Advertisement

NVidia launches patent lawsuit

Started by September 10, 2014 09:04 AM
23 comments, last by Ohforf sake 10 years, 1 month ago
So apparently NVidia is sueing Samsung and Qualcom for infringing on NVidias IP.

http://blogs.nvidia.com/blog/2014/09/04/nvidia-launches-patent-suits/

Some of the patents listed in the lawsuit lend heavily on previous work, at least to my understanding. Others are so old, relating to hardware T&L and having been filed more than a decade ago, that I can't see them still being applicable to any modern hardware.

Some of them I can't judge. But I aways thought Qualcomm's tiled rendering architecture was significantly different from the early programmable GPUs, and that nVidia's Tegras were the only chips that actually reused that old technology. Does anyone have more insight into this?

Here are the claims and the non-technical descriptions. I copied them from the ITC complaint:

March 6, 2001
"Transform, Lighting and Rasterization System Embodied on a Single Semiconductor Platform"

The ?488 Patent discloses a graphics pipeline system for graphics processing, the components of which are positioned on a single semiconductor platform. The operations performed on the single semiconductor platform include transforming graphics data from object space to screen space, performing lighting operations on the data, rendering the data, and executing multiple threads of instructions in parallel on a plurality of logic units while transforming or lighting the data.

January 31, 2006
"Single Semiconductor Graphics Platform System and Method with Skinning, Swizzling and Masking Capabilities,"

The ?667 Patent discloses a graphics pipeline system for graphics processing on a single semiconductor platform. The system transforms, lights, and rasterizes graphics data and is adapted to operate in conjunction with a central processing unit. The system is further capable of performing skinning, swizzling, and masking operations involving the graphics data.

May 2, 2006
"Programmable Graphics Processor for Multithreaded Execution of Programs"

The ?685 Patent discloses a unified approach for computer graphics sample processing. Various kinds of computer graphics samples, such as vertex samples and pixel samples, are processed in a graphics processor with at least one multithreaded programmable computation unit capable of processing different sample types. The graphics processor includes a thread control unit capable of assigning samples to available threads based on a priority among the sample types. The thread control unit is also capable of dynamically balancing the number of samples assigned to the threads. The graphics processor is also able to process more than one sample type simultaneously.

March 21, 2006
"Method and Apparatus for Multithreaded Processing of Data in a Programmable Graphics Processor"

The ?913 Patent discloses an approach for computer graphics sample processing. Various kinds of computer graphics samples, such as vertex samples and pixel samples, are processed in a graphics processor with at least one multi-threaded processing unit. The multi-threaded processing unit is capable of processing samples in an order independent of the order in which the samples were received. For example, instructions in a second thread may be scheduled for execution while instructions in a first thread are stalled waiting for source data.

February 24, 2004
"Rendering Pipeline"

The ?063 Patent describes a rendering computer graphics pipeline system that can use screen space tiling (SST) to reduce the memory bandwidth consumed by the rendering system. The computer graphics pipeline system disclosed in the ?063 Patent performs SST efficiently, while avoiding the breaking up of primitives. The ‘063 Patent also describes a rendering pipeline design that efficiently renders visible fragments by decoupling the scan/conversion depth buffer processing from certain rasterization and shading processes. A “scan/z engine” resolves visibility and allows the rest of the rendering pipeline to shade only visible fragments.

April 24, 2007
"System, Method and Article of Manufacture for a Programmable Vertex Processing Model with Instruction Set"

The ’140 patent relates to programmable graphics operations in a hardware graphics accelerator. For example, in a system with a central processing unit and a hardware graphics accelerator, the hardware graphics accelerator is used to perform programmable operations on graphics data. The programmable operations are performed utilizing instructions from a predetermined instruction set.

February 10, 2004
"System, Method and Article of Manufacturer for Shadow Mapping"

The ’372 patent relates to graphics operations of a programmable shader in a graphics pipeline. In particular, shading calculation may be performed through a shader program, including successive operations on a fragment’s color or texture-related operations in which the output of the first shading calculation is saved for use by the second shading calculation and where the calculations include decoupled variables.

