any ideas on how to get the Bus Master version of the driver working please let me know, it currently hangs just before the login prompt. Also if anyone has any recommendations on how to properly capture screenshots and videos let me know. I've got an OSSC, thinking I could feed that to say an Elgato?

Here is a terrible quality video: https://youtu.be/e5YXYFB6pTw

Here is the link to the drivers and source code provided by Petri: https://archive.org/details/Pyramid3d-Archive

Keep us posted if you are able to get your card working @fragmentfi!

Great, you got the drivers from him too! He indeed found the files so shipping the floppy wasn't necessary.

He did tell me the zip contains some things that shouldn't be shared though. I was going to make a stripped version of it.

Which version of the driver did you test? "P3DNT" with the separate bus master and non-bus master are old (based on the dates, Sept 1997) while in "p3d" and "p3d_mcd" are identical and are from October 1997. The SDK CD has drivers from June 1997.

We'll test the other card soonish.

Got it, I pulled the files for now. Mika said it was okay but let me double check.

Yes the drivers in the P3DNT folder is what I'm currently using. I'll see if I can give the other drivers a shot!

Also, in the README of the latest dated drivers, there is this mentioned for Quake:

Use "gl_ztrick 0" with GLQuake.

I wonder if it has any effect on performance

Without “gl_ztrick 0” the rendering gets trashed so it’s def required.

I tried the drivers in the p3d zip and NT won’t allow me to install, maybe they are Win95 only? The drivers in the “p3d_mcd” folder must be bus master only as they crash the same as the bus master drivers in the “p3dnt” folder. I’ll try a few more things for sure.

What kind of setup do you have (CPU/chipset)? Do you have chipset drivers installed? In the beyond3d forum Nappe1 said you basically need Intel chipset (440BX/EX) to get bus mastering (DMA transfers) working. You could also try another PCI slot and check that the slot does not share resources with other slots.

I imagine the performance should be way better with bus mastering..

I’m using an Asus P3B-F with BX chipset. I’ll try chipset drivers and moving slots as mentioned! I hope you’re right, would really like to see some playable frame rates.

I tried the chipset drivers found here: http://vogonsdrivers.com/getfile.php?fileid=1424&menustate=0

And I've tried moving slot positions and still the system hangs in the same spot. I'll try a few more things. Maybe another Mobo.

Can you see your revision from somewhere?

It says this in the SDK CD, ERRATA.TXT

1Re: Errata List for Pyramid3D TR25202 revision B.03
2
31. PCI Bus Master 
4Status: There is a compatibility problem in the PCI bus master. Include are 
5the P3D API dll for "On" and "Off " bus mastering.  The setup program will 
6install the non-dma version by default.  You have to manually install the 
7dma version.  Please read README.TXT in SDK\WIN95 for instructions.

Maybe the problem is in the board/chip itself and you cannot get running with DMA..

Yes, it's possible this revision just doesn't work with PCI Bus Master.

According to some documentation (vs203_103.pdf) the default clock is 50MHz. The NT driver changes the clock to 60MHz for rev 2 and later chips. The VLSI branded (they continued the work with different focus after Tritech was out) chips were run at 200MHz and up with active cooling. What I'm saying is the NT driver could be tweaked to run the latter card/chip revisions at much higher speeds.

With the drivers for Win95 (dist20x.zip) you can change the clock and DMA setting from the .inf file. For example by default the TR25202 is set to 73MHz and the TR25204 driver has the clock configured to 90MHz.

We're finally testing my friend's card tonight in my NT machine. It is probably a very early card as it doesn't have any Tritech markings for example.

I'll post photos and results! Currently we have no idea if the card even works.

Good luck!

fragmentfi wrote on 2024-01-22, 11:27:

According to some documentation (vs203_103.pdf) the default clock is 50MHz. The NT driver changes the clock to 60MHz for rev 2 and later chips. The VLSI branded (they continued the work with different focus after Tritech was out) chips were run at 200MHz and up with active cooling. What I'm saying is the NT driver could be tweaked to run the latter card/chip revisions at much higher speeds.

That really sounds like there was a die shrink along the way then right? A 4x increase in clocks would be wild for respins even going from passive to active cooling.

Yes absolutely. There is some info about the plans in this doc: https://vgamuseum.info/images/doc/unreleased/ … resentation.pdf

It says:
• Initial design (sampling 3Q97)
– 100 MHz clock frequency
– 0.6 µm 3 layer metal process
– 13 × 13 mm2 area (est.)
• Production version (1Q98)
– 160 MHz clock frequency
– 0.35 µm 4 layer metal process
– 8.6 × 8.6 mm2 area (est.)

Anyway, good news! The card works! Tested in a P200 WinNT4 machine first. It didn't boot but removed my PCI NIC and it worked. The drivers installed fine and card was detected as TR25202 R1. However that info was visible only if the machine was booted with VGA mode. With regular boot the display changed into static blocky black and white pattern and the machine hung.

I thought the card was dead but decided to try in my P133 Win95 machine. Got a resource conflict error in BIOS. Forced the IRQs and again removed the PCI NIC. Card was detected and the SDK demos worked with some success. Torus seemed to always work, game demo crashed half way. Warrior and Eagle changed display resolution to something my monitor or capture device couldn't handle. Esc quit fine though. My friend left me the card to play with so I'll do more testing later this week. I also have a capture device (Startech USB3HDCAP) so I'll try to get the demos captured.

As the chip is R1, it probably runs nicely only at 50MHz. I'll also add a fan for some extra cooling..

This is the card:

The card has no logos or text. On the other side there is a "11-97" marking, assuming week 11, making it very early revision.

Very cool to see! Also, am I crazy or does that PowerPoint presentation imply quads as a supported primitive?

What an awesome video card, I remember how much I wish I had one to test. How many samples exist more or less?

According to some articles there were "stacks" of boards done and going around.

To me the most interesting part of the card is the programmable pixel pipeline. Pixel shaders were really only available when GeForce 3 was launched in 2001. Not sure yet how flexible in practice it is as it was mainly used only for bump mapping in the demos. They are called as "mixer" in the SDK.

Some more findings..

I got all the SDK demos working. I had used an old (~2004) LCD screen for convenience which has max refresh rate of 75hz. I tested with a CRT and all demos are displayed nicely.

Torus runs at 60 hz
Warrior and Eagle run at 88 hz
Game runs at 73 hz. It still crashes at part 3.

My capture device detects the refresh rate correctly but won't display a picture. It also shows the resolution is 640x385 for Warrior and Eagle. Maybe the resolution is actually 512x384 as it is one of the resolution options in the SDK.

In Windows 95 DirectX shows DirectDraw to work but Direct3D as not being available. At this point I'm not sure if the drivers even support it. I tested with DirectX versions 5.0, 6.1 and 7.0.

Then there probably is no d3d support. Have you tried some DirectX 2/3 games? At least give the D3dtest.exe a go.

Does Final Reality work?