`Tel: 571-272-7822
`
`
`
`
`Paper No. 22
`Entered: May 10, 2019
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`AMAZON WEB SERVICES, INC., AMAZON.COM, INC., and
`VADATA, INC.,
`Petitioner,
`
`v.
`
`SAINT REGIS MOHAWK TRIBE,
`Patent Owner.
`____________
`
`Case IPR2019-00103
`Patent 7,149,867 B2
`____________
`
`
`Before KALYAN K. DESHPANDE, JUSTIN T. ARBES,
`and CHRISTA P. ZADO, Administrative Patent Judges.
`
`ZADO, Administrative Patent Judge.
`
`
`
`
`DECISION
`Denying Inter Partes Review
`35 U.S.C. § 314
`
`
`
`
`
`
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`I. INTRODUCTION
`
`A. Overview
`Amazon Web Services, Inc., Amazon.com, Inc., and VADATA, Inc.
`(collectively, “Petitioner”)1 filed a petition requesting inter partes review of
`claims 1, 3–9, and 11–19 (the “challenged claims”) of U.S. Patent No.
`7,149,867 B2 (Ex. 1001, “the ’867 patent”). Paper 1 (“Pet.”). Saint Regis
`Mohawk Tribe (“Patent Owner”)2 filed a Preliminary Response. Paper 20
`(“Prelim. Resp.”).
`35 U.S.C. § 314 provides that an inter partes review must not be
`instituted “unless . . . the information presented in the petition . . . shows that
`there is a reasonable likelihood that the petitioner would prevail with respect
`to at least 1 of the claims challenged in the petition.” 35 U.S.C. § 314(a).
`Upon considering the evidence and arguments presented, we determine the
`Petition does not demonstrate a reasonable likelihood that Petitioner would
`prevail with respect to at least one of the challenged claims.
`Accordingly, we do not institute an inter partes review.
`
`B. Related Proceeding
`The parties advise that the ’867 patent has been subject to, or relates
`to, the following district court proceeding: SRC Labs and Saint Regis
`Mohawk Tribe v. Amazon Web Services, Inc., Amazon.com, Inc., and
`VADATA, Inc., No. 2:18-cv-00317 (W.D. Wash.). Pet. 2; Paper 17, 1.
`
`
`1 Petitioner identifies only itself as real parties-in-interest to the Petition.
`Pet. 1–2.
`2 Patent Owner identifies only itself as a real party-in-interest to this
`proceeding. Paper 17, 1.
`
`2
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`C. The ’867 Patent
`The ’867 patent, titled “System and Method of Enhancing Efficiency
`and Utilization of Memory Bandwidth in Reconfigurable Hardware,”
`generally relates to “implementing explicit memory hierarchies in
`reconfigurable processors that make efficient use of off-board, on-board,
`on-chip storage and available algorithm locality.” Ex. 1001, 1:18–21.
`The ’867 patent explains that microprocessors “enjoyed annual
`performance gains averaging about 50% per year,” wherein most of the
`gains were attributable to higher clock processor speeds, more memory
`bandwidth, and increasing utilization of instruction level parallelism (“ILP”)
`at execution time. Id. at 1:26–30. However, as microprocessor speeds
`increased, challenges arose to designing memory hierarchies that could keep
`up. Id. at 1:31–33. The ’867 patent identifies two measures of the gap
`between microprocessor and memory hierarchy speeds—bandwidth
`efficiency and bandwidth utilization. Id. at 1:35–37. Because potential
`performance gains from using a faster microprocessor were reduced or
`negated by corresponding drops in bandwidth efficiency and bandwidth
`utilization, significant effort had been spent, according to the ’867 patent, on
`development of memory hierarchies that could maintain high bandwidth
`efficiency and utilization. Id. at 1:45–50.
`The ’867 explains that one approach to bridging the gap was the
`utilization of cache memories. Id. at 1:51–53. In designing cache
`memories, a number of considerations had to be taken into account. Id. at
`59–60. For example, for programs that exhibit a high degree of spatial
`locality (i.e., it is likely that other data within the same cache line will be
`needed), wide cache lines are efficient. Id. at 1:64–2:4. However, for
`programs that have low levels of spatial locality, narrow cache lines are
`
`3
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`more efficient. Id. at 2:4–7. The ’867 patent provides additional examples
`of considerations in cache design. Id. at 2:14–3:40. The ’867 patent states
`that the various considerations and tradeoffs made cache design challenging
`for a multipurpose computer that executes a wide variety of programs. Id. at
`3:30–32. Cache designers tried to derive the program behavior of the
`“average” program, and optimize the cache for the “average” program. Id.
