VOGONS


Reply 20 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++
The Serpent Rider wrote:
PhilsComputerLab wrote:

To my eyes these look nicer than the screenshots posted?

Not really. Blocky pattern is still there:

Are we looking at the same screenshots?

This is the one you supplied, which looks horrible. I'm not getting anything even close to that, so not sure what's going on. When did you create your screenshots and what driver and settings did you use? Would be good supplying those to avoid stumbling in the dark?

file.php?id=33525&mode=view

YouTube, Facebook, Website

Reply 21 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++
PhilsComputerLab wrote:

This is the one you supplied, which looks horrible.

I've posted only 1600x1200 screenshots on Q3DM5. Without texture compression.

PhilsComputerLab wrote:

When did you create your screenshots and what driver and settings did you use?

They were made around 2-3 years ago.
You'll get the same result on any GeForce 1-7 and any driver.
Settings are identical to your screenshot.

Last edited by The Serpent Rider on 2017-04-21, 20:34. Edited 1 time in total.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 23 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

And now someone can provide screenshots of various Radeon cards to compare. Because I'am to lazy to do this myself.

Last edited by The Serpent Rider on 2017-04-21, 20:46. Edited 1 time in total.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 26 of 123, by spiroyster

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote:
PhilsComputerLab wrote:

To my eyes these look nicer than the screenshots posted?

Not really, blocky pattern is still there:

I'm with Phil on this one. No block pattern in any of Phil's screen shots. However there seems to be some consistent banding in yours? How do you capture these images? They show decoding artifacts?

Phils screen shots (both sizes) show no banding what so ever for the sky texture.

Interpretation of vision (eyes) can lie, numbers (image data) don't 😀

Open image in Gimp...

Filter -> Edge-Detection -> Edge -> Set Amount to 10.0 -> Select 'Laplace'
bandedgecomp.jpg
Filename
bandedgecomp.jpg
File size
342.35 KiB
Views
1866 views
File comment
Clockwise from top left: Kyro, GeForce, Phil's, Voodooh
File license
Fair use/fair dealing exception

Black regions denote banding, for no banding it should be noise (like Phils images, 1600x1200 and the smaller one). Black region denotes no difference across adjacent pixels (a loss of precision when representing the gradient colour change across that region). Full noise implies each pixel has distinct difference, no patches/regions with the same colour when rendered.

The Serpent Rider wrote:
swaaye wrote:

I'd like to see what the actual texture image looks like.

That would require software rendering, with linear texture filtering no less. VSA100 or modern GPU are pretty much as close as you can get to the reference.

As in the raw texture image data, not rendered. Even software rendering will apply filtering/interpolation. Ain't got Q3 personally but this could probably easily be extracted from the game data files? .pak ... or is that Q2?

Reply 27 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

spiroyster
I dunno what you want to see that way:

PhilsComputerLab wrote:

I'll take a few VGA captures in Quake III.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 28 of 123, by spiroyster

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote:

spiroyster
I dunno what you want to see that way:

PhilsComputerLab wrote:

I'll take a few VGA captures in Quake III.

All your screen shots show banding just as bad as each other, in the case of the Kyro and Geforce they are pretty indistinguishable, and to the naked eye (without zooming in on the image to observe the compression artefacts) I doubt anyone could tell the difference between the Voodoo and Kyro/Geforce.

Phils VGA captures should be WORSE that direct framebuffer grabs (which I'm assuming you provided), due to *possible* signal degradation through the VGA cable, yet image analysis on his screen shots show 0 banding or compression artefacts, so I don't think there are any IQ issues with the Geforce2 GTS (or VGA bandwidth at that res). o.0

While the renderer may have used uncompressed textures, when you saved the resultant-screen shot/frame-grab, you then saved the image in jpg and not at 100% quality, ergo it saved it in a lossy format (reduced quality for compression) and this is the banding issues seen in your images, not bad IQ from any GeForce 1-7.

Of course the size of the images suggest this. Phil's 1600x1200 image is 4.28MB, and your 1600x1200 images are 730-750 KB, so they are obviously compressed and that amount of compression implies a lossy one. Avoid JPG when IQ is important, you can't get better then lossless PNG 😀

Reply 29 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

I can supply a quick DVI capture later today if I find some time! Would be from a Quadro2 Pro however as the GeForce doesn't have DVI.

YouTube, Facebook, Website

Reply 30 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++
spiroyster wrote:

While the renderer may have used uncompressed textures, when you saved the resultant-screen shot/frame-grab, you then saved the image in jpg and not at 100% quality, ergo it saved it in a lossy format (reduced quality for compression) and this is the banding issues seen in your images, not bad IQ from any GeForce 1-7.

That's not an issue.

Attachments

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 31 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

DVI capture of Quadro2 Pro 1600x1200 attached.

Attachments

YouTube, Facebook, Website

Reply 32 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

And another one with a dark cloud?

Attachments

YouTube, Facebook, Website

Reply 33 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++
PhilsComputerLab wrote:

And another one with a dark cloud?

Pretty much same thing:

Quake 3 GeForce edge Phil's DVI capture.png
Filename
Quake 3 GeForce edge Phil's DVI capture.png
File size
462.46 KiB
Views
1781 views
File license
Fair use/fair dealing exception

Looks like Quadro series are plagued as well.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 34 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

I've got your two screenshots open, as well the one I took, and yours (V4 and GeForce) looks noticeably worse than the one I supplied. The banding in yours really stand out and your images appear blockier also. What are your thoughts? Can you see the banding difference?

Could be the way you took the screenshot, this can often be a common issue with screenshots not fully showing what comes out of the graphics card.

YouTube, Facebook, Website

Reply 35 of 123, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++
PhilsComputerLab wrote:

looks noticeably worse than the one I supplied

Another map, different sky texture. Q3DM5 please.
Pattern didn't change though - color transitions are still blocky.

Last edited by The Serpent Rider on 2019-10-14, 01:38. Edited 2 times in total.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 37 of 123, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

Well guys I've supplied my captures, I'm hoping you and others will contribute and shed some light on this.

I haven't made up my mind, but a few things don't add up, so for the time being I'm politely skeptical 🤣

EDIT: Split the thread into its own, I think we derailed the original one enough 😊

YouTube, Facebook, Website

Reply 38 of 123, by appiah4

User metadata
Rank l33t++
Rank
l33t++

I'll try to provide some ATI R200 screenshots tomorrow if I find the time, for comparison's sake. And yes, for IQ comparison JPEGs are a very bad idea, lossless PNG is preferrable, even though most of the time JPEG compression artifacts and texture compression artifacts are distinguishable.

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 39 of 123, by spiroyster

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote:
spiroyster wrote:

While the renderer may have used uncompressed textures, when you saved the resultant-screen shot/frame-grab, you then saved the image in jpg and not at 100% quality, ergo it saved it in a lossy format (reduced quality for compression) and this is the banding issues seen in your images, not bad IQ from any GeForce 1-7.

That's not an issue.

This PNG is the JPG you posted earlier simply saved as a PNG?

It doesn't work like that unfortunately. Saving a 2/3 year old JPG which has already had lossy compression applied as a PNG doesn't bring back all the discarded data from the original encoding. o.0

@Phil

Massive tear! 😀

PhilDVItear.jpg
Filename
PhilDVItear.jpg
File size
1.95 MiB
Views
1715 views
File license
Fair use/fair dealing exception