IBM Power Systems POWER9 Processor/Server and IBM i Update Daniel R Sundt Partner Technical Advocate BP Technical Sales & Enablement 3605 Highway 52 North IBM Power Systems Rochester, MN 55901 Tel 507 253 3228 Mobile 507 261 5329 [email protected] John J. Nelson Tim Spencer Client Systems Manager Client Technical Sales Industrial Market IBM Power Systems Industrial Market Mobile 612 414 9272 Mobile 414 870 0832 [email protected] [email protected] IBM Power Systems Strategy Hybrid Cloud Innovation: Resilient, Scalable, & Secure: Application Modernization: Agility, Flexibility & Automation Architectural Strength & Enterprise AI, Advanced On-prem and Public Cloud Superiority Analytics, Containerization, Extension via Microservices AIX IBM i Enterprise Linux IBM Public Cloud Extend resilient, scalable and Extend the most integrated Deliver optimized infrastructure Leverage Power Systems in the IBM mission critical Unix environment operating system with for workload modernization and Public Cloud across AIX/IBM i and with new innovative on-prem and innovative resiliency platform growth on Power Systems Enterprise Linux as clients adopt public cloud offerings with in Db2 Mirror, open source through SAP HANA, SAS Viya, mission-critical workloads in public consistent industry automation. languages and AI integration. OpenShift & Cloud Paks. clouds. Built on the Innovative POWER Processor Roadmap in conjunction with stack integration across Hardware Systems and System Software POWER9 14nm POWER10 7nm POWER11 © Copyright IBM Corporation 2020 IBM POWER Processor Roadmap POWER10 POWER7+ POWER9 POWER8 14 nm 32 nm 22 nm POWER7 POWER8NV 2+-4+ GHz 45 nm 3-4+ GHz POWER6+ 22 nm POWER6 65 nm 3-4+GHz POWER5™ POWER5+ 65 nm 90 nm . Enhanced Thread Extreme 130 nm 3-5GHz performance Analytics 1.5 – 2.3GHz . Multi Core (up to 8) Multi Core (up to 12) . Analytics . Dual Core . On-Chip eDRAM (L3) Double L1 data cache Optimization . High Frequencies . On Chip eDRAM (L3) . Dual Core . Power Optimized . Extreme Big Data . Virtualization + . L4 on Memory . Enhanced Scaling Cores Optimization . Memory Subsystem + . cCards . SMT 2 Mem Subsystem ++ . Next Gen Memory . Altivec . Bandwidth +++ . Distributed Switch + SMT 4 . Instruction Retry . SMT 8 . On-chip accelerators . Core Parallelism + Reliability + . Dynamic Energy . Reliability ++ . Open CAPI . FP Performance + VSM & VSX (AltiVec) . Mgmt . On Chip Power Mgmt . NVLink . Memory bandwidth + Protection Keys+ . SMT 2+ PCIe Gen 3 . BW enhancements . Virtualization . Protection Keys . Transistors 1.2B . Transistors 4B+ . Transistors 790M . POWER7+ 2.1B . Transistors 276M Transistors 8B 2004 2007 2010 2012 2014 2016 2017 2020 2021 2020 IBM Systems Technical University 3 © Copyright IBM Corporation 2020 POWER9 Increases Performance Over Prior Generations 2020 IBM Systems Technical University 4 © Copyright IBM Corporation 2020 POWER9 Increases Performance Over Prior Generations POWER9 vs x86 Xeon SP (Skylake) Performance per st ~ TB/s 1 1 chip 1 2x core BW into with PCIe4 chip 2 4GHZ 7TB/s 2.6x RAM per socket PEAK On chip BW FREQUENCY Memory bandwidth >15 3 >24B per socket MILES OF VIAS 1.8x WIRE 8 17 POWER9 with NVLink vs x86 Xeon BILLION LEVELS TRANSISTORS OF METAL 4 CPU to accelerator 9.5x bandwidth (1) 2X performance per core is based on IBM Internal measurements as of 2/28/18 on various system configuration and workload environments including (1) Enterprise Database (2.22X per core): 20c S922L (2x10-core/2.9 GHz/256 GB memory): 1,039,365 Ops/sec versus 2-socket Intel Xeon Skylake Gold 6148 (2x20-core/2.4 GHz/256 GB memory): 932,273 Ops/sec. (2) DB2 Warehouse (2.43X per core): 20c S922L (2x10-core/2.9 GHz/512 GB memory): 3242 QpH versus 2-socket Intel Xeon Skylake Platinum 8168 (2x24-core/2.7 GHz/512 GB memory): 3203 QpH. (3) DayTrader 7 (3.19X per core): 24c S924 (2x12-core/3.4 GHz/512 GB memory): 32221.4 tps versus 2-socket Intel Xeon Skylake Platinum 8180 (2x28-core/2.5 GHz/512 GB memory): 23497.4 tps. (2) 2.6X memory capacity is based on 4TB per socket for POWER9 and 1.5TB per socket for x86 Scalable Platform Intel product brief: https://www.intel.com/content/dam/www/public/us/en/documents/product-briefs/xeon-scalable-platform-brief.pdf?asset=14606 (3) 1.8X bandwidthy is based on 230 GB/sec per socket for POWER9 and 128GB/sec per socket for x86 Scalable Platform Intel product brief: https://www.intel.com/content/dam/www/public/us/en/documents/product-briefs/xeon-scalable-platform-brief.pdf?asset=14606 (4) 9.5X is based on POWER9 and next-generation NVIDIA NVLink peak transfer rate is 150 GB/sec = 48 lanes x 3.2265625 GB/sec x 64 bit/66 bit encoding compared to x86 PCI Express 3.0 (x16) peak transfer rate is 15.75 GB/sec = 16 lanes X 1GB/sec/lane x 128 bit/130 bit encoding. 