`at 3:32–36. As a result, the cache was sub-optimal for most programs,
`because most programs that actually run on the microprocessor are not
`“average.” Id. at 3:36–39.
`Because of the above-discussed issues, there was a growing need,
`according to the ’867 patent, to develop improved memory hierarchies that
`limited overhead of a memory hierarchy without also reducing bandwidth
`efficiency and utilization. Id. at 3:57–60. To address the need, the ’867
`patent describes a system including a memory hierarchy and a
`reconfigurable processor that includes a data prefetch unit. Id. at 4:4–10,
`5:60–62, 6:9–13, 7:34–48. “Unlike conventional static hardware platforms,”
`the memory hierarchy is reconfigurable so that computational demands and
`memory bandwidth can be matched. Id. at 7:17–22. The ’867 patent
`explains:
`An important feature of the present invention is that many
`types of data prefetch units can be defined so that the
`prefetch hardware can be configured to conform to the
`needs of the algorithms currently implemented by the
`computational logic. The specific characteristics of the
`prefetch can be matched with
`the needs of
`the
`computational logic and the format and location of data in
`the memory hierarchy.
`Id. at 7:49–55. The ’867 patent provides an example of configuring the data
`prefetch unit depending on the needs of the computational logic. For
`
`4
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`example, Figures 9A and 9B show an external memory organized into a 128
`byte (16 word) block structure that is optimized for stride 1 access of
`a cache. Id. at 7:56–59. However, the data prefetch unit can be configured
`to extract only 8 bytes of data in the memory block, discarding the
`remaining 120 bytes if only the 8 bytes are needed. Id. at 8:3–11. In another
`example relating to a computational intensive matrix multiplication problem,
`the ’867 patent explains that
`On a conventional microprocessor with static execution
`resources, these loops [representing matrix multiplication]
`would be arranged to give stride-one data access where
`possible and also block or tile these uses to facilitate data
`cache hits on the B and A matrices, which are read many
`times. With the configurable memory hierarchy of the
`present invention, matrix B may be stored in on-board
`BRAM memory 307 and rows of matrix A in registers.
`Id. at 10:33–40.
`
`D. Asserted Grounds of Unpatentability
`Petitioner challenges claims 1, 3–9, and 11–19 of the ’867 patent on
`the following grounds. Pet. 3.
`Reference
`
`
`
`3 Holger Lange & Andreas Koch, “Memory Access Schemes for
`Configurable Processors,” Field-Programmable Logic and Applications:
`The Roadmap to Reconfigurable Computing, 10th International Conference,
`FPL 2000, Villach, Austria, 615–25 (Aug. 27–30, 2000) (Ex. 1003)
`(“Lange”).
`4 Peixin Zhong & Margaret Martonosi, “Using Reconfigurable Hardware to
`Customize Memory Hierarchies,” High-Speed Computing, Digital Signal
`Processing, and Filtering Using Reconfigurable Logic, SPIE—The
`International Society for Optical Engineering, Boston, MA, 237–248 (Nov.
`5
`
`Lange3
`Zhong4
`
`Ground
`§ 103(a)
`§ 103(a)
`
`Claims
`1, 3–9, 11–19
`1, 4, 6, 7, 9
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`Petitioner relies on the declaration of Brad L. Hutchings, Ph.D., to
`support the Petition. Ex. 1002 (“Hutchings Declaration”).
`
`instantiates an
`
`E. Challenged Claims
`Of the challenged claims, claims 1, 9, and 13 are independent.
`Claim 1, reproduced below, is illustrative.
`1. A reconfigurable processor
`that
`algorithm as hardware, comprising:
`a first memory having a first characteristic memory
`bandwidth and/or memory utilization; and
`a data prefetch unit coupled to the memory, wherein the
`data prefetch unit retrieves only computational data
`required by the algorithm from a second memory of
`second characteristic memory bandwidth and/or memory
`utilization and places the retrieved computational data in
`the first memory wherein the data prefetch unit operates
`independent of and in parallel with logic blocks using the
`computational data, and wherein at least the first memory
`and data prefetch unit are configured to conform to needs
`of the algorithm, and the data prefetch unit is configured
`to match format and location of data in the second
`memory.
