Soitec's Engineered Substrates For Edge Computing

1y ago
23 Views
3 Downloads
1.61 MB
23 Pages
Last View : Today
Last Download : 2m ago
Upload by : Gia Hauser
Transcription

FD-SOI Substratesfor Edge ComputingMichael REIHAFD-SOI Business Unit ManagerExane BNP Paribas - Edge Computing ConferenceDecember 15th, 2020

DisclaimerSOITEC PROPRIETARY - NO COPYING OR DISTRIBUTION PERMITTEDAny unauthorized reproduction, disclosure, or distribution of copies by any person of any portion of this work may be a violation of Copyright Laws, could result in the awardingof Damages for infringement, and may result in further civil and criminal penalties.All rights reserved. Copyright 2020 SOITEC.Dec. 20202FD-SOI Substrates for Edge ComputingSoitec proprietary

Soitec megatrendsSemiconductormegatrendsDifferentiated engineered substrates to serveour strategic end marketsKey figures in H1’21254 M sales5G30.4%EBITDAmarginAI102 M operatingcashflowEE**Energy EfficiencyDec. 20203FD-SOI Substrates for Edge ComputingSoitec proprietary

OutlineDec. 202041What is Edge computing?2How is Edge computing utilized?3Why is FD-SOI seamless for Edge computing?FD-SOI Substrates for Edge ComputingSoitec proprietary

OutlineDec. 202051What is Edge computing?2How is Edge computing utilized?3Why is FD-SOI seamless for Edge computing?FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Decentralized data processing and analysisEdge computing integrates intelligence to edge devicesData is processed and analyzed in real time near the sensor nodeSource: www.alibabacloud.comDec. 20206FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Intelligent analysis autonomous from the cloudReal time analytics, increasing privacy/safety, providingnew value & experienceSenseThinkConnectActThings (devices) with varioussensorsIntelligence, analysis,managementWired & a ProcessingDevice ControlData AnalysisDec. 20207FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Evolution from cloud to on-device Edge computingDec. 2020BeforeNowFutureCloudEdgeOn-device EdgeAI training in the CloudAI training in the CloudAI training at the Edge Inference in the CloudInference at the EdgeInference at the Edge8FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Chip features and applicationsLow ComplexityMedium ComplexityHigh ComplexityChip FeaturesDynamic Always-ON / UbiquitousEnergy Efficient ComputingHighly Reliable / PredictiveExamples ofApplicationsSmart City / Smart Home (Sensors)Smart Devices / WearablesSmart Vehicles / Smart MachinesDec. 20209FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Towards a trillion of connected ‘things’ConsumerForecasted number of AI enabled Edge systemsby Industry (M units)1,000Transportation750 30%CAGRCAGR500250 15% 24Consumer2025Industrial 60% CAGR0Source: IHS Markit Artificial Intelligence - Status of the Market Report 2019Dec. 202010FD-SOI Substrates for Edge ComputingSoitec proprietary

OutlineDec. 2020111What is Edge computing?2How is Edge computing utilized?3Why is FD-SOI seamless for Edge computing?FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Training vs. InferenceTrainingCreating AI model from existing dataInferenceApplying AI model to new dataNew data intotrained modelEvolutionCloud Cloud and EdgeCloud and Edge On-device EdgeLimitationsPower ceiling due to thermal inefficiencyEnergy limited due to battery capacityRequirementsHigh-Performance (TOPS)Energy efficiency (mJ / frame), Competitive costMain ArchitecturesCPU, GPU, TPU, FPGALow Power FPGA, NPU, MCUTechnologiesFinFETFD-SOI, 2.5D-3D packagingDec. 202012FD-SOI Substrates for Edge ComputingSoitec proprietarySource: Nvidia, Soitec

Edge computing – Requires new paradigm for efficient designThroughput vs LatencyInferences per second, frame rate,sub-ms delayHardware cost of ownershipRobustnessOn-chip storageNumber of processing elementsChip areaProcess technologySoft-error rate(20x better than bulk technologies)Environmental factors (up to 125 C)Energy and PowerAccuracy vs EfficiencyBattery lifetime ( 10 years),Energy/operation, Memory bandwidthRange of deep neural networks modelsInterconnect designArchitecture selectionConnectivity and IntegrationSource: Soitec, industry dataDec. 202013FD-SOI Substrates for Edge ComputingSeamless integration of Radio and DSPEmbedded compute IPSoitec proprietary

