I think I remember my Voodoo3 PCI behaving the same way in one setup, then working fine in another.
Maybe I updated the bios?

first test, untuned, not played with the settings, here it is:

wing commander II intro

Amazing quality!

leileilol wrote:

I'd like to see VGA2PCI through a Voodoo2 😀 so we can finally have screenshots of it's infamous dac filter.

So here we go:
ftp://78.47.153.163/video/unreal_flyby_60fps.avi
(currently uploading, final size 122.128.050)

This is a grab of Unreal Flyby Intro from Voodoo2 SLI, later in the video timedemo runs to give some framerate impression. CPU is a P3 933MHz.
Video is 640x480 at 60 fps XVID 1.3.2/MP3, Quantization 4.00.

Something about the video grabbing:
I really wanted to grab at 1024x768 at 60 fps but I noticed that my C2Q 2.83 GHz is a bit too slow for this and I just reach lower fps. Unreal Flyby is a scene with a lot of movement and this really needs CPU power. For 800x600 I can reach about 45 fps and 1024x768 about 22 fps. Saving the video uncompressed is also no option, since the amount of data streamed is much too high. For the above video I already reduced the motion search from ultra high precision to very high precision. It is probably not noticeable.
Oh and also deactivate any postprocessing decoder filters that are made for movies. They usually blur all the edges, which will have the video look like recorded in bad quality.

Still I would appreciate good advices or experience for optimizing the encoding more.
Maybe schlang has a more powerful machine and can show some full 60 fps video at 1024 resolution 😀.

edit: The above grab is ofcourse after the Voodoo2s DAC and should in this aspect differ from common video from framebuffer grabs. I haven't done an in detail comparison yet.

unfortunately I only have a voodoo1 in my K6, so I'm unable to show 1024x768 😀

€dit: please add more keyframes to future AVI files 😀

luckybob wrote:

Actually, VGA IS the same as component. I've done it in the past.

HOW? VGA and Component are almost completely different.

VGA sends Red, Green, Blue, Horizontal Sync, and Vertical Sync

Component sends, Difference from Red, Difference from Blue, and Composite Sync. The display or conversion device estimates Green.

So how could they be compatible like that?

I think he was mixing up RGB VGA instead of YUV

Try grabbing with something like huffyuv and re-encode it afterwards, you'll need a lot of space though.

Elianda, could you post your Gravis Ultrasound 2.x driver installs somewhere? I can only find 3.59 and 4.x 😖

mod: And btw, this is an awesome card... 😳
I have to buy one for this christmas...a surprise for myself 😀

In all the mess testing around with different video recording, codec, CPU 32 bit vs 64 bit setting - I somehow uploaded not the final Unreal Flyby video but some video I made while testing. So I removed it from the ftp (this also explains why video stops after 50% of the video and audio goes on...).

Anyway here are some 640x480 captures (still uploading):
ftp://78.47.153.163/video/unreal_640x480.avi
ftp://78.47.153.163/video/unreal_800x600.avi at 'near 60 fps'.

also following captures from:
ftp://78.47.153.163/video/3dmark99max.avi
Demo running in 640x480 16 Bit, no blurred dithering artefacts

ftp://78.47.153.163/video/final_reality.avi
Final Reality 1.0, there is a mode switch after the 2D tests where the card syncs to the new mode after a few seconds.
Walking Mechs, space ship city flight scene

ftp://78.47.153.163/video/xlr8r.avi
XLR8R Profiler from Madonion, Spaceship Valley, Active Kill Switch for Nuclear Reactor, Mirror Cube, Women lifting out of water...
Damn, I like this music from Teque in the second part.

XVID Video Encoder setting are slightly tuned for speed vs quality for the 640x480 vids.

I hope my VGA Multiplier Box arrives here soon, so I can add some lag free human interaction.

edit: I also uploaded a video to youtube and it shows a 480p resolution but the quality reduction is massive. So is the trick to scale the videos up to 720p, to get a useable 640x480 quality?

@DBob see PM.

Depending on the PC, could you use a dual head video card? With VGA + DVI and use a DVI > VGA adapter in clone mode?

Mau1wurf1977 wrote:

Depending on the PC, could you use a dual head video card? With VGA + DVI and use a DVI > VGA adapter in clone mode?

Ofcourse, but usually this is no option for Socket7 and earlier Systems where most of the PCI or ISA cards just have one VGA Out.
I could have used Clone Mode for the Direct3D demo recordings on the Radeon 9800XT, but it wasn't necessary.
The videos should be uploaded by now btw.

how do other users upload HQ vids to youtube? every official music video has very good quality on youtube, even without HD

I read a bit and I think I have to encode in H.264 since this is native youtube format and the expected quality loss is lowest. Until now I used XVID which encodes in MPEG4-ASP and this is probably reencoded.

unfortunately realtime x264 encoding with 60fps is requires quite potent hardware

Yes but x264 should also be more efficient in using multiple CPUs...
Looks like YT can not work with mkv containers that I used to set for 'DOS-Mode' recordings at 720x400 an aspect ratio of 4:3. While the video looks correct, all sound gets converted to random scratches (audio is a usual 128 kBit/s MP3 stream). Not quite convincing...

I have a C2D E8400, we'll see how that goes 😀

Update:
I tried some lossless capturing as recommended by TheLazy1 and used Lagarith encoder. This requires at 1024x768 60fps a HDD that can write ~50 MB/s (or about 30 mins of video 100 GB size).
Then I experimented a bit with x264 encoding and did a 2-pass encoding at a bitrate of 8192 kbps (this bitrate is usually used for blueray HD material at 1920x1080 and my material is at 60 fps). This is ofcourse a bit more work, but I do not need a real high end PC when capturing.
So if you thought the quality was already amazing, check this out:
ftp://78.47.153.163/video/unreal_intro.mp4
(still uploading)
requires H264/AAC decoder, 1024x768 full 60 fps, no framedrops
This is Voodoo2 SLI with VSync disabled. Even the sometimes visible lines due to SLI asynchroneous framedraw are visible in the video material.
It won't get much better.

awesome 😀

So I practiced a bit more with the workflow to get a very good conversion of the captured video and made these:
ftp://78.47.153.163/video/3dmark2000.mp4
in 1024x768 at 60 fps.
I must say I was really impressed by this conversion myself, since all the small flaws the voodoo shows are also visible in the video.
On 3D initialization one frame of the previous framebuffer content is visible before it is cleared (a image from the helicopter scene from a previous test demo run)
Sometimes a bit Tearing due to disabled VSync, 16 Bit Alpha Blending artefacts, sometimes stripes from the 2 V2s rendering asynchroneously.
I would say thats not distinguishable from the real demo run.

and a yet better version of the XL-R8R capture in x264:
ftp://78.47.153.163/video/xlr8r.mp4
640x480 60 fps