So, basically nVidia owns the idea of a GPU.

The graphics mafia want to extort protection money from all GPU vendors, and if you don't pay up, they'll intercept shipments of your devices and cut you off from the marketplace.

Legally, they're perfectly entitled to do this. Personally I'd love to see some anti-patent activism from developers, leading to activism from gamers deciding to boycott nVidia for such immoral behavior. I guess that shows where I stand on the patent debate laugh.png

I guess their absence suggests that this also means that AMD and Intel did pay their protection fees.

Advertisement

NVIDIA basically owns the patents for almost any GPU, even newer versions. This is why AMD and Intel have cross license agreements with NVIDIA. They (NVIDIA) decided to only use 7 patents within their patent portfolio of nearly 7000 patents in this lawsuit/ITC complaint as these are the most obvious ones and easiest ones to defend in court.

It's very obvious that Method and Apparatus for Multithreaded Processing of Data in a Programmable Graphics Processor is a violation.

So, basically nVidia owns the idea of a GPU.

The graphics mafia want to extort protection money from all GPU vendors, and if you don't pay up, they'll intercept shipments of your devices and cut you off from the marketplace.

Legally, they're perfectly entitled to do this. Personally I'd love to see some anti-patent activism from developers, leading to activism from gamers deciding to boycott nVidia for such immoral behavior. I guess that shows where I stand on the patent debate laugh.png

I guess their absence suggests that this also means that AMD and Intel did pay their protection fees.

It looks like you're not in the computer hardware business. NVIDIA has spent billions developing hardware and having patents for that hardware. A cross license agreement is very easy to do and doesn't cost as much money as you would believe. In addition NVIDIA didn't want to go to court. They were talking to Samsung for 2 years before they decided to do so. Samsung is trying to put the blame on Qualcomm of course which is why they think they shouldn't have to pay.

K1 has no design wins; NV sues.
(Samsung were apparently considering a K1 based device but dropped it for *reasons* which might also have triggered this)

I also find it interesting (amusing?) that, in the press release/blog post, they claim to have invented the unified shader when ATi had devices using it shipping while NV were still banging the separate pipelines drum and playing down the whole idea.

K1 has no design wins; NV sues.
(Samsung were apparently considering a K1 based device but dropped it for *reasons* which might also have triggered this)

I also find it interesting (amusing?) that, in the press release/blog post, they claim to have invented the unified shader when ATi had devices using it shipping while NV were still banging the separate pipelines drum and playing down the whole idea.

When NVIDIA started patent negotiations with Samsung 2 years ago the current K1 wasn't even on the table at that time. I believe in the court room at the end of the day Qualcomm will be the big loser here.

Advertisement


I guess their absence suggests that this also means that AMD and Intel did pay their protection fees

Lovely. :)

beautiful. And soon intel will claim to ARM, AMD, Via, Qualcomm, Mips, Dec, Sony etc. to stop their production because they have a patent "electronic integrated chip that performs programmable operations" ?

This nonsense needs to stop. right now.

seriously.


Method and Apparatus for Multithreaded Processing of Data in a Programmable Graphics Processor

~ How long does such a patent remain in force?

beautiful. And soon intel will claim to ARM, AMD, Via, Qualcomm, Mips, Dec, Sony etc. to stop their production because they have a patent "electronic integrated chip that performs programmable operations" ?
This nonsense needs to stop. right now.

seriously.

This has most likely already happened, in the sense that practically all of these companies have cross licensing agreements. It is essentially impossible to develop a modern CPU without Intel's licenses, or a modern GPU eithout NVIDIA's licenses. Most of the time, a a cross license deal will placate both sides without any money actually exchanging hands. Quallcomm either did not want to play ball or NVIDIA wasn't interested in what they had to share, demanding cash instead.

This IS the current state of the computing industry, for basically every company doing any kind of significant semiconductor business.
SlimDX | Ventspace Blog | Twitter | Diverse teams make better games. I am currently hiring capable C++ engine developers in Baltimore, MD.

This topic is closed to new replies.

Advertisement