`
`II. DISCUSSION
`A. Level of Ordinary Skill
`Petitioner asserts that a person of ordinary skill in the art in the field
`of the ’867 patent in the relevant time frame would have had a bachelor’s
`degree in electrical engineering, computer engineering, or a related field,
`with two to three years of experience working with reconfigurable systems.
`Pet. 3 (citing Ex. 1002 ¶ 24). Petitioner asserts that “[w]ith more education,
`
`
`20–21, 1996) (Ex. 1004) (“Zhong”).
`
`6
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`such as additional graduate degrees or study, less experience is needed to
`attain the ordinary level of skill.” Id.
`The Preliminary Response provides no assessment of the level of
`ordinary skill in the art.
`For purposes of this decision and based on the record before us, we
`adopt Petitioner’s assessment of the level of ordinary skill in the art.
`
`B. Claim Construction
`In an inter partes review involving a petition filed before
`November 13, 2018, claim terms in an unexpired patent are given their
`broadest reasonable construction in light of the specification of the patent.
`37 C.F.R. § 42.100(b) (2016). Consistent with this standard, we assign
`claim terms their ordinary and customary meaning, as would be understood
`by one of ordinary skill in the art at the time of the invention, in the context
`of the entire patent disclosure. See In re Translogic Tech., Inc., 504 F.3d
`1249, 1257 (Fed. Cir. 2007). Only those terms that are in controversy need
`be construed, and only to the extent necessary to resolve the controversy.
`See Nidec Motor Corp. v. Zhongshan Broad Ocean Motor Co., 868 F.3d
`1013, 1017 (Fed. Cir. 2017) (citing Vivid Techs., Inc. v. Am. Sci. & Eng’g,
`Inc., 200 F.3d 795, 803 (Fed. Cir. 1999)). We determine that the term “data
`prefetch unit” requires construction.
`Each of the challenged independent claims recites a “data prefetch
`unit.” Claim 1 recites
`a data prefetch unit coupled to the memory, wherein the
`data prefetch unit retrieves only computational data
`required by the algorithm from a second memory of
`second characteristic memory bandwidth and/or memory
`utilization and places the retrieved computational data in
`the first memory wherein the data prefetch unit operates
`independent of and in parallel with logic blocks using the
`7
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`computational data, and wherein at least the first memory
`and data prefetch unit are configured to conform to needs
`of the algorithm, and the data prefetch unit is configured
`to match format and location of data in the second memory
`Ex. 1001, 12:43–54.
`Claim 9 recites one or more reconfigurable processors,
`wherein at least one of the reconfigurable processors
`includes a data prefetch unit to read and write only data
`required for computations by the algorithm between the
`data prefetch unit and the common memory wherein the
`data prefetch unit operates independent of and in parallel
`with logic blocks using the computational data, and
`wherein the data prefetch unit is configured to conform to
`needs of the algorithm and match format and location of
`data in the common memory.
`Id. at 13:17–26.
`Claim 13 recites “transferring data between a memory and a data
`prefetch unit in a reconfigurable processor,” id. at 14:2–3, and further recites
`that the data prefetch unit is
`to needs of an algorithm
`configured
`to conform
`implemented on the computational unit and transfer only
`data necessary for computations by the computational
`unit, and wherein the prefetch unit operates independent
`of and in parallel with the computational unit.
`Id. at 14:6–11.
`Petitioner proposes to construe the term “data prefetch unit” as
`“a functional unit that retrieves computational data needed to complete the
`algorithm instantiated on the reconfigurable processor during processing.”
`Pet. 6–8. Petitioner states that this construction was proposed by Patent
`Owner in the related district court proceeding, and asserts that for purposes
`of the Petition, Patent Owner’s construction should be used. Id. Petitioner
`does not explain why this construction is correct, or provide any arguments
`
`8
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`or evidence to support this claim construction. See generally id.; Prelim.
`Resp. 17.
`Patent Owner responds that Petitioner’s proposed construction is
`incorrect. Prelim. Resp. 16–17. Patent Owner argues that the ’867 patent
`expressly defines the term “data prefetch unit,” and therefore the term
`should be construed in accordance with the express definition provided in
`the patent. Id. Patent Owner asserts that the ’867 patent provides a heading
`labeled “Definitions” in the Detailed Description, and under this heading,
`defines “data prefetch unit.” Id. at 16. Patent Owner argues, therefore, that
`“the patentee has clearly set forth a definition of the disputed term with
`reasonable clarity, deliberateness, and precision.” Id. at 17.