2020 IBM Systems Technical University 5 ©© Copyright Copyright IBM IBM Corporation Corporation 2020 2020 POWER9 Systems Forbes: “IBM Power Systems play where reliability, longevity, and performance are key.” https://www.forbes.com/sites/patrickmoorhead/2018/02/20/ibm-targets-installed-base-with-second-phase-of-power9-systems/#1064fdb43341 2020 IBM Systems Technical University 6 © Copyright IBM Corporation 2020 IBM Power Systems POWER9 Family POWER9 servers and solutions are built to crush today’s most advanced data applications … from the mission critical applications you run today to the next generation of AI workloads 2017-2020 POWER9 Mission Critical Workloads Big Data Workloads Enterprise AI Workloads LC922/LC921 2Q 2018 S922/S914/S924 E950 E980 AC922 H922/H924/L922 3Q 2018 3Q 2018 Q4 2017 1Q 2018 IC922 1Q 2020 Next Gen AI Core Infrastructure Workloads 7 2020 IBM Systems Technical University 7 © Copyright IBM Corporation 2020 A Portfolio for the AI Era From Mission-Critical workloads to AI and Cloud Computing leadership Accelerated Data, Inferencing, Compute and Cloud PowerVM and Big Data high RAS IC922 AC922 NEW! LC922 / LC921 L922 • Industry first and only in • Storage dense, high • Big Data server with up to advanced IO with 2nd bandwidth server – up to 24 • Industry leading reliability 1 120TB storage capacity, large Generation CPU - GPU NVMe or SAS/SATA in 2U and computing capability form factor support, KVM NVLink delivering ~5.6x • Advanced IO with PCIe Gen4 support, leveraging P9 compute • PowerVM ecosystem focus higher data throughput for outstanding utilization • Optimized inferencing server for a composable design • Up to 4 integrated NVIDIA • Focus on memory capacity with up to 6 Nvidia T4 GPUs at • 1U and 2U form factors with up to 4TB of RAM “Volta” GPUs air cooled GA and additional accelerators (GTH) and up to 6 GPUs with in roadmap1 • Advanced IO with PCIe water cooled (GTX) version 4.0/CAPI 2.0 • OpenCAPI support1 • OpenCAPI support • Price/performance server • Up to 44 cores (2U) or 40 cores • Memory coherence (1U) at lower frequency © Copyright IBM Corporation 2020 POWER9 Servers (Highlights) Full Portfolio refresh (started in 2017) CPU SCM packaging for all 2-socket and 4-socket systems Eliminates SW licensing issues associated with DCM designs Lowers latency for CPU to CPU xfers due to simpler CPU fabric topology Up to 4x increased CPU fabric bandwidth for max scalability Dynamic CPU Frequency enables higher frequencies for typical workloads (EnergyScale) Increased Memory capacities over POWER8 (2x and 4x) Leveraging IS DIMMs to provide more competitive offerings in 2-socket and 4-socket space Increased I/O bandwidth with PCIe GEN4 Slots and future PCIe GEN4 Expansion Drawer 25Gb ports for High Speed GPU/OpenCAPI acceleration Integrated NVMe Flash device support Basic form factors and power requirements remain the same External DVD 2020 IBM Systems Technical University 9 © Copyright IBM Corporation 2020 IBM Power Systems POWER9 Scale-Out Family Scale Out for AIX, IBM i, and Linux Built for HANA Optimized for Linux S922 S914 S924 H922 H924 L922 9009-22A 9009-41A 9009-42A 9223-22H 9223-42H 9008-22L • 1,2-socket, 2U • 1-s, 4U & Tower • 2-socket, 4U • 1,2-socket, 2U • 2-socket, 4U • 1,2-socket, 2U • 4, 8,10 cores/skt • 4,6,8 cores/skt • 8,10,11,12 • 4, 8,10 cores/skt • 8,10,12 cores/skt • 8,10,12 cores/skt • 32 IS DIMM slots • 16 IS DIMM slots cores/skt • 32 IS DIMM slots • 32 IS DIMM slots • 32 IS DIMM slots • 4TB memory • 1TB memory • 32 IS DIMM slots • 4TB memory • 4TB memory • 4TB memory • Internal RDX • 4TB memory • Internal RDX Media • Internal RDX Media Media • PowerVM • PowerVM • PowerVM • PowerVM • PowerVM • PowerVM • AIX, IBM i up to 25% • AIX, IBM i up to 25% • Linux • Linux Technology • Cloud enabled - Embedded virtualization capabilities with PowerVM Leadership • Up to 4TB in 2 socket - DDR4 Industry Standard memory RDIMMs • High Speed 25Gb/s external ports – one per socket • Embedded Analytics and Algorithms on the chip help run POWER9 at an always optimized frequency 2020 IBM Systems Technical University 10 © Copyright IBM Corporation 2020 POWER9 4U Scale Out Server S924 9009-42A, H924 9223-42H 4U server - 19” Rack enclosure POWER9 Scale-Out SMT8 processor (12-core, 11-core, 10-core, 8-core offerings) Up to 4TB Total DDR4 Industry Standard memory RDIMMs . Up to 340 GB/s total system memory bandwidth . 32 IS RDIMM slots (no risers) 11 PCIe Gen3/Gen4 slots, Full Height, Half Length (2-sockets populated) . Five PCIe GEN4 slots (4 CAPI 2.0 enabled) . Six PCIe GEN3 slots (1 reserved for Ethernet adapter) 2 High Speed 25Gb/s ports for OpenCAPI / GPU Acceleration 12 or 18 SFF (2.5”) bay options Two internal storage controller slots . Single or Split backplane or Dual RAID write