S3's new antialiasing patent

[Ailuros]

Credit goes to Demirug for making me aware of the patent:

http://v3.espacenet.com/textdoc?DB=EPODOC&IDX=US6828983&F=0&QPN=US6828983

Highly interesting if you ask me

 

[Dave Baumann]

Filed 2000. One of the inventors is "Konstantine I. Iourcha - San Jose CA", I know that there is a Konstantine at ATI's CA office and considering the influx of former S3 employee's to ATI and S3 technology enhancments in recent ATI parts (3Dc from S3TC, "Fast Trilinear" in RV350 onwards) I wonder if its a development of this that has been reference in relation to the new form of AA we've heard.

(Edit: In fact I believe it was Konstantine that was the inventor of S3TC)

 

[Dave Baumann]

I've just done a search for Konstantine Iourcha in the USPTO database, and it turns up a number of patents - one of which is this: Method and apparatus for block based image compression with multiple non-uniform block encodings - the inventors are: "Iourcha, Konstantine; (San Jose, CA) ; Pomianowski, Andrew S.C.; (Palo Alto, CA) ; Koduri, Raja; (Santa Clara, CA). Raja and Andrew are at ATI at the moment, so it looks like Konstanine is as well.

It has, however, been pointed out to me that this could very well be a form of Compressed Multisampling - remember, ATI's colour compression only works with AA enabled.

There is an application that has Eric's name on it which may be more along the lines of the hints that Erice has been dropping: Method and apparatus for sampling on a non-power-of-two pixel grid

"Using the present invention with 6 sample multisampling gives the same visual antialiasing quality as 8 samples using a prior art technique but uses less memory."

[Edit] Actually, this look like its just R300 MSAA as it refers to a 12x12 grid, which is what I'd calculated to sample grid to be for a while. There's also another Eric application that talks about super tiling - it seems that ATI are just very late in filing for patents on much of their R300 technology.

 

[Simon F]

(Edit: In fact I believe it was Konstantine that was the inventor of S3TC)

One, at least, and seems quite a nice chap.

 

[Nick]

Actually, this look like its just R300 MSAA as it refers to a 12x12 grid, which is what I'd calculated to sample grid to be for a while. There's also another Eric application that talks about super tiling - it seems that ATI are just very late in filing for patents on much of their R300 technology.

Yes, I haven't looked at the exact details, but for all I can tell this is classical multisampling. It's a new European patent but maybe it was filed a long time ago in the U.S.A.

 

[Dave Baumann]

It's a new European patent but maybe it was filed a long time ago in the U.S.A.

That's from the USPTO database, which is the US Patent Filing system - filing date was only Feb 2004!

 

[Geo]

Filed 2000. One of the inventors is "Konstantine I. Iourcha - San Jose CA", I know that there is a Konstantine at ATI's CA office and considering the influx of former S3 employee's to ATI and S3 technology enhancments in recent ATI parts (3Dc from S3TC, "Fast Trilinear" in RV350 onwards) I wonder if its a development of this that has been reference in relation to the new form of AA we've heard.

(Edit: In fact I believe it was Konstantine that was the inventor of S3TC)

Heh. I can see this post from world-recognized 3d expert Wavey being offered in evidence by S3's IP attorneys. Or are they broadly cross-licensed?

 

[tEd]

This sounds very similar to what matroxs fragment antialiasing does with addition of the z-edges which fragment AA had problems with at least in the first implementation(was fixed in later hardware wasn't it?)

 

[Xmas]

I thought it was similar to FAA as well. What worries me is that they only mention 2x2 and 4x4 ordered grid sampling for the edges. But considering DeltaChrome's sparse sampling pattern, if they ever come up with an implementation of this patent it will probably feature a better pattern.

 

[Megadrive1988]

would this likely have anything to do with S3's nextgen part, codenamed Destination Films?

 

[tEd]

would this likely have anything to do with S3's nextgen part, codenamed Destination Films?

i'm a little surprised that they didn't implemented it til now i mean it has been filed in 2000. Maybe we'll see it one day but as xmas mentioned hopefully they use a better sample pattern than just og