![16
[AMD Official Use Only]
Open APIs
Open
Libraries
Compilers
Developer
Tools
Kernel /
Runtime
HPC
Frameworks
ISV Apps
Open-
Source
Codes
Operating
Systems
Deployment
Tools
Mgmt Tools
ML
Frameworks](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-16-320.jpg)
![17
[AMD Official Use Only]
Drivers/Runtimes
Programming
models
Libraries
Compilers & Tools
Deployment Tools
Compiler
OpenMP API HIP API OpenCL™
RedHat, CentOS, SLES & Ubuntu Device Drivers and Run-Time
BLAS FFT
RAND
SPARSE
Debugger
Profiler
ROCm Validation Suite ROCm Data Center Tool
SOLVER
TENSILE
ALUTION THRUST MIOpen
MIVisionX
Tracer
RCCL
MIGraphX PRIM
hipify
ROCm SMI](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-17-320.jpg)
![18
[AMD Official Use Only]
AMD Infinity Hub
Containerized HPC Apps and ML Frameworks
Purpose-built accelerators for HPC and AI workloads
Full range of leading OEMs/ODMs supplying AMD
Accelerated systems to HPC and AI market segments
Open software platform for developers to build
HPC applications on AMD Accelerators
Single location for researchers and data scientists to
download containerized HPC apps and ML
frameworks
Compilers, Libraries, Dev
Tools, APIs, Kernels/Runtimes
Validated, Optimized Systems & Platforms](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-18-320.jpg)
![19
[AMD Official Use Only]
DRIVING MAINSTREAM ADOPTION & ECOSYSTEM ENABLEMENT
19
EXPANDED
OPTIMIZED
ENABLING
SUPPORT FOR AMD INSTINCT™
MI200 & AMD RADEON™ PRO
W6800 GPUS
COMPILER & LIBRARY
OPTIMIZATIONS FOR HPC &
AI/ML
NEW ROCm DOCUMENTATION
PORTAL & IMPROVED DEBUG
TOOLS](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-19-320.jpg)
![20
[AMD Official Use Only]
Re-architected ROCm Documentation
Support Guides
Installation & Deployment Guides
API / SDK Documentation
Access to ROCm Learning Center
GPU programming tutorials, videos and labs
https://docs.amd.com/](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-20-320.jpg)
![21
[AMD Official Use Only]
Molecular Dynamics Academic / Research Oil & Gas / Geoscience
NAMD
LAMMPS
GROMACS
Computer Aided
Engineering (CAE)
Weather Machine Learning
Reverse Time Migration (RTM) –
miniMOD sample
SPECFEM3D (Cartesian)
SPECFEM3D (Globe)
CP2K
Quantum Espresso
NWChem
VASP
MPAS
TempoQuest AceCAST
ICON
NEMO
Chroma
MILC
GRID
TensorFlow
PyTorch
ONNX-Runtime
MLPerf
AMBER
OpenMM
Relion
Quantum Chemistry Quantum Physics
OpenFOAM® (CFD)
PYFR (CFD)
Cascade CharLES (CFD)
Ansys Mechanical (FEA)
Target availability 1H22](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-21-320.jpg)
![22
[AMD Official Use Only]
AMD INFINITY HUB ROCm™ APP CATALOG
COMMERCIAL ISVs
[LINK]
[LINK]
•
•
•
•
•
*Ansys Mechanical 2022 R2, Cascade CharLES, TempoQuest AceCAST](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-22-320.jpg)
![23
[AMD Official Use Only]
• HPC Apps: CHROMA*, CP2K*, GRID*, GROMACS*,
HACC, LAMMPS, MILC, NAMD*, OpenMM*, Relion,
SPECFEM3D (Cartesian)*, SPECFEM3D (Globe)*
• HPC Apps: AMBER*, ICON, MPAS, NWCHEM,
OpenFOAM, PYFR, QuantumEspresso, WRF, NEMO
• AI/ML: PyTorch*, TensorFlow*
• Benchmarks: HPL, NBODY
• Benchmarks: MLPerf (SSD, Resnet50, Transformer),
HPCG
Additional MI200 Support Planned for 1H22
MI200 Support Planned for 2H21
* Available on InfinityHub with MI100 support today
Performance Results for Select Apps / Benchmarks](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-23-320.jpg)
![24
[AMD Official Use Only]
AMD Instinct GPUs:
AMD Instinct™ MI210 GPU page: https://www.amd.com/en/products/server-accelerators/instinct-mi210
