While it does jump around. Not as epileptic inducing a true z-fighting.

Scali wrote:

...or perhaps a bug I once had in the windowed mode of my engine, where it didn't resize the offscreen buffer to the exact size of the client area.

.
This sounds like it could be the culprit.

Not my mouse pointer, windowed mode? 😀

Ah got you now, if you stored the interpolated normal at the pixel of each object's surface (at that pixel) in the buffer, after performing the test, could be used to shade the internal surface o.0? luckily GF2 is a bit extreme in terms of what we should be supporting 😀 could probably skip that one and move onto something with a bit more flexaiblity in terms of custom gfx buffer resources....

Good stuff, did not know you could do something like this (let alone get sustained, decent FPS! with it).

spiroyster wrote:

Ah got you now, if you stored the interpolated normal at the pixel of each object's surface (at that pixel) in the buffer, after performing the test, could be used to shade the internal surface o.0? luckily GF2 is a bit extreme in terms of what we should be supporting 😀 could probably skip that one and move onto something with a bit more flexaiblity in terms of custom gfx buffer resources....

Each surface *is* shaded. It's rendered in 2 passes:
1) Mark pixels in stencil/zbuffer
2) Render geometry with proper stencil/ztest

Thing is just that you are looking at the inside of the cylinder, so they are backfaces, hence no shading.
You could ofcourse create inside-out cylinders when you want to shade the inside, by just flipping the vertex normals.

spiroyster wrote:

Good stuff, did not know you could do something like this (let alone get sustained, decent FPS! with it).

Yea, don't tell me "you need shaders" unless you have a really REALLY good reason 😀
DOOM 3 is also a nice example, it does per-pixel lighting even on GF2, by doing a truckload of renderpasses, using the dot3 extension.
I've also done my own per-pixel Blinn-Phong rendering on GF2, as you may have found on FlipCode already: http://www.flipcode.com/archives/01-16-2004.shtml
The good thing about GF2GTS/Ultra was that they had tons of fillrate, and efficient z-buffering, so doing 3-4 renderpasses was not a big deal.
I believe DOOM 3 needed as many as 6 passes in total (also did stencil shadows): https://www.youtube.com/watch?v=pEFWdiWiBa0

Scali wrote:

You could ofcourse create inside-out cylinders when you want to shade the inside, by just flipping the vertex normals.

This is what I was thinking, ofcourse forogt they would be inside out 😀... very little extra overhead for nice diffuse shaded interiors. skip shadows o.0.

Scali wrote:

Yea, don't tell me "you need shaders" unless you have a really REALLY good reason :) DOOM 3 is also a nice example, it does per- […]
Show full quote

Yea, don't tell me "you need shaders" unless you have a really REALLY good reason 😀
DOOM 3 is also a nice example, it does per-pixel lighting even on GF2, by doing a truckload of renderpasses, using the dot3 extension.
I've also done my own per-pixel Blinn-Phong rendering on GF2, as you may have found on FlipCode already: http://www.flipcode.com/archives/01-16-2004.shtml
The good thing about GF2GTS/Ultra was that they had tons of fillrate, and efficient z-buffering, so doing 3-4 renderpasses was not a big deal.
I believe DOOM 3 needed as many as 6 passes in total (also did stencil shadows): https://www.youtube.com/watch?v=pEFWdiWiBa0

It has been quite eye opening in regards to those GF2's. I bet if you scratched the surface of the chip your see handwritten "Project Quadro". 🤣 On my trail of learning with all this, as wise person once said to me in regards to 3D accelration "What you have there is a glorified matrix calculator, use it wisely... yes! i thought, time to make pretty pictures"

Can't say I remember it on flipcode, while that model does impose a bit of familiarity (looks like oe that did the rounds). Did anyone send you a 'modern' graphics card to test o.0

spiroyster wrote:

Can't say I remember it on flipcode, while that model does impose a bit of familiarity (looks like oe that did the rounds). Did anyone send you a 'modern' graphics card to test o.0

The model is of a Ferrari F50, which ATi used in a demo for their Radeon 9700: https://youtu.be/92bIPSh6r6Y
They also included it in some sample code on their website.
So I took their file format and modified their parser to work in my engine.
I later made my own 3dsmax exporter (the problem with normalmapping is getting the whole tangentspace vectors exported into your vertex data, which most standard export formats don't support).
The mesh is actually only half the car, split in the middle, to save memory. You just render the mesh twice, mirroring it the second time to create the full car.