`When the specification of a patent provides a special definition for a
`claim term, even if it differs from the term’s ordinary meaning, then the
`inventor’s lexicography governs. In re Morris, 127 F.3d 1048, 1054 (Fed.
`Cir. 1997) (applying “the broadest reasonable meaning of the words in their
`ordinary usage as they would be understood by one of ordinary skill in the
`art, taking into account whatever enlightenment by way of definitions or
`otherwise that may be afforded by the written description contained in the
`applicant’s specification”); see also Thorner v. Sony Computer Entm’t Am.
`LLC, 669 F.3d 1362, 1365 (Fed. Cir. 2012) (“To act as its own
`lexicographer, a patentee must ‘clearly set forth a definition of the disputed
`claim term,’” and “‘clearly express an intent’ to redefine the term.”).
`We are persuaded that the ’867 patent clearly sets forth a definition
`for the term “data prefetch unit.” The ’867 patent provides an express
`definition for “data prefetch unit” under the heading “Definitions,” thereby
`indicating the patentee intended to accord a special definition to the term.
`Ex. 1001, 5:18, 5:40–43. The definition provides “Data prefetch Unit—is a
`
`9
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`functional unit that moves data between members of a memory hierarchy.
`The movement may be as simple as a copy, or as complex as an indirect
`indexed strided copy into a unit stride memory,” wherein a memory
`hierarchy “is a collection of memories.” Id. at 5:39–43.
`Petitioner’s argument that its construction is the same as that proposed
`by Patent Owner in district court is not persuasive. Pet. 6–8. Petitioner does
`not cite any intrinsic or extrinsic evidence to support its proposed
`construction, much less explain why its construction is correct. See
`generally id. Petitioner has not explained, nor do we discern, a reason to
`deviate from the express definition for “data prefetch unit” provided in the
`’867 patent.
`Therefore, on the record before us, we construe “data prefetch unit” in
`accordance with the definition set forth in the ’867 patent, namely as “a
`functional unit that moves data between members of a memory hierarchy.
`The movement may be as simple as a copy, or as complex as an indirect
`indexed strided copy into a unite stride memory,” wherein a “memory
`hierarchy” is “a collection of memories.” Ex. 1001, 5:39–43.
`
`C. Lange (Ex. 1003)
`Lange, titled “Memory Access Schemes for Configurable Processors,”
`generally describes a scalable, device-independent memory interface that
`supports both irregular access (via configurable caches) and regular access
`(via pre-fetching stream buffers). Ex. 1003, 615. Lange states that “[b]y
`hiding specifics behind a consistent abstract interface, it is suitable as a
`target environment for automatic hardware compilation.” Id. Lange
`explains that reconfigurable compute elements can achieve considerable
`performance gains over standard central processing units (“CPUs”). Id.
`According to Lange, these reconfigurable elements often are combined with
`10
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`a conventional processor, which provides control and I/O services that are
`more efficiently implemented with fixed logic. Id. In combined systems,
`design tools address hardware and software issues separately. Id.
`According to Lange, whereas the level of support for software is suitable,
`the same level of support is not provided for hardware. Id. Lange states that
`it therefore presents a “hardware target” for hardware compilers that is
`analogous to a software target for conventional computers. Id. The
`hardware target is a Memory Architecture for Reconfigurable Computers
`(“MARC”). Id. Figure 4 of Lange, reproduced below, shows an overview
`of the MARC architecture.
`
`
`Id. at 618. Figure 4 shows a MARC core with a caching port interfaced with
`front-end ports (CachePorts and StreamPorts) that interface with User Logic,
`and a streaming port interfaced with back-end ports interfaced with dynamic
`random access memory (DRAM) and n static random access memories
`(SRAMs). Id. The MARC core also interfaces, through a Back-End port,
`
`11
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`with a bus interface unit (“BIU”) to an I/O bus. Id. The MARC core
`includes a First-In First-Out (“FIFO”) buffer and Random Access Memory
`(“RAM”).
