奎芯科技 MSquare Technology
目录 CONTENT
公司简介 01
行业愿景 02
大事记 03
公司荣誉 05
知识产权 06
产品综述 07
⊚ ML100 I0 Die 08
⊚ HBM3 IP 09
⊚ D2D (UCle) IP 10
⊚ ONFI5.0 IP 1
⊚ ONFI5.1IP 12
⊚ LPDDR4X IP 3
⊚ LPDDR5X IP 14
⊚ eDP IP 15
⊚ PCle 4.0 IP 16
公司简介
奎芯科技(MSquareTechnology)于2021年注册成立,是一家专业的集成电路IP和Chiplet产品供应商。公司专注于解决智慧经济时代下芯片互联和应用垂直整合的挑战。目前在上海,东京,悉尼,圣何塞等地拥有办公室,拥有一支数百人的团队,其中研发人员占比高达 80% 奎芯科技致力于构建一个AI与Chiplet的开放生态服务平台,旨在为集成芯片和芯粒产业的创新和发展提供全方位的支持。奎芯科技(MSquareTechnology)命名来自于双胞胎星系。该星系由两个重叠的螺旋星系构成,其闪耀的光芒照亮了其半径6倍的区域,天文学家借助它的指引观察周围的星系。奎芯科技以此自喻,期望凭借先进的的产品和服务引领半导体产业的繁荣和发展,与产业链上下游共同打造共赢且开放的生态,照亮整个半导体领域。
作为半导体产业链上游的技术引领者,奎芯科技通过整合系统设计、ChipletDie及2.5D封装联盟的定制业务,勇攀芯片设计技术的巅峰。我们立志成为AI时代的先锋领航者,专注于互联接口IP与Chiplet产品的创新研发,广泛覆盖人工智能、数据中心、汽车电子、物联网及消费类电子等多元领域。通过提供先进的半导体IP和Chiplet研发与定制服务,奎芯科技致力于打造一个开放、协同的生态体系,以引领行业标准,推动硬科技创新的深层发展。
产品愿景
PRODUCTVISION
奎芯科技旨在打造国际芯粒品牌,成就互联IP龙头。凭借丰富的IP产品组合和先进的技术实力,奎芯科技正积极构建业界领先的IP基础设施平台,引领AI与Chiplet技术创新。奎芯科技的IP已经成功在一些知名厂商的工艺节点得到验证并实现量产,目前已成功开发5nm到1 |80\mathsf{n m} 、覆盖五大晶圆厂的400多个不同制程节点的IP产品。研发团队已陆续推出UCle、HBM、LPDDR、ONFI、eDP、PCle、USB等互联接口IP,以及以M2LINK为代表的Chiplet产品解决方案,打造开放生态的服务平台,持续赋能数字化转型。
10月 ONFI5.112nm交付 LPDDR5X 6nm交付 HBM312nm交付
国际认证体系
ISOCERTIFICATIONS
HIGHLIGHTS
MSquare'sML100IODieisahigh-bandwidthmemorysolutionthatintegratesefficientUCle(Die-to-Die) interconnect IPand HBM3IP.TheUCleIPfollowstheUCle1.1Specificationandsupportsbothstandard andadvancedpackagingtechnologiesofferingupto1TB/softransferbandwidthinasinglemodule configuration.TheUClesupportsthe {\mathsf{A X}}|4.0 interfacestandard,enablingultra-lowlatency,high-speed interconnectivitybetweentwodies.AndtheintegratedHBM3IPadherestotheHBM3JESD238standard andsupportsIOtransferratesofupto6400Mbps.
| Integrates16completeUClemodules,following the standardUCle1.1 standard |
| IntegratsnecmletMnldingthcontrlleandHllwingthMD38standard |
| SupportsinterconnectivitywiththeSOCviatheAMBAAXI4.0standard |
| TheAXi4.0bussupportsbandwidthtrafficandlatencystatistics |
| Supportsvariousperipherals,i.e.,NORFlash,QSPI,2CandART |
| Maximumbandwidthis819.2GB/s,matchingthebandwidthof6400MbpsHBM3DRAM |
| Utilizesa2.5Dpackagingsolution,UCleinterconnectedwiththeSOCthroughsubstrate |
| DecouplestheHBMfromtheSOC,reducingthethermaleffectfromSOCtoHBMDRAM |
MSquare'sHBM3IP(the 3^{\mathsf{r d}} generationof High-BandwidthMemory)isspecificallytailoredforapplications thatrequirehighmemorythroughputandlowlatency,complyingwiththeJESD238memorystandard.lt includesbothPHYandMemoryControllercomponentssupportingHBM3SDRAMspeedsrangingfrom 4.8Gbps/pinto6.4Gbps/pin.Flexibleconfigurationsareavailable,includingPHYOnly,PHY ^+ Controller, andControllerOnlytoaccommodate diversecustomerdesignspecifications.Additionally,thechip's footprintandpowerconsumptionarehighlycompetitivewithintheindustry.
