`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`APPLE INC., HTC CORPORATION, HTC AMERICA, INC., SAMSUNG
`ELECTRONICS CO. LTD, SAMSUNG ELECTRONICS AMERICA, INC., AND
`AMAZON.COM, INC.,
`Petitioners,
`
`v.
`
`MEMORY INTEGRITY, LLC,
`Patent Owner.
`____________
`
`Case IPR2015-00163
`Patent 7,296,121
`____________
`
`
`
`
`
`PETITIONERS’ REPLY TO PATENT OWNER RESPONSE
`PURSUANT TO 37 C.F.R. § 42.23
`
`
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`
`
`
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`TABLE OF CONTENTS
`
`I.
`
`INTRODUCTION .............................................................................................. 1
`
`II. CLAIM CONSTRUCTIONS ............................................................................. 1
`
`A. “States” ........................................................................................................ 2
`
`B. “Programmed” ............................................................................................. 5
`
`III. THE INSTITUTED GROUNDS ANTICIPATE OR RENDER OBVIOUS
`CLAIMS 4–6, 11, 12, AND 19–24 ............................................................................ 8
`
`A. Koster Discloses the Claimed “States” ........................................................ 8
`1. MI’s Focus on Koster’s Tags is Inapposite ................................................ 9
`2. Koster’s Snoop Filter Filters Requests Based on State Information
`Contained in the Local State Memory ............................................................. 11
`
`B. Koster Discloses that Each of its Processing Nodes is Programmed in the
`Manner Recited in Claim 11 ............................................................................ 13
`
`C. Koster Discloses the Temporary Storage Recited in Claim 12 ................. 15
`
`D. Koster in Combination with Smith Discloses Claims 19-24 ..................... 19
`
`IV. CONCLUSION ................................................................................................ 20
`
`
`
`
`
`
`
`1
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`
`
`
`
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`EXHIBITS
`
`APPL-1001
`
`APPL-1002
`
`APPL-1003
`
`APPL-1004
`
`APPL-1005
`
`U.S. Patent Number 7,296,121 to Morton et al. (“the ‘121
`Patent”)
`
`Excerpts from the Prosecution History of the ‘121 Patent (“the
`Prosecution History”)
`
`U.S. Patent Application Publication Number 2002/0053004 to
`Pong (“Pong”)
`
`David Chaiken et al., “Directory-Based Cache Coherence in
`Large-Scale Multiprocessors,” Computer vol. 24, issue 9 (Jun
`1990) (“Chaiken”)
`
`Daniel Lenoski et al., “The Directory-Based Cache Coherence
`Protocol for the DASH Multiprocessor,” ISCA ‘90 Proceedings
`of the 17th annual international symposium on Computer
`Architecture, pp. 148-159 (May 1990) (“Stanford DASH”)
`
`APPL-1006
`
`U.S. Patent Number 6,490,661 to Keller et al (“Keller”)
`
`APPL-1007
`
`APPL-1008
`
`Excerpts from Jose Duato et al., INTERCONNECTION NETWORKS
`– AN ENGINEERING APPROACH (1997) (“Duato”)
`
`Michael John Sebastian Smith, APPLICATION-SPECIFIC
`INTEGRATED CIRCUITS (1997) (“Smith”)
`
`APPL-1009
`
`U.S. Patent No. 7,698,509 to Koster et al. (“Koster”)
`
`APPL-1010
`
`U.S. Patent No. 7,315,919 to O’Krafka et al. (“O’Krafka”)
`
`APPL-1011
`
`U.S. Patent No. 6,338,122 to Baumgartner et al.
`(“Baumgartner”)
`
`
`
`2
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`APPL-1012
`
`
`
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`Anant Agarwal et al., “An Evaluation of Directory Schemes for
`Cache Coherence,” Conference Proceedings of 15th Annual
`International Symposium on Computer Architecture (1988)
`
`APPL-1013
`
`Louis G. Johnson, “Multiprocessors,” ECEN 6253 Lecture
`Notes (April 28, 2003)
`
`APPL-1014
`
`Declaration of Dr. Robert Horst
`
`APPL-1015
`
`APPL-1016
`
`APPL-1017
`
`APPL-1018
`
`APPL-1019
`
`Excerpts from Merriam-Webster's Collegiate Dictionary - 10th
`Ed. (2001)
`
`Redacted Letter of March 28, 2014 from Memory Integrity’s
`Counsel to Samsung’s Counsel in Memory Integrity LLC v.
`Samsung Electronics Co., Ltd. et al., Case No. 1:13-cv-01808-
`GMS, including “Response to Samsung’s Allegation of a Rule
`11 Violation”
`
`Luca Benini and Giovanni De Micheli, “Networks on chips: a
`new SoC paradigm,” Computer vol. 35, issue 1 (Jan. 2002)
`(“Benini”)
`
`“HyperTransport™ Technology I/O Link - A High-Bandwidth
`I/O Architecture” (Jul. 20, 2001) (“HyperTransport”)
`
`U.S. Publication No. 2005/0228952 to Mayhew et al.
`(“Mayhew”)
`
`APPL-1020
`
`U.S. Patent No. 6,662,277 to Gaither (“Gaither”)
`
`APPL-1021
`
`U.S. Patent Application Serial No. 10/966,161, as filed
`
`APPL-1022
`
`U.S. Patent Application Serial No. 10/288,347, as filed
`
`APPL-1023
`
`U.S. Patent No. 7,003,633 to Glasco (“Glasco”)
`
`APPL-1024
`
`
`
`
`Affidavit of Mr. Michael Rueckheim in Support of Petitioner
`Apple Inc.’s Motion for Pro Hac Vice Admission
`
`
`
`3
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`
`
`
`
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`APPL-1025
`
`Reply Declaration of Dr. Robert Horst
`
`APPL-1026
`
`APPL-1027
`
`Deposition Transcript of Dr. Vojin G. Oklobdzija Vol. 1, November
`23, 2015
`Deposition Transcript of Dr. Vojin G. Oklobdzija Vol. 2, November
`23, 2015
`
`APPL-1028
`
`David E. Culler et al., Parallel Computer Architecture: A
`Hardware/software Approach (1st Ed.) (1998)
`
`APPL-1029
`
`APPL-1030
`
`
`
`
`
`"InfiniBand Architecture Specification Volume 1 Release 1.0.a"
`(June 19, 2001)
`James Laudon and Daniel Lenoski, Proceedings of the 24th Annual
`International Symposium on Computer Architecture, "The SGI
`Origin: A ccNUMA Highly Scalable Server" (1997)
`
`
`4
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`I.
`
`Introduction
`Petitioners submit this Reply to Memory Integrity’s (“MI”) Response (Paper
`
`
`
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`31). MI relies upon improper claim construction proposals that have already been
`
`considered and rejected in the Board’s Institution Decision (Paper 18). MI’s
`
`proposals ignore the actual claim language and improperly seek to narrow the
`
`broadest reasonable construction of the terms without support.
`
`MI’s validity arguments are similarly deficient. MI inconsistently argues
`
`that the use of the term “state” has a specific meaning in the field of cache
`
`coherency (PO Resp. at 5), while, at the same time, argues that the Board should
`
`ignore Koster’s (a reference that MI’s expert admitted was written for the cache-
`
`coherency field) use of the term “local state memory” (PO Resp. at 23). MI’s
`
`validity arguments are also highly attenuated and reflect a flawed understanding of
`
`the Koster reference. Indeed, MI’s expert admitted that he is unfamiliar with the
`
`relevant technical features of Koster’s “Infiniband” embodiment, an embodiment
`
`which moots MI’s validity arguments with respect to the “temporary storage”
`
`limitation. Moreover, MI’s expert declaration does not provide support as it fails
`
`to identify any legal understanding applied in rendering a patentability opinion.
`
`II. Claim Constructions
`In an effort to avoid Koster’s anticipating disclosure, MI “engages in a post
`
`hoc attempt to redefine the claimed invention by impermissibly incorporating
`
`language appearing in the specification into the claims.” In re Paulson, 30 F.3d
`1
`
`
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`1475, 1480 (Fed. Cir. 1994). MI’s proposals should be rejected as there is no clear
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`definition, in the ’121 Patent or elsewhere, that warrants narrowing the terms.
`
`A. “States”
`MI’s proposed construction improperly seeks to add the very limitation that
`
`MI is trying to add with its Motion to Amend. Motion to Amend at 2 (seeking to
`
`add “wherein said states comprise cache coherency states of a cache coherence
`
`protocol” to substitute claims). This attempt by MI belies its argument that
`
`“states” is already limited to “cache coherence protocol states.” MI should not be
`
`allowed to use claim construction to add claim limitations without amendment.
`
`Further, the Board has already considered intrinsic and extrinsic evidence
`
`and found that the term “states … is not limited to cache coherence protocol states
`
`and is broad enough to include the condition of presence—i.e., what is stored in
`
`cache memory.” Institution Decision, pp. 11-12. MI effectively repeats its earlier
`
`arguments, essentially citing to the same disclosure within the ’121 specification,
`
`and has presented no new evidence to diminish the Board’s preliminary findings.
`
`Additionally, MI’s proposal contains the word “state” that it seeks to define,
`
`exposing MI’s attempt to narrow the broadest reasonable interpretation of this
`
`term.
`
`“[T]he PTO should only limit the claim based on the specification or
`
`prosecution history when those sources expressly disclaim the broader definition.”
`
`
`
`2
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`In re Bigio, 381 F.3d 1320, 1325 (Fed. Cir. 2004). Here, none of the passages
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`cited by MI amount to an express disclaimer. To the contrary, as noted in the
`
`Institution Decision, all of the examples in the specification to which MI (again)
`
`points are couched in broad language stating that “particular implementations may
`
`use a different set of states” and “[t]he techniques of the present invention can be
`
`used with a variety of different possible memory line states.” Institution Decision,
`
`p. 11.
`
`Moreover, the claims at issue recite “states associated with selected ones of
`
`the cache memories.” This recital is broader than the individual “memory line”
`
`states described in each of the ’121 Patent passages quoted by MI. PO Resp., pp.
`
`6-7 (“the cache state for a specified line”; “memory line state information”; and
`
`“the memory line states are modified, owned, shared, and invalid”) (emphasis
`
`added). MI does not and cannot explain why the meaning of states for individual
`
`“line” content (these passages), which may be present in a variety of states, should
`
`be the same as the state of the whole cache memories (as claimed). Indeed, MI’s
`
`expert admitted that MI’s proposal (which he did not help in developing) “would
`
`not apply” when referring to entire “cache memories.” Ex. 1026, 113:15-114:13;
`
`112:13-16; 16:15-17:13.
`
`Additionally, the dependent claims reveal that the patentee used express
`
`language when seeking to limit a term to the context of cache coherency, and it
`
`
`
`3
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`clearly omitted such language with respect to the term “states.” In particular, claim
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`3, which depends indirectly from independent claim 1, recites that the probe
`
`filtering unit corresponds to an additional node that “comprises a cache coherence
`
`controller.” Claim 3 demonstrates that, where MI intended for a claim term to be
`
`limited to the context of cache coherency, MI explicitly used the term “cache
`
`coherence” in the claim language. The omission of “cache coherence” from claim
`
`1 is telling, and, under the broadest reasonable interpretation standard should be
`
`read in contrast with claim 3 to preserve the breadth of claim 1.
`
`MI also relies upon extrinsic evidence as supportive of its position, citing to
`
`two references by Daniel Sorin, one of which was published nearly nine years after
`
`the priority date sought by MI. PO Resp., pp. 4-5. However, “the fact that [MI]
`
`can point to definitions or usages that conform to their interpretation does not make
`
`the PTO’s definition unreasonable when the PTO can point to other sources that
`
`support its interpretation.” In re Morris, 127 F.3d 1048, 1056 (Fed. Cir. 1997).
`
`Here, the Board cited to the Microsoft Computer Dictionary for support of its
`
`construction. Institution Decision, p. 12. MI’s criticism of this source is hardly
`
`credible as MI endorses the use of the same Microsoft Computer Dictionary, as
`
`well as a general purpose dictionary, for its “programmed” proposal. PO Resp.,
`
`pp. 14-15. Moreover, far from simply being the “broad” technical dictionary
`
`alleged by MI, the Microsoft Computer Dictionary is a source consistent with the
`
`
`
`4
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`background of a POSITA to which both experts agreed: a bachelor’s degree in
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`electrical engineering, computer engineering, or computer science, and at least two
`
`years of experience in the design of multiprocessor systems. See Ex. 1014, ¶ 8; Ex.
`
`2016, ¶ 8.