`
`D. Zhong (Ex. 1004)
`Zhong, titled “User Reconfigurable Hardware to Customize Memory
`Hierarchies,” generally describes implementing mechanisms like victim
`caches and prefetch buffers in reconfigurable hardware to improve
`application memory behavior. Ex. 1004, 237. Zhong states that
`microprocessor speeds have increased much more quickly than memory
`speeds. Id. As a result, there is a processor-memory performance gap such
`that many significant applications suffer from substantial memory
`bottlenecks, according to Zhong. Id. Zhong explains that typically cache
`memories are used to bridge the performance gap, but that cache memory
`still fails to provide high performance for certain applications. Id. To
`address issues with cache performance, Zhong states that prefetching
`techniques and use of victim caches (e.g., memory for storing data recently
`evicted from cache) may hide some latencies, but that these techniques result
`in waste of transistor space on CPU chips. Id.; see also id. at 239
`(describing victim caches). Zhong proposes to address these issues by using
`programmable logic, such as field-programmable gate arrays (FPGAs), that
`can be reconfigured and customized for different functions during different
`sessions. Id. at 237. Part of Figure 1 of Zhong is reproduced below.
`
`12
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`
`Id. at 239. The portion of Figure 1 reproduced above illustrates a computer
`architecture that includes configurable logic C1 on the same chip as a
`conventional Processor. Id. Applying one possible chip boundary, the chip
`is shown as including a Processor, C1, and an L1 cache, whereas the L2
`cache and additional configurable processor C2 are off-chip. Id. The figure
`also shows an alternative chip boundary, in which the chip also includes the
`L2 cache and C2. Id. In both alternatives, the L2 cache is connected to
`Memory and I/O Bus. Id.
`Zhong also discloses a prefetch buffer “to initiate main memory
`accesses in advance, so that the data will be closer to the processor when
`referenced.” Id. at 240–241. The prefetch buffer comprises several
`independent slots, each of which holds several cache lines of data and works
`like a FIFO buffer. Id. at 241.
`
`13
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`E. Principles of Law
`Section 103(a) forbids issuance of a patent when “the differences
`between the subject matter sought to be patented and the prior art are such
`that the subject matter as a whole would have been obvious at the time the
`invention was made to a person having ordinary skill in the art to which said
`subject matter pertains.” In Graham v. John Deere Co., 383 U.S. 1 (1966),
`the Court set out a framework for applying the statutory language of § 103:
`under § 103, the scope and content of the prior art are to be determined;
`differences between the prior art and the claims at issue are to be
`ascertained; and the level of ordinary skill in the pertinent art resolved.
`The Supreme Court has made clear that we apply “an expansive and
`flexible approach” to the question of obviousness. KSR Int’l Co. v. Teleflex,
`Inc., 550 U.S. 398, 415 (2007). Whether a patent claiming the combination
`of prior art elements would have been obvious is determined by whether the
`improvement is more than the predictable use of prior art elements according
`to their established functions. KSR Int’l Co., 550 U.S. at 417. Reaching this
`conclusion, however, requires more than a mere showing that the prior art
`includes separate references covering each separate limitation in a claim
`under examination. Unigene Labs., Inc. v. Apotex, Inc., 655 F.3d 1352,
`1360 (Fed. Cir. 2011). Rather, obviousness requires the additional showing
`that a person of ordinary skill at the time of the invention would have
`selected and combined those prior art elements in the normal course of
`research and development to yield the claimed invention. Id.
`
`F. Patentability
`As we discussed above, supra Sec. II.B, each of the challenged
`independent claims recites a “data prefetch unit.” Petitioner’s arguments,
`however, are based on a construction that we do not adopt. Petitioner does
`
`14
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`not provide any arguments under the construction set forth above. For the
`reasons discussed below, Petitioner has not demonstrated a reasonable
`likelihood that Petitioner would prevail with respect to at least one of the
`challenged claims.
`
`1. Asserted Obviousness over Lange
`Petitioner asserts that, for purposes of the Petition, “the broadest
`reasonable interpretation of ‘a data prefetch unit’ is ‘a functional unit that
`retrieves computational data needed to complete the algorithm instantiated
`on the reconfigurable processor during processing.’” Pet. 17. However, as
`we discussed above, we interpret “data prefetch unit,” in accordance with the
`definition set forth expressly in the ’867 patent, as “a functional unit that
`moves data between members of a memory hierarchy. The movement may
`be as simple as a copy, or as complex as an indirect indexed strided copy
`into a unit stride memory.” Supra Sec. II.B; see also Ex. 1001, 5:40–43. In
`addition, in accordance with the ’867 patent’s express disclosure, we
`interpret a “memory hierarchy” as “a collection of memories.” Supra Sec.