Anonymous Coward wrote:

I finally had time to take a closer look at this thread. I am very disappointed that the "enhanced" version of the Compaq card is slower in DOS than the original!

How do you mean? Q-vision was the enhanced card. It beat everything... and I mean everything, hands down with Doom benchmarking.

Shadow Lord wrote:

Excellent thread here. EISA video cards have been rarities, outside of the Compaq QVision ones that show up all the time. I find the results for the Mach32 a bit surprising. The Mach32 was an excellent DOS card, and I am very happy with my 2MB MCA version. Of course I have an Elsa 4MB and a 4MB? Miro Crystal 32 so not too worried 😉. Thanks again for the excellent info.

Thanks, it took quite a while to accumulate results, glad it's appreciated.

You have an MCA version of Mach 32!? If so, do you have any pictures? It would be interesting to see. Also, if you could post pictures of you EISA cards, maybe over on this thread: -

EISA Graphics / Video Cards

This too would also be appreciated, thanks!

dogchainx wrote:

What EISA cards compared to their ISA counterpart for gaming?

I'm playing around with my EISA Gateway 486 system and enjoying the fiddling with SCSI EISA, but there's no way I'm going to procure an EISA Mach32 anytime soon, so I'm left with ET4000AX ISA, Stealth PRO ISA, Stealth64 ISA.

I might have to do my own benchmarks. 🤣

I have not yet done much testing between both ISA and EISA... this is something in the pipeline. From memory, cards that have the same chipset but are available in both ISA and EISA bus are: -

ATI Mach 32
S3's 805, 911, 924, 928
Artists GPX

and, I remember seeing an EISA Cirrus Logic a while back, can't remember which chipset revision.

EISA ATI Mach 32's are extremely rare... If I see another one, I would probably still jump on it for a spare. I've only ever seen one for sale... the one I bought!

brassicGamer wrote:

I have just acquired an EISA / VLB combo board. Although it came with an EISA SCSI controller, I have no other EISA cards. I've got about 3 VLB cards and a selection of ISA ones too so, as soon I can hunt down some EISA equivalents, I'll be able to contribute to this cool project 😀

Thanks for the comments, It would be great if you (...and others) were to contribute at some point in any way you can and would be much appreciated.

Look forward to any benchmark data you may have.

386_junkie wrote:

How do you mean? Q-vision was the enhanced card. It beat everything... and I mean everything, hands down with Doom benchmarking.

Yes, but it gets owned in everything else. 😀

386_junkie wrote:

You have an MCA version of Mach 32!? If so, do you have any pictures? It would be interesting to see. Also, if you could post pictures of you EISA cards, maybe over on this thread: -
EISA Graphics / Video Cards

It might take me a while but I will do my best to get them there. The cards are installed in systems that are hard to get too at the moment,

386_junkie wrote:

I have not yet done much testing between both ISA and EISA... this is something in the pipeline. From memory, cards that have th […]
Show full quote

I have not yet done much testing between both ISA and EISA... this is something in the pipeline. From memory, cards that have the same chipset but are available in both ISA and EISA bus are: -

ATI Mach 32
S3's 805, 911, 924, 928
Artists GPX

and, I remember seeing an EISA Cirrus Logic a while back, can't remember which chipset revision.

Well, the best cards for testing IMHO would be the Elsa cards. You can just flip them over an install them in an ISA slot. It is about as close to a perfect twin comparison study you can get to because you are using the same chipset, memory, even PCB just different buses.

I wanted a Mach32 EISA for completeness sake until I saw your bench marks. I think even though the ELSA is a bit slower the higher resolutions (due to 4MB of memory) more then makeup for it!

Shadow Lord wrote:

386_junkie wrote:

I have not yet done much testing between both ISA and EISA... this is something in the pipeline. From memory, cards that have th […]
Show full quote

I have not yet done much testing between both ISA and EISA... this is something in the pipeline. From memory, cards that have the same chipset but are available in both ISA and EISA bus are: -

ATI Mach 32
S3's 805, 911, 924, 928
Artists GPX

and, I remember seeing an EISA Cirrus Logic a while back, can't remember which chipset revision.

Well, the best cards for testing IMHO would be the Elsa cards. You can just flip them over an install them in an ISA slot. It is about as close to a perfect twin comparison study you can get to because you are using the same chipset, memory, even PCB just different buses.

