PTAB

IPR2018-00102

MediaTek, Inc. v. ATI Technologies ULC

Log In

1. Case Identification

2. Patent Overview

  • Title: Parallel Pipeline Graphics System
  • Brief Description: The ’506 patent discloses a graphics processing system featuring a back-end comprised of multiple parallel pipelines. Each pipeline processes geometry for a specific screen portion ("tile") and includes a "unified shader," which the patent defines as a single unit capable of performing both pixel color shading and texture address shading.

3. Grounds for Unpatentability

Ground 1: Obviousness over Rubinstein in view of Collodi - Claims 1-9 are obvious over Rubinstein in view of Collodi.

  • Prior Art Relied Upon: Rubinstein (Patent 7,102,646) and Collodi (Application # US 2003/0076320A1). Petitioner noted Rubinstein incorporates by reference Patent 6,697,063 (“Zhu ’063”) and Patent 6,856,320 (“Zhu ’320”).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Rubinstein discloses a graphics processor with nearly all elements of the claimed invention, including a front-end geometry processor and a back-end rendering engine that processes geometry for screen "tiles." The back-end pipeline in Rubinstein includes a scan/z engine, a rasterizer, a shading unit, and a color buffer. However, Rubinstein did not explicitly disclose multiple parallel pipelines or a programmable shader meeting the patent’s definition of "unified." Collodi was argued to supply these missing elements. Collodi disclosed a programmable per-pixel shader that performs both color shading and texture shading operations, thus meeting the ’506 patent’s specific definition of a "unified shader." Furthermore, Collodi explicitly taught that rendering efficiency could be increased by "providing multiple per-pixel processors running in parallel in order to multiply the pixel throughput," directly addressing the parallel pipeline limitation.
    • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine the teachings of Rubinstein and Collodi. Both references are in the same field of graphics processing and disclose complementary components of a graphics pipeline. A POSITA seeking to improve the performance of the system in Rubinstein would have been motivated to implement the known technique of parallel processing and incorporate the improved programmable shader taught by Collodi. The motivation was to achieve the known benefits of increased processing speed and user customization.
    • Expectation of Success: A POSITA would have had a reasonable expectation of success in this combination. The petitioner contended that swapping the shading unit in Rubinstein for the programmable shader in Collodi was a simple substitution, as both references use similar pipeline architectures. Implementing parallel pipelines was a routine and well-understood improvement in the art at the time, with predictable results of increased performance.

Ground 2: Obviousness over Rubinstein, Collodi, and Zatz - Claims 8-9 are obvious over Rubinstein in view of Collodi, and in further view of Zatz.

  • Prior Art Relied Upon: Rubinstein (Patent 7,102,646), Collodi (Application # US 2003/0076320A1), and Zatz (Patent 6,809,732).
  • Core Argument for this Ground:
    • Prior Art Mapping: This ground was presented as an alternative for claims 8 and 9, which require the unified shader to be "operative to loop back to process operations for color shading and/or texture address shading." While Petitioner argued the primary combination taught this, Zatz was introduced to provide more explicit support. Zatz disclosed a programmable shader architecture specifically designed for multipass operations. It featured an explicit "Return Path" that functions as a loopback, allowing the output from a combiner unit to be re-processed by the shader core. This directly taught the loopback functionality required by claim 8.
    • Motivation to Combine: A POSITA, having already combined Rubinstein and Collodi to create a parallel processing system, would have been motivated to incorporate the teachings of Zatz to further improve the shader's capabilities. Multipass rendering was a known technique for creating more complex visual effects. Zatz provided a clear, established method for implementing this functionality via its loopback path. The motivation was to gain the known benefit of improved control and flexibility in shading operations.
    • Expectation of Success: Incorporating the loopback path from Zatz into the combined Rubinstein/Collodi system would have been a predictable modification. The addition was an enhancement to the shader module, a discrete component within the overall pipeline, and would have been expected to successfully enable multipass rendering.

4. Key Claim Construction Positions

  • "Z-Buffer Logic Unit": Petitioner proposed this term be construed as "a logic unit that facilitates visibility testing by comparing depth values." This construction was argued to be consistent with the specification's description of the Z buffer's function.
  • "Hierarchical Z-Interface": Proposed construction was "an interface with a z-buffer logic unit that provides for visibility testing at a coarse level, including, for example, for an entire tile or primitive." This was based on the patent’s disclosure of performing initial Z-tests on coarse-level tiles to quickly reject occluded geometry.
  • "Early Z-interface" / "Late Z-interface": Petitioner argued "early Z" means an interface providing visibility testing prior to shading and texturing, while "late Z" provides testing after shading and texturing. These constructions were central to mapping the double-z buffering scheme of the prior art to the claims.

5. Key Technical Contentions (Beyond Claim Construction)

  • Definition of "Unified Shader": Petitioner emphasized that the ’506 patent provided its own unique definition for "unified shader" as a unit performing both pixel color shading and texture address shading. This was contrasted with the modern industry definition, which typically refers to a shader that can process both vertex and pixel data. This distinction was critical, as the prior art (Collodi) was argued to clearly disclose a shader meeting the patent’s specific, older definition, even if it did not meet the modern one.

6. Relief Requested

  • Petitioner requested the institution of an inter partes review and the cancellation of claims 1-9 of the ’506 patent as unpatentable.