`II.B; Ex. 1001, 5:39.
`Patent Owner argues that the Petition is based on an erroneous claim
`construction. Prelim. Resp. 22. Patent Owner argues that “[t]he Board is
`not required to ‘play archeologist with the record’ or endeavor to discover a
`challenge that might have been asserted had the Petitioner identified the
`correct claim construction.” Id. at 22 (citing United Microelectronics Corp.
`v. Lone Star Silicon Innovations LLC, Case IPR2017-01513, slip op. at 9
`(PTAB May 22, 2018) (Paper 10)). We agree.
`Applying its proposed construction of “data prefetch unit,” Petitioner
`argues that Lange’s MARC core with its front-end port interfaces
`incorporates data prefetch units. Pet. 17. Petitioner argues that the MARC
`
`15
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`core, when used with the front-end ports, performs the function of
`prefetching the computational data needed to complete the algorithm
`instantiated in Lange’s user logic. Id. at 17–18.
`The Petition, however, does not specify with particularity how Lange
`teaches a memory hierarchy, and moving data between members of a
`memory hierarchy, as required under our interpretation of the term “data
`prefetch unit.” Our rules require that a petition specify with particularity
`where each element of a claim is found in the prior art, and include a
`detailed explanation of the relevance of the prior art to the claim.
`37 C.F.R. § 42.104(b)(4) (“[t]he petition must specify where each element of
`the claim is found in the prior art patents or printed publications relied
`upon”); id. § 42.22(a)(2) (“[e]ach petition . . . must include . . . a detailed
`explanation of the significance of the evidence including material facts”); id.
`§ 42.104(b)(5) (the petition must “identify . . . the relevance of the evidence
`to the challenge raised, including identifying specific portions of the
`evidence that support the challenge”). As the Federal Circuit has explained,
`“[i]n an IPR, the petitioner has the burden from the onset to show with
`particularity why the patent it challenges is unpatentable.” Harmonic Inc. v.
`Avid Tech., Inc., 815 F.3d 1356, 1363 (Fed. Cir. 2016).
`In its contentions regarding independent claims 1, 9, and 13,
`Petitioner does not identify a memory hierarchy in Lange, much less assert
`that Lange teaches moving data between members of a hierarchy. With
`regard to claim 1, Petitioner asserts that Lange discloses a first memory (i.e.,
`either the FIFO memory in the MARC core or BlockSelectRAM in the
`FPGA) and a second memory (i.e., SRAM and/or DRAM accessed by the
`MARC core back-end ports), as recited in the claim, but Petitioner does not
`specify that these memories comprise a memory hierarchy or explain why
`
`16
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`that would be the case. Pet. 15–17 (asserting a first memory); id. at 21–22
`(asserting second memory); see generally id. at 15–22 (failing to specify a
`memory hierarchy). With regard to claim 9, Petitioner identifies a memory,
`asserting that Lange discloses a “common memory” (i.e., DRAM in
`Figure 5), as recited in the claim, but Petitioner does not specify a memory
`hierarchy comprising a collection of memories. Id. at 33–34. With regard to
`claim 13, Petitioner asserts that Lange discloses a “memory” (i.e., the second
`memory of claim 1), as recited in claim 13, but Petitioner does not specify a
`memory hierarchy comprising a collection of memories. Id. at 37–39.
`Similarly, the Petition does not address whether Lange teaches
`moving data between members of a memory hierarchy. See generally id. at
`13–26, 33–34, 37–39.
`By failing to address whether Lange teaches a memory hierarchy, and
`movement of data between members of the memory hierarchy, Petitioner has
`placed the burden on the Board to ascertain how the prior art allegedly reads
`on the challenged claims—a task that we do not undertake. The burden is on
`Petitioner, not the Board, to specify with particularity how Lange teaches a
`memory hierarchy, and moving data between members of a memory
`hierarchy.
`For the reasons stated above, on this record, we are not persuaded that
`Petitioner has demonstrated a reasonable likelihood that it will prevail in
`showing unpatentability of claims 1, 3–9, and 11–19 as obvious over Lange.
`2. Asserted Obviousness over Zhong
`As we discussed above, we interpret “data prefetch unit,” in
`accordance with the definition set forth expressly in the ’867 patent, as “a
`functional unit that moves data between members of a memory hierarchy.
