Careless talk costs wives
|
|
|
Microsoft is granted a GPU video encoding patent
SOMEONE at the US Patent and Trademark Office (USPTO) must have screwed up by granting Microsoft a patent on using a graphics processing unit (GPU) to encode video.
The patent, entitled "Accelerated video encoding using a graphics processing unit" was granted to Microsoft yesterday, despite being filed nearly six years ago. Not only is it surprising, to say the least, that a patent would be granted on such a wide ranging and seemingly fundamental process but that it was awarded to Microsoft, of all companies, and not a GPU hardware and driver software design company.
Unlike many patents, this one really does do what it says on the tin. Microsoft describes it as, "A video encoding system uses both a central processing unit (CPU) and a graphics processing unit (GPU) to perform video encoding. The system implements a technique that enables the GPU to perform motion estimation for video encoding. The technique allows the GPU to perform a motion estimation process in parallel with the video encoding process performed by the CPU."
So basically encoding video by using both a CPU and a GPU is what's being touted here as a revolutionary new technology worthy of receiving royalties. It makes you wonder what other software processes can be granted a patent. Perhaps Microsoft has something in the pipeline for GPU accelerated web browsers that it will spring on the competition in order to stay ahead in the browser wars.
The abstract essentially describes what both AMD and Nvidia have been harping on about for years regarding the potential and performance of programmable GPUs. As Microsoft's patent application puts it, "The performance of video encoding using such a system is greatly accelerated as compared to encoding using just the CPU." You'd be forgiven for thinking that perhaps an Nvidia employee wrote the application.
The application goes into further detail on the specifics of encoding and mentions the use of DirectX 9's pixel and vertex shader engines to parallelise pixel encoding. The problem for firms that flog GPU enabled video encoding software is that the wording in the application is vague. For instance, prior to the mention of DirectX 9, the application states, "GPU 121 [a reference GPU] may be configured in a variety of manner." With language like that in a patent you know that sooner or later the lawyers are going to get busy, and rich.
Both AMD and Nvidia have been shipping software that does pretty much what might arguably be covered by the patent that Microsoft has just been awarded. It remains to be seen if the two firms will end up ponying up the cash to use the Vole's now patented technology or whether they will try to invalidate the patent through the courts and challenges at the USPTO.
If challenged, Microsoft will argue that back in October 2004 it was slightly ahead of the game with its patent application, but given the wide-ranging nature and vague terminology of this patent, it seems likely that the Vole will face resistance should the firm actually try to enforce it. µ
Share this:
Share
I am a patent attorney/agent who works with patents on a daily basis.
This article appears to have been written by someone who is unfamiliar with patent law and patent interpretation. The starting point for defining the monopoly defined by a patent is the language of the patent claims, which are the numbered paragraphs at the end of the document. You don't start with the title of the patent, which is often misleadingly broad. Tell us what the claims of the patent say. And then tell us why the specific combination of elements recited in the claims, is not new or obvious, by reference to a similar disclosure in a document or product publicly available before the patent's filing date. Then your criticism will carry some weight.
It is true that some patents are granted which are overly broad; and such patents can be knocked out by opponents in re-examination or litigation proceedings. But the approach taken in this article is simply the wrong approach and fails to demonstrate that Microsoft was not entitled to a patent in this case.
So since we will mix a lot of GPU and CPU in the future does that mean we have to write them all as new patents?
Do you add a new patent when you plant to use a new CPU instruction set in your program?
Ridiculous!
OpenCL is here deal with it.
The main problem here is that MS is patenting the implementation of a well-known algorithm *on a particular architecture*. It's not a patent on a new way of doing motion estimation, just the concept of implementing it on a particular processor. It's about as novel as patenting the implementation of a standaard motion estimation algorithm on a SPARC processor.
Not to mention that SGI (among others) was doing motion estimation exactly like this back in the mid-late 90's. Sure, they called the hardware a "coprocessor" rather than a GPU, but the mechanics of the implementation looks to be identical.
Before C&P Latest on BING now adding Facebook likes to Bings Database, Lets look at Liability of Vague Patent. this Commentos' Is: Pat Pend.