`
`Finally, the Petition contends that, even within the specific field of cache
`
`coherence, the scope of the term “states” is “broad enough to include the condition
`
`of presence,” as found by the Board. See, e.g., Petition, p. 10 (identifying the
`
`Chaiken reference’s use of “status” to reference presence condition). The ‘121
`
`patent accords by also using the term “state” to refer to presence. See ’121 patent
`
`at 13:60-61 (“In the invalid state, a memory line is not currently available in cache
`
`associated with any remote cluster.”). Indeed, the Sorin and Culler references that
`
`MI cites also identify a “Not Present State.” Ex. 2010 at 89; Ex. 2002 at 4; see Ex.
`
`1026 at 38:3-39:11 (Sorin and Culler are reliable references).
`
`Accordingly, it is reasonable, for purposes of this proceeding in which the
`
`broadest reasonable construction standard applies, to consider the term “states” as
`
`being broad enough to encompass the condition of presence.
`
`B. “Programmed”
`MI advanced arguments regarding the term “programmed” in its preliminary
`
`response, without explicitly construing the term. MI now improperly seeks to
`
`effectively amend the claim to overcome the Board’s preliminary endorsement of
`
`
`
`5
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`the Koster reference, by advancing a narrowing interpretation of this term. MI
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`asserts that “the term ‘programmed’ should be construed to refer to a device that
`
`has been ‘configured by a sequence of instructions’” (PO Resp., p. 13) and argues
`
`that “‘programmed’ is not broad enough to encompass hardwired logic” (PO Resp.,
`
`p. 17). However, these proposals are not supported by the evidence. See Ex. 1025,
`
`¶¶ 2-6.
`
`The ’121 Patent does not use the words “instruction” or “execute.” The ’121
`
`Patent makes no distinction between hardwired logic and the relevant portion of
`
`the processors that dictate the completion of a memory transaction. MI has not
`
`cited to any language in the ’121 Patent which provides a clear limitation to the
`
`nature of the programming. See PO Resp., pp. 13-14. Indeed, the ’121 Patent is
`
`clear that these cited sections are “merely exemplary for the purpose of describing
`
`a specific embodiment of the present invention.” Ex. 1001, 28:2-5. Further, even
`
`the dictionaries that MI cites include reasonable definitions for programming that
`
`defy MI’s proposal, such as “to work out a sequence of operations to be performed
`
`by (a mechanism).” Ex. 2014, p. 931.
`
`Indeed, hardwired logic is designed to perform a sequence of operations.
`
`For example, a field programmable gate array (FPGA) is effectively an array of
`
`logic gates that can be interwired in different configurations according to a
`
`manufacturer’s programming. Ex. 1025, ¶ 3. Contrary to MI’s contention that the
`
`
`
`6
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`term “programmed” is limited to devices that execute a series of instructions, a
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`field programmable gate array does not execute any instructions. Id. Rather, a
`
`designer uses an interface to program the physical interconnections between the
`
`logic gates, and the logic gates within the FPGA each perform a logical operation
`
`corresponding to the gate’s type and its input signals. Id. Indeed, MI’s expert
`
`admitted that “everybody use[s]” the term “programmable” with respect to an
`
`FPGA. Ex. 1026, 123:12-124:11.
`
`MI’s evidence states that “in hardwired logic systems the physical
`
`interconnections of the elements govern the routes by which data flows between
`
`the processing elements and thus the sequence of processing operations performed
`
`on the data.” Ex. 2015 at 15/3. Yet, this is entirely consistent with the operation of
`
`the aforementioned field programmable gate array. Ex. 1025, ¶ 5. By its very
`
`name a field programmable gate array is “programmed” by a designer. Indeed,
`
`MI’s own specification uses the term “programmable” when referring to devices
`
`that do not execute instructions, teaching that “the cache coherence controller 230
`
`is a specially configured programmable chip such as a programmable logic device
`
`or a field programmable gate array.” Ex. 1001, 7:49-52 (emphasis added).
`
`Accordingly, MI’s attempt to narrow the term “programmed” to exclude
`
`hardwired systems is inconsistent with the use of the term “programmable” in the
`
`’121 Patent and with the extrinsic evidence cited by MI. Indeed, to read the
`
`
`
`7
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`limited passages of the ‘121 Patent cited by MI any differently would be a clear
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`violation of In re Bigio, as the section of the ‘121 Patent in which these passages
`
`appear are clear that the examples given are not meant to be limiting. See Ex.
`
`1001, 28:2-5. Instead, consistent with its usage in the context of FPGAs and the
`
`dictionary cited by MI, the term “programmed” should be construed at least
`
`broadly enough to encompass “designed to perform a sequence of operations,”
`
`regardless of whether this design is in hardware or software. See Ex. 1025, ¶ 6.
`
`III. The Instituted Grounds Anticipate or Render Obvious Claims 4–6, 11, 12,
`and 19–24
`In its Response, MI largely reiterates arguments raised in its Preliminary
`
`Response. These arguments amount to attempts to improperly import limitations
`
`from the specification into the claims, as discussed above in Section II supra, and
`
`to otherwise misinterpret the plain teachings of the Koster reference. The Board
`
`properly found such arguments deficient in its Institution Decision and should do
`
`so again, as none of the evidence now proffered by MI establishes a valid legal
`
`basis for altering the Board’s initial findings.
`
`A. Koster Discloses the Claimed “States”
`MI’s argument rests entirely upon its claim construction of the term “state,”
`
`which is improperly narrow for the reasons described above. See Ex. 1027,
`
`186:17-24 (admitting that Koster discloses “states” under the Board’s preliminary
`
`construction). The Board should uphold its preliminary determinations of the
`
`
`
`8
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`scope of the term “state,” and, in so doing, the Board should reject MI’s arguments
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`as moot. However, even if the Board were to reverse its preliminary
`
`determinations—which it should not—for reasons noted below, the Board should
`
`find that Koster meets the claims by describing even MI’s narrow definition of
`
`“states.”
`
`As noted in the Petition, “Koster specifically describes the shadow tag
`
`memory 164 as ‘local state memory . . . that stores the tags of data stored in the
`
`local cache memories.’” Petition, p. 28 (citing Ex. 1009 at 6:11-13) (emphasis
`
`added). In particular, Koster discloses that the “shadow tag memory ... may use a
`
`MOESI (Modified Owner Exclusive Shared Invalid) cache coherency protocol.”
`
`Ex. 1009 at 6:33-38. As acknowledged by the ‘121 Patent, the five components of
`
`the MOESI protocol (Modified, Owner, Exclusive, Shared, and Invalid) are
`
`“memory line states.” See Ex. 1001, 14:30-36. Thus, Koster’s shadow tag
`
`memory 164 includes probe filtering information representative of states associated
`
`with selected ones of the cache memories. MI’s arguments to the contrary can be
`
`dismissed for either of at least two reasons.
`
`1. MI’s Focus on Koster’s Tags is Inapposite
`
`MI initially attempts to draw attention away from Koster’s explicit teaching
`
`of a “local state memory” that “uses a MOESI . . . cache coherency protocol” by
`
`focusing on Koster’s additional teaching of “tags” stored in that memory and
`
`
`
`9
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`comparing MI’s incomplete description of these tags to its own narrow
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`construction of “states.” However, regardless of whether these “tags” meet MI’s
`
`narrow construction of “states,” MI’s own logic in construing the term “states”
`
`dictates that information representative of cache coherency states is stored by
`
`Koster’s “local state memory.”
`
`Under MI’s logic, because the field of the ‘121 Patent is cache coherency,
`
`the ‘121 Patent’s use of the term “state” would necessarily be understood by a
`
`person of ordinary skill in the art to mean a “cache coherency state.” Specifically,
`
`in Section II(A) of its Response, MI cites to use of the phrase “cache coherency”
`
`within the background section and within various portions of the detailed
`
`description as proof that “the teachings of the ’121 Patent make it clear that its
`
`inventions are directed to the specific field of cache coherency.” PO Resp., pp. 3-
`
`4. MI then argues that “the term ‘state’ connotes a specific meaning in the field of
`
`cache coherency—a cache coherency state.” PO Resp., p. 4.
`
`By the same logic, because Koster is directed to the specific field of cache
`
`coherency, its use of the word “state” is equally applicable to a “cache coherency
`
`state.” Koster is titled “Snooping-Based Cache-Coherence Filter for a Point-To-
`
`Point Connected Multiprocessing Node.” Ex. 1009, Title (emphasis added).
`
`Indeed, Koster uses the term “cache-coheren[ce/cy/t]” over thirty times throughout
`
`its specification. See, e.g., Ex. 1009, title, 3:15, 3:17, 3:44, 3:49, 3:51, 3:55, 4:22,
`
`
`
`10
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`4:26, 4:28, 4:43, 4:51, 4:54, 4:58, 4:60, 4:64, 5:1, 5:7, 5:9, 5:15, 5:17, 5:18, 5:21,
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`6:7, 6:10, 6:37, 9:15, 10:36. Koster is no less directed to the specific field of cache
`
`coherency than the ’121 Patent. Thus, assuming, arguendo, that MI’s construction
`
`argument is correct, Koster’s description of a “local state memory” would
`
`necessarily be read to be a local cache coherent state memory. This is consistent
`
`with the title, which specifies that the snoop filter 162 is a cache-coherence filter.
`
`See Ex. 1026 at 147:14-148:15 (Koster is written in the field of cache coherency
`
`and claims use of MOESI protocol).
`
`2.
`
`Koster’s Snoop Filter Filters Requests Based on State
`Information Contained in the Local State Memory
`
`MI’s attempts to explain away Koster’s use of the word “state” and
`
`description that the local state memory “may use a MOESI . . . cache-coherency
`
`protocol” are inconsistent with the logic underpinning its own narrow
`
`interpretation of the teachings of Koster. MI argues that “despite Koster’s
`
`reference to such tags as ‘state’ information, the tags merely indicate an address
`
`and are not representative of cache coherency states.” PO Resp., p. 23. However,
`
`Koster doesn’t reference the tags as “state information,” as implied by MI. Rather,
`
`Koster discloses that “the snoop filter 162 has local state memory (referred to and
`
`shown in FIG. 7 as ‘shadow tag memory’) 164 that stores the tags of data stored in
`
`the local cache memories.” Ex. 1009, 6:11-13. This description indicates one of
`
`two things: (1) the local state memory stores information representative of cache
`11
`
`
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`coherency states in addition to the tags; or (2) Koster’s tags themselves would have
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`been understood by a skilled artisan to represent cache coherency state information
`
`in addition to location information. See Ex. 1025, ¶ 39.
`
`Regarding the latter, Koster’s tags do not merely provide the snoop filter
`
`with addresses identifying locations of copies of requested data. Instead, they
`
`provide the snoop filter with addresses identifying locations of copies of requested
`
`data that are in a valid cache coherency state. After all, a person of ordinary skill
`
`in the art would understand that Koster’s snoop filter 162 uses the tags to forward
`
`data requests only to those processors known by the snoop filter 162 to have valid
`
`copies, i.e., only those processors able to productively respond to the data request.
`
`In only identifying the location of valid copies, Koster’s tags themselves reflect
`
`cache coherency state information in addition to location information. See 1025, ¶
`
`40.
`
`MI further argues that “Koster does not disclose that shadow tag memory
`
`164 stores the MOESI information. Nor does Koster disclose that the snoop filter
`
`uses the MOESI information in the process of determining where to send a probe.”
`
`PO Resp., pp. 23-24. However, a person of ordinary skill in the art would
`
`understand that, in order to “use a MOESI . . . cache-coherency protocol,” the
`
`shadow tag memory 164 would store information about the protocol (i.e., “MOESI
`
`information”). See 1025, ¶ 41. Moreover, Koster discloses the only functionality
`
`
`
`12
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`of the snoop filter 162 as “observ[ing] snooping-based cache-coherence broadcasts
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`for requested data and the responses thereto,” and filtering the broadcasts
`
`according to the information stored in the local state memory 164. See Ex. 1009,
`
`6:7-32. Therefore, a person of ordinary skill in the art would understand the snoop
`
`filter’s “use” of the MOESI cache-coherency protocol as related to its snooping
`
`and filtering requests.
`
`For at least these reasons, Koster discloses that the snoop filter 162 filters
`
`probes “with reference to probe filtering information representative of states
`
`associated with selected ones of the cache memories,” as recited in claims 1 and
`
`16.
`
`B. Koster Discloses that Each of its Processing Nodes is Programmed
`in the Manner Recited in Claim 11
`MI’s argument rests entirely upon its claim construction of the term
`
`“programmed,” which is improperly narrow for the reasons described above.
`
`Because this construction is too narrow, MI’s arguments are moot. Indeed, the
`
`Board already addressed nearly identical arguments from the preliminary response
`
`regarding the prior art’s teaching of “programming,” and justifiably found that, “by
`
`discussing microprocessors, [the prior art] inherently discloses that this is done via
`
`some sort of programming.” Institution Decision, p. 21.