I wanted a Mach32 EISA for completeness sake until I saw your bench marks. I think even though the ELSA is a bit slower the higher resolutions (due to 4MB of memory) more then makeup for it!

IMHO I agree with you, but unfortunately the problem is not having a lack of ISA/EISA hardware... it comes down to time, constraints, and the lack of hours that it takes to do this. For the last month i've not even been in the same country... all my hardware is in the UK, whilst i've been in the US, working!

"until I saw your bench marks" ... What? half of them you mean... there were not even any Windows benchmarks! The results for the Mach 32 was on one OS/system and without driver support i.e. gimped performance. Trying the card with a different OS/system... and with drivers puts it pretty much on top. I thought you would know this!?

386_junkie wrote:

Shadow Lord wrote:

"until I saw your bench marks" ... What? half of them you mean... there were not even any Windows benchmarks! The results for the Mach 32 was on one OS/system and without driver support i.e. gimped performance. Trying the card with a different OS/system... and with drivers puts it pretty much on top. I thought you would know this!?

Huh? What drivers? There are no drivers for DOS so not sure why it is gimped? You have the results for DOS performance based on DOOM and 3DBench. The ATI card is marginally faster then the ELSA card based on the numbers. Its not like there is a 10 fps difference between the cards. Now the ATI w/ proper drivers maybe faster then the ELSA card in windows as well BUT I doubt it will be significant or more importantly even matter in Windows 3.x environments. However, the fact that you have 4MB of VRAM on the ELSA card (not your model of course) means higher resolution and color depth. That is a difference you can see and appreciate even in Win 3.x. Basically the ATI card did not blow things out of the water which is what I had expected.

Disclaimer: Part of the issue could be your benchmarking - i.e. in a more complex game (say Quake) the differences may become more pronounced. Although since both cards are 2D accelerators (i.e. no OpenGL HW) they may be more CPU dependent then video dependent. At the end of the day I think either of the cards would work fine in everyday use and DOS game play but you can drive a higher res monitor with the 4MB Elsa.

Shadow Lord wrote:

Huh? What drivers? There are no drivers for DOS so not sure why it is gimped? You have the results for DOS performance based on DOOM and 3DBench. The ATI card is marginally faster then the ELSA card based on the numbers. Basically the ATI card did not blow things out of the water which is what I had expected.

I didn't realise you were being DOS dependent, the Mach 32 is a Windows accelerator and was not really designed I think for DOS in mind. I personally would not use a Mach card for DOS, there are of course better alternatives. The ATI in DOS was never going to blow things out of the water, but you are right that there is not much between it and the ELSA.

Once I get the chance, I will record and post more results in a different system and under Windows, with ISA equivalents.

Shadow Lord wrote:

Disclaimer: Part of the issue could be your benchmarking - i.e. in a more complex game (say Quake) the differences may become more pronounced. Although since both cards are 2D accelerators (i.e. no OpenGL HW) they may be more CPU dependent then video dependent. At the end of the day I think either of the cards would work fine in everyday use and DOS game play but you can drive a higher res monitor with the 4MB Elsa.

I wouldn't disagree, though under Windows the ATI does make a difference.

Have you done any benchmarking?

386_junkie wrote:

I didn't realise you were being DOS dependent, the Mach 32 is a Windows accelerator and was not really designed I think for DOS in mind. I personally would not use a Mach card for DOS, there are of course better alternatives. The ATI in DOS was never going to blow things out of the water, but you are right that there is not much between it and the ELSA.

Well with a 386 (or even a 486) EISA system you are pretty much limited to DOS and Win 3.x. You certainly are not gaming in Win 3.x so gaming is going to be DOS based. the "acceleration" for Windows 3.x while nice is not really essential specially now. Unless you are doing cutting edge work (for 1994) on your Win 3.x system you are mostly just goofing off on it. The most any of these cards will do is accelerate window dragging and will allow for full windows vs. outline dragging 😉.

386_junkie wrote:

Have you done any benchmarking?

Unfortunately not formally enough. I keep meaning too but real life gets in the way... 😢

I fired up Doom on my 486 DX50 with EISA Mach32 and the performance is dismal. Single figure frame rate. I don’t remember doom being so slow on a 486 of this sort of speed.
I wonder, is this due to the slowish 486 itself or due to the graphics card? I was hoping for better to be honest, never having had en EISA system before