backtop


Print 70 comment(s) - last by Johnmcl7.. on Jul 30 at 6:57 AM


An alleged screen capture of the ATI version of Rydermark (saved and reposted to preserve metadata)

An alleged screen capture of the NVIDIA version of Rydermark (saved and reposted to preserve metadata)

A difference map of the two images
"Rydermark developers" make bold claims which turn out to be nothing more than a Photoshop hoax

It would appear The Inquirer was quick to jump the gun on a story accusing NVIDIA of lying about full DirectX 9 support. The story accused NVIDIA of not allowing developers to use 24-bit or 32-bit shader precision. Instead it claims NVIDIA forces developers into using 16-bit shader precision as the technique is faster. This is a problem as DirectX 9 compliancy requires 24-bit shader precision or better.  "Rydermark" is not a commercially shipping application yet, and has had very little information published to confirm its authenticity.

The original story lacked any type of physical evidence and The Inquirer claimed its sources were developers for the program. Images allegedly proving that NVIDIA forces developers into using 16-bit shader precision were posted on The Inquirer. The posted images compared a rendered scene in Rydermark 2006 between ATI and NVIDIA graphics cards.

It turns out the images "proving" NVIDIA’s wrongdoings were nothing more than poorly done Photoshopped images.  The NVIDIA rendered image appeared to have blurrier water while the ATI rendered image had sharper water detail. However, the ATI rendered image just didn’t look right with poor cut offs and a creation date three minutes after the NVIDIA rendered image.  A difference image of the two JPG files can be seen to the right, with the outline of the modified  area clearly visible in the ATI image.  This would suggest the NVIDIA image was the original source image, and that the ATI version was modified afterwards.

A difference of the metadata from both images reveals that the NativeDigest delimiter is identical for both images, but has two different InstanceIDs.  This would be consistent with an image that was modified and saved twice.  In the author's defense, images that are created and saved on his computer have distinct metadata tags that are very easily identifiable.  These are not present in the two images supplied by The Inquirer for "Rydermark" -- suggesting the images may not have been modified by the author.

There’s been an outcry of The Inquirer images on various forums including Ace’s Hardware, AnandTech and Beyond3D.

Update 07/19/2006: The Inquirer has posted something resembling a rebuttal to this article. Incredibly, a user from the ModTheater.com forums managed to track down some of the stock art used in the screen renders, and believes the entire image is actually fraudulent.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

By Justin Case on 7/20/2006 8:06:45 PM , Rating: 1
Pixel shaders, as the name implies, operate on pixels. Specifically, when applying a pixel shader to selected areas of a pre-rendered image, the shader will affect the areas defined by the mask. That mask can be based on the original material IDs, on a mathematically defined shape, or it can be a simple grayscale mask created using... a brush in a paint program.

If the benchmark author was claiming that the whole water effect was the direct product of a pixel shader, applied to a 3D scene, I'd say "bullshit". But he isn't. For all you or I know, the "pixel shader" might be running on a pre-rendered image (a 2D bitmap), based on a hand-drawn mask.

Do I think the images in question came straight out of a 3D engine? No. In fact, I'm 99.9% sure they didn't. But, AFAIK, the benchmark author never claimed that they had.

Maybe the pixel shader in question is a lot like a sharpen filter (pixel shaders can and do run on screen data, not just mapped textures), and maybe the benchmark author simply drew a basic mask to define where the shader would run.

Without more information about what the benchmark is doing (or what it's supposed to be doing), there's simply no way to tell.

Like several people have pointed out, even a five year old can follow the water's outline better than whoever drew that mask over the image (hell, Photoshop's magnetic lasso will do it automatically for you). So I think it's perfectly possible that what we're seeing as signs of a "fake" are simply signs that the "shader" was something applied to a bitmap, with a quickly-drawn mask, and not to textures mapped onto polygons on a 3D scene.

Until the benchmark author makes it clear what that image is supposed to be, both possibilities remain valid.

And none of this has anything to do with the fundamental issue of the pipeline precision. Even if the images turn out to be mock-ups made by a 4 year old, that still doesn't prove anything about the shader precision, one way or the other.



"So if you want to save the planet, feel free to drive your Hummer. Just avoid the drive thru line at McDonalds." -- Michael Asher











botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki