Persistent Memory Productization driven by AI & ML Danny Sabour VP Marketing, Avalanche Technology
Persistent Memory Usage from Cloud to Node CLOUD Compute Storage Deep Learning Training Big data processing Data Warehousing EDGE Data Visualization Data caching Data filtering MES ERP SCADA Systems IoT NODE Inference Realtime Data processing 2
Node Architecture : From Data Fusion to AI Sense V2X CoProc Radar CoProc Ladar CoProc UltraS CoProc Sensor Fusion Think Processor Processor Transmit + ACT Neural Processor The Node: Collects Data Aggregates data Sends data back to Edge Make decisions per input Learn based on local stimuli Camera Motion Speed CoProc NV Store Needs Persistence Low Power Smallest die size Inference: Data Fusion Local Decision at the Node Making 3
IoT Roadmap ` Embedded IoT Memory 28nm NVM MACRO 8-64Mb 28nm SRAM MACRO 8-64Mb 28nm NVM MACRO 64-256Mb 14nm SRAM MACRO 8-64Mb Discrete IoT Memory Gen 1 SPI 8Mb Gen 2 QSPI 32Mb Gen 3 OSPI 256Mb Technology 4
Edge/Cloud Architecture CPU Core 1 Processing ProcessorCPU Core 2 CPU Core 3 Core X PCH Storage Cache NVDIMM-N AI/Machine Learning GPU PCIe Cache The Cloud and the Edge Architecture are similar, but with the possible variations in: GPU Cores which can be used for Machine learning PCIe Cache for performance Communication to other domains in the edge for inference and learning Needs Persistence Speed Storage Training : Learning New Capabilities 5
Speeding up the Cloud/Edge Computing: Cache & GPU Memory CPU Core 1 Processing ProcessorCPU Core 2 CPU Core 3 Core X PCH Storage Cache Storage AI/Machine Learning GPU PCIe Cache Bus Interface Controller SSD Medium Density Persistent RAM (MD-RAM) Bus Interface 6
Universal Memory in the Cloud: NVDIMM NVDIMM-P NVDIMM-N AVA-DIMM: Persistent replacement CPU Core 1 Processing ProcessorCPU Core 2 CPU Core 3 Core X NVDIMM-P: Hides Persistent Memory Endurance Controller + Battery + Cache Cache PCH AI/Machine Learning GPU PCIe Cache NVDIMM-N: Brought Persistence to Controller + Battery + Flash Storage Cache Storage High Density Persistent RAM (HD-RAM) 7
High and Mid Density Persistent RAM High-Density Persistent RAM Gen 3 HD-AVRAM DDR 8-32Gb DDR 64Gb Mid-Density Persistent RAM DDR 4-8Gb DDR 4-16Gb DDR 1-4Gb DDR 4-8Gb Gen 1 MD-AVRAM Gen 2 MD-AVRAM Technology 1S1R 1T1R Regular Small Tiny Micro 8
Where does MRAM fit into the picture Endurance (Cycles) 10 18 10 15 10 12 10 9 10 6 10 3 1 SRAM E-SRAM MD-RAM HD-RAM SPnvSRAM PCM RRAM E-NVM NOR HDD NAND 10-9 10-8 10-7 10-6 10-5 10-4 10-3 Write Speed 1 Magnetic structure Embedded: High Performance High Temperature Stand Alone: IoT Avalanche Specification for different MRAMs available at: www.avalanche-technology.com/products Med Density RAM High Density RAM 9
Summary MRAM products are making inroads from the cloud to the IoT Node It is not one product MRAM is unifying the Software defined world Compute Storage Network Built on a Unified CPU & Memory platform based on a unified MTJ architecture for Compute, Storage, Network Across the Cloud, Edge, Node Phased product introduction has begun 10