AMD Instinct™ MI Series Product Page: https://www.amd.com/en/graphics/instinct-server-accelerators
AMD Instinct™ HPC Solutions Page: https://www.amd.com/en/graphics/servers-instinct-mi-powered-servers
AMD Instinct™ Machine Learning Solutions Page: https://www.amd.com/en/graphics/servers-instinct-deep-learning
AMD CDNA2 Architecture: https://www.amd.com/en/technologies/cdna2
CDNA2 WP: https://www.amd.com/system/files/documents/amd-cdna2-white-paper.pdf
AMD ROCm™ open software platform:
AMD ROCm™ pages: https://www.amd.com/en/graphics/servers-solutions-rocm
AMD Infinity Hub: https://www.amd.com/en/technologies/infinity-hub
AMD Accelerator Cloud: https://www.amd.com/en/solutions/accelerated-computing
ROCm Information Portal (DOCs & Learning Ctr.): https://docs.amd.com/
HPC & AMD page: www.AMD.com/HPC
For AMD Instinct™ GPU and ROCm™ marketing assets, contact: Guy.Ludden@AMD.com or
Sydney.Freeman@AMD.com](https://image.slidesharecdn.com/universalgpuwebinar-220420232242/85/Supermicro-s-Universal-GPU-Modular-Standards-Based-and-Built-for-the-Future-24-320.jpg)
Supermicro’s Universal GPU: Modular, Standards Based and Built for the Future
- 1. Confidential Better Faster Greener™ © 2022 Supermicro
- 2. Confidential Supermicro’s Universal GPU: Modular, Standards Based and Built for the Future Josh Grossman, Principal Product Manager April, 2022 Better Faster Greener™ © 2022 Supermicro
- 3. Confidential Agenda • Introduction to AI Market • Universal GPU Systems with MI250 • Martin Huarte on AMD GPU Software Stack v
- 4. Confidential AI Market Projection • 13 trillion dollar overall market Size According to Mckinsey • AI market size (USA) will expand at a Compound Annual Growth Rate (CAGR) of 40.2% from 2021 to 2028. • 83% of companies share that having access to AI is a top priority in their business plans. v
- 5. Confidential 4/20/2022 Better Faster Greener™ © 2022 Supermicro 5 AI augmentation” will create $2.9trn of “business value” and save 6.2bn man-hours globally. A survey by McKinsey last year estimated that AI analytics could add around $13trn, or 16%, to annual global GDP by 2030. Retail and logistics stand to gain most (see chart 2). (Economist, 2022)
- 6. Confidential 4/20/2022 Better Faster Greener™ © 2022 Supermicro 6
- 7. Confidential 4/20/2022 Better Faster Greener™ © 2022 Supermicro 7
- 8. Confidential Rack Scale AI Solutions 4/20/2022 Better Faster Greener™ © 2022 Supermicro 8
- 9. Confidential 9 Summary Universal GPU Server • The Most Optimized and Flexible GPU Server Platform available today o CPU MB Support • AMD H12 Milan • Intel X12 Ice Lake o GPU Support • NVIDIA Redstone with GPU to GPU NVLink • AMD MI-250 with Infinity Fabric xGMI • Traditional PCIe Form Factor GPU • Modular Design for Flexibility • Improved Thermal Capability o Support up to 500W/700W GPU, 280W AMD CPU and 350W/400W Intel CPU • 1U Expansion Module available for all 4U Servers UBB/OAM Intel PVC Redstone AMD MI- 250 PCIe Supermicro Confidential
- 10. Confidential 4U/5U Rackmount Dual X12, X13 and H12 Processors 32 DIMM Slots Up to 10 PCIe Low Profile 5.0 Slots Up to 10 PCIe with up to 2 AIOM/OCP 3.0 NIC Slots Up to 10 Drives of 2.5” NVMe/SAS/SATA 4x 3000W Redundant Titanium (2+2) / Platinum Level Power Supplies Universal GPU Product Series 4U uGPU Universal GPU Server Performance Modular, Standards Based Modular design supports a variety of GPU technologies and configurations Supports industry leading high performance GPUs from NVIDIA, AMD and Intel. Standardize on one GPU Platform for all your data center needs Next Generation Supermicro Universal GPU Servers Subject to change without notice 10 Better Faster Greener™ © 2022 Supermicro 5U uGPU Universal GPU with 1 U Expansion Module One GPU Platform