`The movement may be as simple as a copy, or as complex as an indirect
`
`17
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`indexed strided copy into a unit stride memory,” wherein a “memory
`hierarchy” is “a collection of memories.” Supra Sec. II.B; see also
`Ex. 1001, 5:39–43. The Petition, however, applies a different claim
`construction. Pet. 48. As we discussed above with regard to Lange, supra
`Sec. II.F, Patent Owner argues that the Petition is based on an erroneous
`claim construction. Prelim. Resp. 22. We agree with Patent Owner.
`The Petition does not specify with particularity how Zhong teaches a
`memory hierarchy, and moving data between members of a memory
`hierarchy, as required under our interpretation of the term “data prefetch
`unit.” As we discussed above, our rules require that a petition specify with
`particularity where each element of a claim is found in the prior art, and
`include a detailed explanation of the relevance of the prior art to the claim.
`37 C.F.R. §§ 42.104(b)(4), 42.22(a)(2), 42.104(b)(5); see also Harmonic,
`815 F.3d at 1363 (explaining that “[i]n an IPR, the petitioner has the burden
`from the onset to show with particularity why the patent it challenges is
`unpatentable”).
`Applying its proposed construction of “data prefetch unit,” Petitioner
`relies on Zhong’s disclosure of a prefetch generator depicted in Figure 4 of
`Zhong and Zhong’s “prefetching engine” for disclosure of a “data prefetch
`unit.” Pet. 48. However, Petitioner does not identify a memory hierarchy in
`Zhong, much less assert that Zhong teaches moving data between members
`of a hierarchy. With regard to claim 1, Petitioner asserts that Zhong
`discloses a first memory (i.e., prefetch buffers) and a second memory (i.e.,
`main memory or L2 cache), as recited in the claim, but Petitioner does not
`specify that these memories comprise a memory hierarchy or explain why
`that would be the case. Pet. 47 (asserting a first memory); id. at 52
`(asserting second memory); see generally id. at 46–53 (failing to specify a
`
`18
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`memory hierarchy). With regard to claim 9, Petitioner identifies a memory,
`asserting that Zhong discloses a “common memory” (i.e., main memory in
`Zhong Figure 1), as recited in the claim, but Petitioner does not specify a
`memory hierarchy comprising a collection of memories. Id. at 60.
`Similarly, the Petition does not address whether Zhong teaches
`moving data between members of a memory hierarchy. See generally id. at
`44–56, 59–61.
`By failing to address whether Zhong teaches a memory hierarchy, and
`movement of data between members of the memory hierarchy, Petitioner has
`placed the burden on the Board to ascertain how the prior art allegedly reads
`on the challenged claims—a task that we do not undertake. The burden is on
`Petitioner, not the Board, to specify with particularity how Zhong teaches a
`memory hierarchy, and moving data between members of a memory
`hierarchy.
`For the reasons stated above, on this record, we are not persuaded that
`Petitioner has demonstrated a reasonable likelihood that it will prevail in
`showing unpatentability of claims 1, 4, 6, 7, and 9 as obvious over Zhong.
`
`G. Additional Arguments by Patent Owner
`Patent Owner argues that we should exercise our discretion to deny
`the Petition under 35 U.S.C. § 314(a). Prelim. Resp. 5–12. Patent Owner
`also argues that we should deny the Petition for failure to satisfy the
`requirement of 37 C.F.R. § 42.104(b)(3) that the petition set forth how the
`challenged claims are to be construed. Id. at 18–21. Because we deny the
`Petition on other grounds, we need not, and do not, address Patent Owner’s
`arguments regarding § 314(a) and § 42.104(b)(3).
`
`19
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`III. CONCLUSION
`
`For the foregoing reasons, we determine that Petitioner has not
`demonstrated a reasonable likelihood it will prevail in showing
`unpatentability of at least one claim of the ’867 patent. Because Petitioner
`has not satisfied the threshold for institution as to at least one claim, we do
`not institute inter partes review.
`
`IV. ORDER
`
`Accordingly, it is
`ORDERED that the Petition is denied and no trial is instituted.
`
`
`
`
`
`
`20
`
`
`
`IPR2019-00130
`Patent 7,149,867 B2
`
`PETITIONER:
`
`J. David Hadden
`Saina Shamilov
`FENWICK & WEST LLP
`dhadden-ptab@fenwick.com
`sshamilov-ptab@fenwick.com
`
`