`
`Under the proper construction of the term “programmed,” either of the
`
`methods MI argues that Koster’s processors could employ to complete a memory
`
`
`
`13
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`transaction would satisfy the limitations of claim 11. Specifically, MI does not
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`dispute that the processors described by Koster “complete a memory transaction
`
`after receiving a first number of responses to a first probe, the first number being
`
`fewer than the number of processing nodes.” See PO Resp., p. 26. Rather, MI
`
`simply alleges that “nothing in Koster requires that the microprocessor be
`
`configured to complete memory transactions using programming rather than being
`
`hardwired.” See PO Resp., p. 28. MI admits that, in Koster, “a configuration must
`
`be either hardwired into the processor or set via some sort of specific programming
`
`targeted at the internal configuration of the processor itself.” PO Resp., p. 29.
`
`Applying the proper construction, as described in Section II(B) supra, the
`
`claimed “programmed” is satisfied regardless of which of MI’s acknowledged
`
`configurations is used by Koster. That is, regardless of whether in Koster the
`
`claimed functionality is “hardwired into the processor or set via some sort of
`
`specific programming targeted at the internal configuration of the processor itself,”
`
`the proper construction of “programmed” is met because each configuration has a
`
`processor that is “designed to perform a sequence of operations” to complete a
`
`memory transaction after receiving a first number of responses.
`
`Accordingly, Koster discloses that each of its processors is programmed to
`
`complete a memory transaction after receiving a first number of responses to a first
`
`
`
`14
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`probe, the first number being fewer than the number of processing nodes, as recited
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`in claim 11.
`
`C. Koster Discloses the Temporary Storage Recited in Claim 12
`MI argues that Koster does not teach that “the probe filtering unit has
`
`temporary storage associated therewith for holding read response data from one of
`
`the cache memories,” as recited in claim 12. Specifically, MI argues that “the
`
`functionality described by Koster can be implemented without the use of any
`
`temporary storage for read response data in the snoop filter.” PO Resp., pp. 30-31.
`
`In support of this argument, MI relies on a single technique known as “circuit
`
`switching” as proof that Koster “does not require that the response data must
`
`necessarily be stored temporarily in the snoop filter.” PO Resp., pp. 33-34.
`
`However, by advancing this argument, MI demonstrates a misunderstanding of the
`
`teachings of Koster.
`
`In reference to FIG. 9, Koster discloses a case in which “the snoop filter 192,
`
`referencing its shadow tag memory 194, determines that microprocessor 188 has a
`
`copy of the data requested in broadcast A, and thus, the snoop filter 192 forwards
`
`broadcast A to microprocessor 188.” Ex. 1009, 7:6-10. After receiving this
`
`broadcast A, Koster teaches that the microprocessor 188 issues a response B.
`
`Koster explicitly discloses that “response B (having a copy of the requested data) .
`
`. . is forwarded back to the requesting microprocessor 182 through the snoop filter
`
`
`
`15
`
`
`
`Patent No. 7,296,121
`Petitioner’s Reply to Patent Owner Response
`
`192.” Ex. 1009, 7:12-14. In other words, response B is a message that contains a
`
`Attorney Docket No. 39521-0007IP4
`IPR2015-00163
`
`
`
`
`copy of the requested data (i.e., the read response data), and response B is received
`
`at the snoop filter 192. Therefore, snoop filter 192 receives the read response data
`
`and forwards it to requesting microprocessor 182. Ex. 1025, ¶ 45. Indeed, MI
`
`does not argue that the snoop filter 192 does not receive the read response data.
`
`Rather, the issue raised by MI is whether, in the process of receiving and
`
`forwarding the read response data, the snoop filter 192 temporarily stores the read
`
`response data. MI does not dispute that any form of temporary storage at snoop
`
`filter 192 would read on the language of claim 12. Instead, MI’s sole argument is
`
`that Koster’s snoop filter 192 may employ “circuit switching,” a switching
`
`technique in which data can be forwarded through a switch without being buffered,
`
`and that Koster therefore avoids buffering read response data at the snoop filter
`
`192. However, MI’s argument fails because Koster explicitly discloses an
`
`implementation that requires buffering.
`
`In conventional interconnection networks, “[w]hen a message or packet
`
`header reaches an intermediate node, a switching mechanism determines how and
`
`when the router switch is set, i.e., the input channel is connected to the output
`
`channel selected by the routing algorithm.” Ex. 1007, p. 12. Two types of
`
`switching mechanisms are circuit switching and packet switching. “[I]n circuit
`
`switching, all the channels required by a message are reserved before starting
`
`
`
`16
`
`
`
`Patent No. 7