- 11. 11 Universal Design and AMD Instinct MI250 OAM Supermicro Confidential/Internal Only • Significant HPC performance increase over competition • Also good for AI/ML workloads • 128GB HBM2e ECC Memory per OAM • GPU to GPU xGMI Infinity Fabric 2.5TB/s
- 12. CONFIDENTIAL AMD Tools & Solutions for AI/ML and HPC 4/20/2022 Better Faster Greener™ © 2021 Supermicro 12 RTM Reverse Time Migration Datacenter Tools: Profilers & Debuggers, Comm & Math Libraries, Compiler Code Reuse: ONNX Run-time, existing deep learning, HPC code Cross Platform: Open source, supports AMD CPUs, CPU, non-AMD GPUs 3RD GEN AMD INFINITY ARCHITECTURE FIRST MULTI-CHIP GPU • Highest performance • Bigger GPU memory • Higher Flops (FP64, FP32, FP16)
- 13. Confidential Better Faster Greener™ © 2021 Supermicro 13 Specifications CPU – Dual Socket Dual AMD EPYC 7003 CPUs (Socket SP3) up to 280W, 128 Cores/256 Threads Memory – 32 DIMM Slots 32 DIMM, 8TB Reg. ECC DDR4 up to 3200MHz Drives – 10 2.5” Drive-bay Up to 10x HS NVMe U.2 connect to PCIe Switch or 10x HS 2.5” SATA Expansion – 8 PCIe Slots 8x PCIe 4.0 x16 LP (via PLX switch) I/O ports 1x VGA, 1x COM Header, 2x USB 3.0, and 1x Dedicated IPMI Power Supply 4x 3000W (2+2) Titanium Level efficiency power supplies 4U AMD EPYC 7003 Dual CPUs and Four AMD MI250 GPUs Universal GPU System AMD AS -4124GQ-TNMI Subject to change without notice Key Features Universal GPU Server Standards Based Design Modular by Design for Flexibility/Future Proofed Improved Thermal Capability Key Applications Perfect Platform for HPC applications Data Center Infrastructure System Rear View System Front View Supermicro Confidential/Internal Only
- 14. Confidential Better Faster Greener™ © 2021 Supermicro 14 Specifications CPU – Dual Socket Dual AMD EPYC 7003 CPUs (Socket SP3) up to 280W, 128 Cores/256 Threads Memory – 32 DIMM Slots 32 DIMM, 8TB Reg. ECC DDR4 up to 3200MHz Drives – 10 2.5” Drive-bay Up to 10x HS NVMe U.2 connect to PCIe Switch or 10x HS 2.5” SATA Expansion – 10 PCIe Slots 8x PCIe 4.0 x16 LP (via PLX switch) 2x PCIe 4.0 x16 LP or AIOM (via CPU w/ 1U add-on) I/O ports 1x VGA, 1x COM Header, 2x USB 3.0, and 1x Dedicated IPMI Power Supply 4x 3000W (2+2) Titanium Level efficiency power supplies 5U AMD EPYC 7003 Dual CPUs and Four AMD MI250 GPUs Universal GPU System AMD AS -4124GQ-TNMI Subject to change without notice Key Features Universal GPU Server Standards Based Design Modular by Design for Flexibility/Future Proofed Improved Thermal Capability Key Applications Perfect Platform for HPC applications Data Center Infrastructure System Rear View System Front View Supermicro Confidential/Internal Only
- 15. Driving Innovation and Discovery with AMD Instinct™ accelerators on ROCm™ Stack Martin Huarte, Ph.D. Developer Relations Manager, martin.huarte@amd.com
- 16. 16 [AMD Official Use Only] Open APIs Open Libraries Compilers Developer Tools Kernel / Runtime HPC Frameworks ISV Apps Open- Source Codes Operating Systems Deployment Tools Mgmt Tools ML Frameworks