| Designed forhighmemory throughput and low latencyapplications |
| ConsistsPHYandMemoryController |
| Optimizedfor12nmprocess |
| Supportsspeedupto6.4Gbps/pin |
| FeaturesintegratedPLLandIO |
| Flexible configurations available:(PHYonly)or(PHY+Controller)or(Controlleronly) |
| Supportsbothfirmware-based trainingandhardware-basedtraining |
| OptionalAdd-onIPstoachievebestperformance |
| Minimizedfootprintandpowerconsumption |
OVERVIEW
TheDie-to-Dieinterfaceisafunctionalblockthatprovidesadatainterfacebetweentwochipdieswithin thesamepackageMSquaresD2DsolutioncompatiblewiththeUCle1.secificationincludesboth Die-to-Dieadapterlayerandaphysicallayer.Eachlayerfeaturesasidebandinterfacethatprovidesa back-channelforlinktrainingandaccesstotheregistersof thelinkpartnerasshowninFigure1.This uniquehybridanalog/digitalarchitectureofferslowpowerconsumptioncompactfootprintandrobust performancemakingitwell-suitedfortargetapplicationssuchashigh-performancecomputinglr multimediaSoCs,andDie-to-Dieinterconnections.
| CompatiblewithUClev1.1specification |
| Featuressingle-endedsource-synchronousandDDR/Osinaling |
| Supports32-bit(16-bitsTX+16-bitRX)databuspermoduleforstandardpackages |
| Offersahighclockfrequencyupto16GHz |
| Providesdatatransferrateupto32Gbpsperlane |
| Delivers1Tbps(512GbpsTX+512GbpsRX)bandwidthpermoduleforstandardpackages |
| Highenergyeffciencywith~0.8pJ/bitforstandardpackages |
| Built-intestanddiagnostics[Functional(PRBS),scan,at-speedexternal loopback] |
| Theadvanced12nm/6nmprocess |
ONFI5.0IP
OVERVIEW
MSquareoffersa silicon-provenONFI5.0PHYIPthat supportsallmodesof theOpenNANDFlash Interface(ONFI)5.0 specification.ONFIis aninterface that connects external flash particles(NAND Flash)andinternalflashcontroller(NANDController),commonlyusedinSolidStateDrives(SSD)and embeddedmemory(eMMC)products.Tofacilitatecustomerimplementation,theMSquare'sONFI5.0 PHYsupportsexpandablechannelsandcustomizableCE/RBPadNumbers.
| CompliantwithONFI5.0specification |
| SupportsNV-DDR2mode |
| SupportsNV-DDR3,NV-LPDDR4,withamaximumrateof2400MT/s |
| SupportsPHYIndependentTX/RXTrainingmode |
| SupportsDFIwith1:1/1:2clockratiomode |
| Supportsself-testmodewithbuilt-inBIST |
| Supportsloopbacktestwithbuilt-inPRBS |
| SupportsODTprogrammableandZQcalibration |
| Integratedlow-jitterPLL&DLL |
MSquare'sONFI5.1PHYisahigh-performanceandlow-powerPHYIPwhichsupportsallmodesof the Open NAND Flash Interface(ONFI)5.1 specification.ONFI is an interface connecting external flash particles(NANDFlash)andinternalflashcontroller(NANDController),widelyusedinSolidStateDrive (SSD)productsosimlifytheimplementationMSquareONFI5.1PHYsupportsexpandablechannes andcustomizableCE/RBpadnumbers.
HIGHLIGHTS
| CompliantwithONFI5.1specification |
| SupportsNV-DDR3/NV-LPDDR4,withamaximumrateupto3600MT/s |
| SupportsmatchedorunmatchedDQS |
| SupportsWDCA/Per-PinVREFQTrainingforNANDdevices |
| SupportsWTMonitor |
| SupportsSeparateCommandAddress(SCA)toimproveNANDIOefficiency |
| Supports2channelswithasinglePHYPLL&flexiblefloorplanwithdoubleorsinglerowIO |
| SupportsDFIwith1:1/1:2clockratiomode |
| SupportsTX/RXTrainingmode |
| Delay-ChainsupportsbackgroundtrackingforPVTcompensation |
| TestMode:Self-testmodewithbuiltinBIST&Loopbacktestwithbuilt-inPRBS&BIST-Delay-Chain |
MSquare'sLPDDR4XPHYisatransceiverphysical layerIPinterfacesolutiondesignedforASICsand SOCs.CompatiblewiththeuniversalIPprotocolDFI4.0itoperatesatadatarateofupto4267Mbpswith a16-bitdatawidthperchannel.AsdepictedinFigure1,theLPDDR4XPHYblockdiagramforsystem applicationincludesadataandaddresstransmitpathareceiverdatapath,aLLblockandmore. Featuringahybridanalog/digitlarchitcturequaredliverslwpowerconumptionom footprint,androbustperformance,makingitwell-suitedforPDDR4Xapplications.