OutlineDec. 2020141What is Edge computing?2How is Edge computing utilized?3Why is FD-SOI seamless for Edge computing?FD-SOI Substrates for Edge ComputingSoitec proprietary

FD-SOI substrate structure and manufacturing sitesFD-SOI substrate structureUltra-thin top silicon & boxenabling fully-depleted transistoroperation300 mm high volume manufacturingin France and in SingaporePasir Ris, SingaporeBernin 2, FranceAwarded “Factory of the year 2020” in France byL’Usine Nouvelle, thanks to Industry 4.0 initiativesDec. 202015FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Why FD-SOI?FD-SOI is a power-efficient & flexible mixed-signal platformwhich can enable analog/RF integration for edge computing applicationsCost of Edge accelerationEnergy(pJ)Operation8b Add0.0316b Add0.0532b Add0.116b FP Add0.432b FP Add0.98b Multiply0.232b Multiply3.116b FP Multiply1.132b FP Multiply32b SRAM Read (8kB)32b DRAM ReadRelativeEnergy Cost12313››››307103373.71235167640Edge computingrequirements›FD-SOI value propositionLimit data size (8 bit)› Low power devices for efficientOn-chip memory› Low energy eNVM for on/near› Lowest power connectivityReduce number ofconvolutions21333FD-SOI Substrates for Edge Computingdata conversionmemory computeEnergy-efficientarchitectureSource: Horowitz, ISSCC 201416computeReduce read/writeand Energy/MAC1 10 102 103 104 105Dec. 2020› High speed devices for analogSoitec proprietary(BLE, NB-IoT, WiFi)› Design-to-cost technology

Edge computing – FD is the ideal platform for edge inference1000VGG Neural Network modelGreenWaves GAP9Input Image (frame per second)Lattice CrossLink-NX100FD-SOIRockchip RK3399Nvidia Jetson AGX XavierFD-SOIFD-SOI10Nvidia Jetson NanoNvidia Jetson TX2Nvidia Tesla P4Nvidia Tesla V1001Raspberry Pi 3 Intel Neural Compute Stick 2STMicroelectronics STM32H70.10.010.11101001000Power Consumption (W)Source: Soitec, industry dataDec. 202017FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – Edge-based Integrated Circuits using FD-SOILow Power FPGA› CrossLink-NXTM built on theIoT Application Processor› GAP9 IoT, state-of-the-artEdge Inference Processor› Ergo delivers 4 TOPS28FDS Lattice Nexus platformApplication Processor in 22FDXsustained and 55 TOPS/W,for Vision Processingfor the Next Wave ofcapable of processing largeApplicationsIntelligence at the Very Edgeneural networks in 20mW, in22FDXSource: Lattice SemiconductorDec. 202018FD-SOI Substrates for Edge ComputingSource: GreenWaves TechnologiesSoitec proprietarySource: Perceive

Future FD-SOI opportunities for Edge computing applicationsFinFET delivers performanceHeterogeneous packaging offersFD-SOI delivers thermal & power efficiencyFD-SOIHigh Fan-Out InterfaceI/OParallel I/O reduces powerMemoryFinFETMemoryMultiple Scalable OptionsI/OExtremely Low-Leakage››Improved thermal efficiency›››Enhanced cost utilization with FinFET››More competitive cost structureSource: SoitecDec. 202019FD-SOI Substrates for Edge ComputingSoitec proprietaryScalable architectures via a chipletbased designLower power budget potentialsArchitecture renewal for bandwidth vsenergy tradeoffBetter production yield

A growing number of FD-SOI applications for Edge computingConsumerTransportationVision processingfor autonomousdronesFacial recognitionVoice recognitionprocessor20FD-SOI Substrates for Edge ComputingSmart sensors foragricultureVision processingfor ADASWearablesDec. 2020IndustrialSmart metersIndustrial robotsMCUs for automotiveSoitec proprietary

Summary – FD-SOI for Edge computingDec. 20201Efficiency Edge computing reduces network complexity, planning but requires lowered energy per frame2FD-SOI Natural platform for Edge: Energy & Cost Efficiencies, Robustness vs. Environmental factors3Scaling Moore’s law improves gate-density, peak performance but not end-to-end Edge architectures4Challenges Consistent Edge experience requires predictable performance – and reliable technology5Strategy Soitec advancing FD-SOI to lower power potentials and planar nodes, extending the Edge21FD-SOI Substrates for Edge ComputingSoitec proprietary