- 17. 17 [AMD Official Use Only] Drivers/Runtimes Programming models Libraries Compilers & Tools Deployment Tools Compiler OpenMP API HIP API OpenCL™ RedHat, CentOS, SLES & Ubuntu Device Drivers and Run-Time BLAS FFT RAND SPARSE Debugger Profiler ROCm Validation Suite ROCm Data Center Tool SOLVER TENSILE ALUTION THRUST MIOpen MIVisionX Tracer RCCL MIGraphX PRIM hipify ROCm SMI
- 18. 18 [AMD Official Use Only] AMD Infinity Hub Containerized HPC Apps and ML Frameworks Purpose-built accelerators for HPC and AI workloads Full range of leading OEMs/ODMs supplying AMD Accelerated systems to HPC and AI market segments Open software platform for developers to build HPC applications on AMD Accelerators Single location for researchers and data scientists to download containerized HPC apps and ML frameworks Compilers, Libraries, Dev Tools, APIs, Kernels/Runtimes Validated, Optimized Systems & Platforms
- 19. 19 [AMD Official Use Only] DRIVING MAINSTREAM ADOPTION & ECOSYSTEM ENABLEMENT 19 EXPANDED OPTIMIZED ENABLING SUPPORT FOR AMD INSTINCT™ MI200 & AMD RADEON™ PRO W6800 GPUS COMPILER & LIBRARY OPTIMIZATIONS FOR HPC & AI/ML NEW ROCm DOCUMENTATION PORTAL & IMPROVED DEBUG TOOLS
- 20. 20 [AMD Official Use Only] Re-architected ROCm Documentation Support Guides Installation & Deployment Guides API / SDK Documentation Access to ROCm Learning Center GPU programming tutorials, videos and labs https://docs.amd.com/
- 21. 21 [AMD Official Use Only] Molecular Dynamics Academic / Research Oil & Gas / Geoscience NAMD LAMMPS GROMACS Computer Aided Engineering (CAE) Weather Machine Learning Reverse Time Migration (RTM) – miniMOD sample SPECFEM3D (Cartesian) SPECFEM3D (Globe) CP2K Quantum Espresso NWChem VASP MPAS TempoQuest AceCAST ICON NEMO Chroma MILC GRID TensorFlow PyTorch ONNX-Runtime MLPerf AMBER OpenMM Relion Quantum Chemistry Quantum Physics OpenFOAM® (CFD) PYFR (CFD) Cascade CharLES (CFD) Ansys Mechanical (FEA) Target availability 1H22
- 22. 22 [AMD Official Use Only] AMD INFINITY HUB ROCm™ APP CATALOG COMMERCIAL ISVs [LINK] [LINK] • • • • • *Ansys Mechanical 2022 R2, Cascade CharLES, TempoQuest AceCAST
- 23. 23 [AMD Official Use Only] • HPC Apps: CHROMA*, CP2K*, GRID*, GROMACS*, HACC, LAMMPS, MILC, NAMD*, OpenMM*, Relion, SPECFEM3D (Cartesian)*, SPECFEM3D (Globe)* • HPC Apps: AMBER*, ICON, MPAS, NWCHEM, OpenFOAM, PYFR, QuantumEspresso, WRF, NEMO • AI/ML: PyTorch*, TensorFlow* • Benchmarks: HPL, NBODY • Benchmarks: MLPerf (SSD, Resnet50, Transformer), HPCG Additional MI200 Support Planned for 1H22 MI200 Support Planned for 2H21 * Available on InfinityHub with MI100 support today Performance Results for Select Apps / Benchmarks
- 24. 24 [AMD Official Use Only] AMD Instinct GPUs: AMD Instinct™ MI210 GPU page: https://www.amd.com/en/products/server-accelerators/instinct-mi210 AMD Instinct™ MI Series Product Page: https://www.amd.com/en/graphics/instinct-server-accelerators AMD Instinct™ HPC Solutions Page: https://www.amd.com/en/graphics/servers-instinct-mi-powered-servers AMD Instinct™ Machine Learning Solutions Page: https://www.amd.com/en/graphics/servers-instinct-deep-learning AMD CDNA2 Architecture: https://www.amd.com/en/technologies/cdna2 CDNA2 WP: https://www.amd.com/system/files/documents/amd-cdna2-white-paper.pdf AMD ROCm™ open software platform: AMD ROCm™ pages: https://www.amd.com/en/graphics/servers-solutions-rocm AMD Infinity Hub: https://www.amd.com/en/technologies/infinity-hub AMD Accelerator Cloud: https://www.amd.com/en/solutions/accelerated-computing ROCm Information Portal (DOCs & Learning Ctr.): https://docs.amd.com/ HPC & AMD page: www.AMD.com/HPC For AMD Instinct™ GPU and ROCm™ marketing assets, contact: Guy.Ludden@AMD.com or Sydney.Freeman@AMD.com