HIGHLIGHTS
| CompatiblewithLPDDR4X,withamaximumrateupto4267Mbps |
| CompliantwithDFI4.0forPHYandcontrol interfaces |
| Flexiblechannelarchitecture |
| SupportsPHY-independent trainingmodeusinganembeddedprocessor |
| Supportsdualrank |
| DFSsupports4trainedfrequencies |
| SupportsBISTandloopbackmodes |
| SupportsbackgroundtrackingforPVTcompensation |
| Integratedlow-jtterPLLandDLL |
| CompliantwithJEDECstandardJESD209-4C |
OVERVIEW
MSquare'sLPDDR5X/4XPHYisatransceiverphysicallayerIPinterfacesolutiondesigned forASICsand SOCs.CompatiblewiththeuniversalIPprotocolDFI5.0itoperatesatdataratesupto8533Mbpswitha 16-bit datawidthperchannel.Withflexible configurationoptions,theLPDDR5X/4XPHY supportsa varietyofmobileapplicationsusingPDDR5X,PDDR5andLPDDR4XDRAMs.Featuringahybrid analog/dgitlarchitctuequareliverswwrconsumtionmpactootritndrbu performance,makingitwell-suitedforLPDDR5X/4Xapplications.
HIGHLIGHTS
| CompatiblewithJEDECstandardsLPDDR4X,LPDDR5andLPDDR5XSDRAMs |
| Supportsfordatatransferrateupto8533Mbps |
| DFI5.0forPHYandcontrollerinterfaces |
| Supportsbothfirmware-based training andhardware-based training |
| SupportsmultiplegearsofDFS |
| Highperformance,lowjitterDDRIO |
| SingleTapDFEforChannelRobustness |
| Supports4trainedfrequencies |
| User-customizablearbiter(scheduler) |
| CompliantwithJEDECstandardsJESD209-4DandJESD209-5B |
OVERVIEW
MSquaresDPHisahighlyreliablesolutionfordisplayintfacerequirmentsupportingresolutions of4KandhigherByperatingatubstantiallyhigherbitratesthisDPsoltionreducesboththenumer ofwiresandpinsrequiredomparedtointerfaceslieDVandMtisfullcomliantwith v1.5and1.4bstandards,andcapableofdrivingupto8.1Gb/speraneinconfigurationsupto4lanesl consistsofelectricalsub-blocksofmain-linkandAUXchannelofPHYforeDP.MSquareprovideseDPwith superiorPPAandofferscustomizationbasedoncustomerspecifications.
| Compatiblewithembeddeddisplayportv1.5/v1.4bspecification |
| Four-lanemain linkwith support for8.1/5.4/4.32/3.24/2.7/2.43/2.16/1.62Gbps |
| SupportsEnhancedFramingMode |
| AutomaticLink-Trainingwithanoptionforfirmware-controlledLink-Trainingprocedure |
| SupportsFast LinkTraining/NoLinkTraining |
| SupportsAUXChannel foraccesstoDPCDandEDID |
| Supportsamaximum144Hzrefreshrate |
| SupportsvariablerefreshrateandAdaptive-SyncforVESAAdaptiveSync |
| SupportsPanelSelfRefresh(PSR)functionandPSRwithselective update(PSR2)function |
| SupportsVESADSCv1.01,v1.1,v1.2awithFEC |
| SupportsPanel Replay(PR)and AdaptiveRefreshPanel(APR) |
PCle 4.0 IP
OVERVIEW
MSquareExpress4.Hnludeshigseedhighl-ficientandcost-ffctiveransiv to turbocharge today'shigh-performance and high-bandwidth aplications such as data centers, automotive,enterprisecomputing,andhigh-performanceembeddedsystems.
ThePCleIPboastsalowpowerconsumptionandcompactsiliconfootprint.ItsrobustPHYarchitecture tolerateprocessvoltageandtmerature)ariationshisPintgratesigh-speedmixedsia circuitsto supportPCle4.0 traffic at16Gbps.It isbackward compatiblewithPCle3.1datarates at 8.0Gbps,PCle2.1at5Gbps,andPCle1.1at2.5Gbps.hemulti-tabtransceiverdesign,accompand byarobustanmbddedbitrrorrateBstrandaninenaymonitrnaldig tocontroltest,andmonitorsignalintegritywithouttheneedforexpensivetestequipment.
HIGHLIGHTS
| CompileswithPCle4.0,3.1,2.1,1.1andPIPE4.4.1specifications |
| Supportsallpower-savingmodes(P0,POs,P1,P2)asdefined inPIPE4.4.1specifications |
| SupportsL1PM/CPMsubstateswithCLKREQ# |
| SupportstheseparateREFCLKIndependentSSC(SRIS)architecture |
| AccessibleegistercontroLbandwidthde-mphasieveDRbandwidthandEQstreng |
| SupportsbothFOMandDIRmodesforLink-EQTraining |
| SupportsrobustBIST/DFTfunctionsformassproductiontests |
| Supportsbifurcationwith4-lanex1or2-lanex2configurations |
www.m2ipm2.com