Edge computing – GlossaryDec. 202022ADASAdvanced Driving Assistance SystemsASICApplication Specific Integrated CircuitBody-biasBody bias is a technique used to dynamically adjust the threshold voltage of a CMOS transistorCPUCentral Processing UniteMRAMembedded Magnetic Random Access MemoryeNVMembedded Non-Volatile MemoryFD-SOIFully Depleted Silicon on InsulatorFPGAField-Programmable Gate ArrayGPUGraphics Processing UnitInferenceApplying a deep learning model to make predictions on a new dataMACMultiplier-accumulatorMCUMicrocontroller UnitmmWmillimeter WaveMRAMMagnetic Random Access MemoryPCMPhase Change MemoryPVTProcess-Voltage-TemperatureTOPSTrillions or Tera Operations per SecondTrainingCreating a deep learning model from a datasetTPUTensor Processing UnitsVGGVGG is a convolutional neural network model proposed by K. Simonyan and A. ZissermanFD-SOI Substrates for Edge ComputingSoitec proprietary

Thank youFollow us on:Soitec@Soitec FR / @Soitec ENSoitecwww.soitec.comDec. 202023FD-SOI Substrates for Edge ComputingSoitec proprietary

12 Edge computing -Training vs. Inference Training Creating AI model from existing data Inference Applying AI model to new data New data into trained model Dec. 2020 FD-SOI Substrates for Edge Computing Soitec proprietary Source: Nvidia, Soitec Evolution Cloud Cloud and Edge Cloud and Edge On-device Edge Limitations Power ceiling due to thermal inefficiency Energy limited due to .

Related Documents:

Bruksanvisning för bilstereo . Bruksanvisning for bilstereo . Instrukcja obsługi samochodowego odtwarzacza stereo . Operating Instructions for Car Stereo . 610-104 . SV . Bruksanvisning i original

10 tips och tricks för att lyckas med ert sap-projekt 20 SAPSANYTT 2/2015 De flesta projektledare känner säkert till Cobb’s paradox. Martin Cobb verkade som CIO för sekretariatet för Treasury Board of Canada 1995 då han ställde frågan

service i Norge och Finland drivs inom ramen för ett enskilt företag (NRK. 1 och Yleisradio), fin ns det i Sverige tre: Ett för tv (Sveriges Television , SVT ), ett för radio (Sveriges Radio , SR ) och ett för utbildnings program (Sveriges Utbildningsradio, UR, vilket till följd av sin begränsade storlek inte återfinns bland de 25 största

Hotell För hotell anges de tre klasserna A/B, C och D. Det betyder att den "normala" standarden C är acceptabel men att motiven för en högre standard är starka. Ljudklass C motsvarar de tidigare normkraven för hotell, ljudklass A/B motsvarar kraven för moderna hotell med hög standard och ljudklass D kan användas vid

LÄS NOGGRANT FÖLJANDE VILLKOR FÖR APPLE DEVELOPER PROGRAM LICENCE . Apple Developer Program License Agreement Syfte Du vill använda Apple-mjukvara (enligt definitionen nedan) för att utveckla en eller flera Applikationer (enligt definitionen nedan) för Apple-märkta produkter. . Applikationer som utvecklas för iOS-produkter, Apple .

FY’20 half-year report are available in French and English language through the Company and may also be consulted and downloaded on the Company’s website (www.soitec.com).The . such elements cannot be construed as a reliable indication of the Company’s future results or developments. . 4. Soitec F

Films of n-type CdTe:In have been deposited by hot-wall vacuum evaporation (HWVE) on 7059 glass substrates, BaF 2 single crystal substrates, metal (Pt, Cr, Mo, Al) coated glass substrates, and single crystal p-type CdTe substrates. Films deposited on

Basic Description Logics Franz Baader Werner Nutt Abstract This chapter provides an introduction to Description Logics as a formal language for representing knowledge and reasoning about it. It first gives a short overview of the ideas underlying Description Logics. Then it introduces syntax and semantics, covering the basic constructors that are used in systems or have been introduced in the .