- 25. Confidential Thank You 25 Better Faster Greener™ © 2022 Supermicro Please Contact us for Details: Josh Grossman, Principal Product Manager, Supermicro joshg@supermicro.com Martin Huarte, Ph.D., Developer Relations Manager, AMD martin.huarte@amd.com
- 26. Confidential DISCLAIMER Super Micro Computer, Inc. may make changes to specifications and product descriptions at any time, without notice. The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors. Any performance tests and ratings are measured using systems that reflect the approximate performance of Super Micro Computer, Inc. products as measured by those tests. Any differences in software or hardware configuration may affect actual performance, and Super Micro Computer, Inc. does not control the design or implementation of third party benchmarks or websites referenced in this document. The information contained herein is subject to change and may be rendered inaccurate for many reasons, including but not limited to any changes in product and/or roadmap, component and hardware revision changes, new model and/or product releases, software changes, firmware changes, or the like. Super Micro Computer, Inc. assumes no obligation to update or otherwise correct or revise this information. SUPER MICRO COMPUTER, INC. MAKES NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE CONTENTS HEREOF AND ASSUMES NO RESPONSIBILITY FOR ANY INACCURACIES, ERRORS OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION. SUPER MICRO COMPUTER, INC. SPECIFICALLY DISCLAIMS ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE. IN NO EVENT WILL SUPER MICRO COMPUTER, INC. BE LIABLE TO ANY PERSON FOR ANY DIRECT, INDIRECT, SPECIAL OR OTHER CONSEQUENTIAL DAMAGES ARISING FROM THE USE OF ANY INFORMATION CONTAINED HEREIN, EVEN IF SUPER MICRO COMPUTER, Inc. IS EXPRESSLY ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. ATTRIBUTION © 2022 Super Micro Computer, Inc. All rights reserved. 4/20/2022 Better Faster Greener™ © 2021 Supermicro 26
Editor's Notes
- #18: Heterogeneous compute interface for portability
- #22: MLPerf 0.7 (Resnet 50, Transformer, SSD)
- #25: Canned questions: Is HIP a drop-in replacement for CUDA? No. HIP provides porting tools which do most of the work to convert CUDA code into portable C++ code that uses the HIP APIs. Most developers will port their code from CUDA to HIP and then maintain the HIP version. HIP code provides the same performance as native CUDA code, plus the benefits of running on AMD platforms. What APIs and features does HIP support? HIP provides the following: Devices (hipSetDevice(), hipGetDeviceProperties()) Memory management (hipMalloc(), hipMemcpy(), hipFree()) Streams (hipStreamCreate(),hipStreamSynchronize(), hipStreamWaitEvent()) Events (hipEventRecord(), hipEventElapsedTime()) Kernel launching (hipLaunchKernel is a standard C/C++ function that replaces <<< >>>) HIP Module API to control when adn how code is loaded. CUDA-style kernel coordinate functions (threadIdx, blockIdx, blockDim, gridDim) Cross-lane instructions including shfl, ballot, any, all - Most device-side math built-ins. Error reporting (hipGetLastError(), hipGetErrorString()) The HIP API documentation describes each API and its limitations, if any, compared with the equivalent CUDA API. https://rocmdocs.amd.com/en/latest/Programming_Guides/HIP-FAQ.html#what-apis-and-features-does-hip-support