Lets Patent: New Telephone System. Not UnCommon for Patent to be so vague, yet many others have Patented New Telephone Systems, Trying to enforce vague Patent could bring Many Torts, Just as Vague as Original Patent. Whos' To Tell.
Bright Idea Backfires, Costing More $ ever Made. Now from Vole'.new feature rolling out Wednesday will start showing what Facebook friends "like" on the search results page.
On Facebook and sites around the Web, people can click a "like" button to show support or share information with friends. On Bing, if you search for a topic in the news, articles friends have shared on Facebook might appear. Restaurants and movies that friends have "liked" could help you decide what to do on your next date.
Microsoft has been working with Facebook since 2006.
The feature could help distinguish Bing from Google, which only has access to information users make public
vondrashek Inventor
Just because an idea now seems commonplace, doesn't mean it wasn't new when the original patent filing occurred. USPO took SIX YEARS to finally award Microsoft this patent. In 2004, programmable shaders were barely supported in consumer cards, and people weren't even thinking in terms of using the GPU for general processing, or for video encoding. Some had separate chips dedicated for the task, but not using the GPU itself for encoding.
It was only in 2008 that nVidia and ATI even announced any kind of support for encoding on their GPUs. That's great: 4 years after Microsoft's original filing. Late to the party. Neither graphics company had the foresight to patent the idea back in 2004; Microsoft did.
This article was written by someone who doesn't understand patents.
If I developed a new type of head for a hammer, my patent would be titled something like "Impact driven nail installer". That doesn't mean I'm claiming *all* hammers.
This patent covers the one particular method of video encoding in a GPU described by the patent and nothing else.
Claims 1, 18, 26, and 29 specify what the patent covers. They say things like:
"An apparatus for encoding video frames into digital video data comprising: means for collocating a particular video frame into collocated frames to create copies of the frame such that each of the copies of the frame is available for processing in parallel by at least one of multiple channels of a graphics processing unit (GPU), and to map each of a plurality of pixels selected by the GPU to each of a plurality of channels of a texel; means for dividing each of the collocated frames into texel blocks; and means for providing the texel blocks to each channel of the GPU so that the collocated frames can be processed in parallel to obtain motion estimation data associated with the video frame, the processing including offsetting the copies of the frame by a predetermined value, wherein the means for providing is configured to obtain the motion estimation data by the GPU in parallel with encoding the video frames by a central processing unit."
For example, if you have the CPU map the pixels rather than the GPU, you're not violating the patent.
This patent can´t stand any serious challenge. First the term "GPU" is not really defined, it´s just marketing gabberish. Nvidia just can claim that with the new GF100 they have a "Hybrid CPU" and the MS GPU patent can puff off. And second the GPU is nothing else then a coprocessor so all the old coprocessor patents from the 80's should apply, but they are no more valid as of today. It´s not an invention if you take an old concept and just rename the technical terms, but it´s a shame the the patent institut grants such "nonsense" patents
Speculative language should be banned in patent text so the world could, or might be, a better place perhaps? Also, the claims should be ANDed and not ORed, like some patently obvious patent trolls like to do.
"The patent, entitled "Accelerated video encoding using a graphics processing unit" was granted to Microsoft yesterday, despite being filed nearly six years ago."
Before bashing, it is good to know what it is related to. As it was posted in 2004, it might be related to the Xbox gpu (which even if developed by ATI, is owned by MS). As the Xbox was eleased in 2005 and one of its use is as a media centre, it makes sense.
Also, I'm not sure, but did anyone use both CPU and GPU to encode video 6 years ago?
PS - Please, don't stop bashing the USPO, they deserve all you can vent on them...
SHOW ME THE MONEYYYYYYYYYYYY !!!
i'm shocked that this happened. this is the cherry on the cake as far as patents go. USA is going down the drain.
You are just plain stupid
Another disgraceful demonstration of the ignorance of the US Patent Office. Companies will either have to pay up, or risk bankruptcy in expensive court cases to sort out this mess - exactly what the patent system was supposed to stop.