PTAB

IPR2023-01335

Mercedes Benz USA LLC v. Daedalus Prime LLC

Log In

1. Case Identification

2. Patent Overview

  • Title: Dynamically Apportioning a Power Budget for a Multi-Domain Processor
  • Brief Description: The ’833 patent relates to methods for dynamically apportioning a power budget for a processor that includes multiple domains, such as processor cores and graphics domains. The system allocates power at run time based on factors like workload and user-defined policies to manage performance and power consumption.

3. Grounds for Unpatentability

Ground 1: Claims 1-5 and 7 are obvious over Finkelstein.

  • Prior Art Relied Upon: Finkelstein (Application # 2010/0115304).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Finkelstein discloses a processor with multiple cores operating in different "power planes," which correspond to the claimed "domains." Finkelstein’s power management logic allegedly controls the operating frequency and voltage of each power plane independently. Petitioner asserted that Finkelstein’s logic dynamically allocates a shared power budget between these planes at run time, using user-defined preferences (controllable by user-level software) as sharing policy values. Finkelstein's iterative equation for determining the energy budget, which considers the remaining energy from a previous interval (a carried-forward budget), power consumed, and a decay value, was argued to teach the limitations of claim 5. For claim 7, Petitioner contended Finkelstein teaches allocating the "entire budget" to a single power plane performing an intensive workload, which meets the "substantially all" limitation.
    • Motivation to Combine (for §103 grounds): Not applicable (single reference ground). The argument was that Finkelstein's own disclosures render the claims obvious, as implementing its teachings would lead to the claimed invention. A person of ordinary skill in the art (POSITA) would have been motivated to operate cores at independent voltages and frequencies to achieve a greater degree of control and performance.

Ground 2: Claims 13-15 and 17-18 are obvious over Nussbaum.

  • Prior Art Relied Upon: Nussbaum (Application # 2011/0022356).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner contended Nussbaum discloses a System on a Chip (SOC) that maps directly to the system of claim 13. Nussbaum’s SOC includes multiple CPU cores (a first domain), a GPU (a second domain), and a North-Bridge/Memory Controller that serves as the system agent circuitry (a third domain). Petitioner argued Nussbaum teaches that the power to the memory controller remains fixed while power is reallocated between the CPU and GPU domains, thus teaching a fixed power budget for the third domain. The reallocation of "power headroom" between the CPU and GPU based on their "boost sensitivity" (a stored power sharing value) was asserted to teach the claimed power sharing logic. Petitioner also argued that Nussbaum's identification of an application as "GPU-bounded," which triggers throttling of CPU cores to release power margin to the GPU, teaches a preference value favoring the second domain (claim 18).
    • Motivation to Combine (for §103 grounds): Not applicable (single reference ground). Petitioner argued a POSITA would find it obvious to implement Nussbaum's teachings as claimed.

Ground 3: Claims 1-5, 7, 13-15, and 17-18 are obvious over Conroy in view of Finkelstein.

  • Prior Art Relied Upon: Conroy (Application # 2007/0049133), Finkelstein (Application # 2010/0115304).

  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner argued that Conroy discloses a system with multiple subsystems, including a CPU (first domain) and a GPU (second domain), that dynamically redistributes power based on workload. Conroy's system uses a power manager and stored "power distribution tables" to allocate power based on whether the system is in a "CPU-heavy" or "GPU-heavy" state. Petitioner asserted that Conroy teaches all elements of the independent claims but argued a POSITA would look to Finkelstein for specific implementation details.
    • Motivation to Combine (for §103 grounds): A POSITA would combine Conroy and Finkelstein as both address managing power allocation in systems with multiple cores or subsystems. Petitioner argued a POSITA would be motivated to implement Conroy’s general system on a single integrated circuit (IC) chip as taught by Finkelstein to reduce size and weight and improve reliability and power efficiency. Furthermore, a POSITA would have been motivated to add Finkelstein's teaching of user-level software control to Conroy's power distribution tables to provide an additional level of user control over the system.
    • Expectation of Success: A POSITA would have a reasonable expectation of success because combining the references involves routine hardware and software functionalities, and Finkelstein already provides a working example of such an implementation on an IC chip.

  • Additional Grounds: Petitioner asserted additional obviousness challenges, including Claim 7 over Finkelstein in view of Conroy; Claim 16 over Nussbaum in view of Bose; Claims 13-15 and 17-18 over Conroy alone; and Claim 16 over Conroy in view of Bose.

4. Arguments Regarding Discretionary Denial

  • Petitioner argued there is no basis for discretionary denial under §325(d) or §314(a). None of the prior art references asserted in the petition appear in the prosecution history of the ’833 patent.
  • Petitioner further argued that under current USPTO guidance, a concurrent International Trade Commission (ITC) proceeding is not a basis for discretionary denial.
  • The petition also noted it was filed with a motion to conditionally join a pending IPR (IPR2023-00550), arguing this posture makes discretionary denial inappropriate.

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-5, 7, and 13-18 of the ’833 patent as unpatentable.