AMD Instinct MI210 Accelerator

The ThinkSystem AMD Instinct MI210 Accelerator is a compute workhorse optimized for accelerating single precision and double-precision HPC-class systems. The accelerator can also be deployed for training large-scale machine intelligence workloads. The accelerator's powerful compute engine, new matrix math FP64 cores, and advanced memory architecture, combined with AMD's ROCm open software platform and ecosystem, provide a powerful, flexible heterogeneous compute solution designed to help datacenter designers meet the challenges of a new era of compute.

Figure 1. AMD Instinct MI210 PCIe Gen4 Passive Accelerator

A visual representation of the AMD Instinct MI210 PCIe Gen4 Passive Accelerator, showing its physical form factor and key components.

Did you know? ❓

The ThinkSystem SR670 V2 supports 8x MI210 Accelerators, which can be connected together using 2x Infinity Fabric Link Bridge Cards, linking two sets of four GPUs. Each accelerator has three Infinity Fabric links with 300 GB/s of Peer-to-Peer (P2P) bandwidth performance. The AMD Infinity Architecture enables platform designs with direct-connect GPU hives with high-speed P2P connectivity, delivering up to 1.2 TB/s of total theoretical GPU bandwidth.

Part Number Information

The following table shows the part numbers for the MI210 Accelerator.

Table 1. Ordering Information

Part NumberFeature CodeDescription
4X67A81102BP04ThinkSystem AMD Instinct MI210 PCIe Gen4 Passive Accelerator
4X67A82325BRMDThinkSystem AMD Instinct MI210 4x Infinity Fabric Link Bridge Card

The PCIe option part number includes:

Each Infinity Fabric Link Bridge Card connects four MI210 Accelerators.

Features

The AMD Instinct MI210 accelerator offers the following features:

Exascale-Class Technologies for the Data Center

The AMD Instinct MI210 accelerator extends AMD's industry performance leadership in accelerated compute for double precision (FP64) on PCIe form factors for mainstream HPC and AI workloads in the data center. Built on AMD Exascale-class technologies with the 2nd Gen AMD CDNA architecture, the MI210 enables scientists and researchers to tackle pressing challenges from climate change to vaccine research. MI210 accelerators, combined with the AMD ROCm 5 software ecosystem, allow innovators to tap the power of HPC and AI data center PCIe GPUs to accelerate their time to science and discovery.

Purpose-built Accelerators for HPC & AI Workloads

Powered by the 2nd Gen AMD CDNA architecture, the AMD Instinct MI210 accelerator delivers HPC performance leadership in FP64 for a broad set of HPC & AI applications. The MI210 accelerator is built to accelerate deep learning training, providing an expanded range of mixed-precision capabilities based on AMD Matrix Core Technology, and delivers an outstanding 181 teraflops peak theoretical FP16 and BF16 performance to fuel the convergence of HPC and AI.

Innovations Delivering Performance Leadership

AMD innovations in architecture, packaging, and integration are pushing the boundaries of computing by unifying the most important processors in the data center: the CPU and the GPU accelerator. With innovative double-precision Matrix Core capabilities and the 3rd Gen AMD Infinity Architecture, AMD is delivering performance, efficiency, and overall system throughput for HPC and AI using AMD EPYC CPUs and AMD Instinct MI210 accelerators.

Ecosystem without Borders

AMD ROCm is an open software platform allowing researchers to tap the power of AMD Instinct accelerators to drive scientific discoveries. The ROCm platform is built on open portability, supporting environments across multiple accelerator vendors and architectures. With ROCm 5, AMD extends its platform powering top HPC and AI applications with AMD Instinct MI200 series accelerators, increasing ROCm accessibility for developers and delivering outstanding performance across key workloads.

2nd Generation AMD CDNA Architecture

The AMD Instinct MI210 accelerator brings commercial HPC & AI customers the compute engine selected for the first U.S. Exascale supercomputer. Powered by the 2nd Gen AMD CDNA architecture, the MI210 accelerator delivers outstanding performance for HPC and AI. The MI210 PCIe GPU delivers superior double and single precision performance for HPC workloads with up to 22.6 TFLOPS peak FP64|FP32 performance, enabling scientists and researchers globally to process HPC parallel codes more efficiently across several industries.

AMD's Matrix Core technology delivers a broad range of mixed precision operations, enabling work with large models and enhancing memory-bound operation performance for AI and machine learning workloads. The MI210 offers optimized BF16, INT4, INT8, FP16, FP32, and FP32 Matrix capabilities, providing supercharged compute performance for AI system requirements. The AMD Instinct MI210 accelerator handles large data efficiently for training and delivers 181 teraflops of peak FP16 and bfloat16 floating-point performance for deep learning training.

AMD Infinity Fabric Link Technology

AMD Instinct MI210 GPUs provide advanced I/O capabilities in standard off-the-shelf servers with AMD Infinity Fabric technology and PCIe Gen4 support. The MI210 GPU delivers 64 GB/s CPU to GPU bandwidth without PCIe switches, and up to 300 GB/s of Peer-to-Peer (P2P) bandwidth performance through three Infinity Fabric links. The AMD Infinity Architecture enables platform designs with dual and quad, direct-connect, GPU hives with high-speed P2P connectivity, delivering up to 1.2 TB/s of total theoretical GPU bandwidth within a server design. Infinity Fabric helps unlock accelerated computing, enabling a quick and simple onramp for CPU codes to accelerated platforms.

Ultra-Fast HBM2e Memory

AMD Instinct MI210 accelerators provide up to 64GB High-bandwidth HBM2e memory with ECC support at a clock rate of 1.6 GHz, delivering an ultra-high 1.6 TB/s of memory bandwidth to support large data sets and eliminate bottlenecks moving data in and out of memory. Combining this performance with the MI210's advanced Infinity Fabric I/O capabilities allows workloads to be pushed closer to their full potential.

Technical Specifications

The following table lists the MI210 Accelerator specifications.

Table 2. Technical Specifications

FeatureSpecification
Compute units104
Stream Processors:6,656
GPU Memory:64GB HBM2e
ECCYes
Memory Bandwidth:1.6 TB/s
Peak FP64 Vector23 TFLOPS
Peak FP32 Vector23 TFLOPS
Peak FP64 Matrix45 TFLOPS
Peak FP32 Matrix45 TFLOPS
Peak FP16181 TFLOPS
Peak BF16181 TFLOPS
Bus Interface:PCIe Gen 4 x16
AMD Infinity Fabric Link support:4 GPU (4-GPU link support by using part number 4X67A82325)
Board Form Factor:Full-Height, Full-Length, Dual-Slot
Thermal Solution:Passively Cooled
Standard Max Power:300W TDP
OS Support:Linux 64-bit
ROCm Software Platform:ROCm 5.0
Programming Environment:OpenMP, OpenCL, ISO C++ (via HIP conversion tool), CUDA (via HIP conversion tool)

Server Support

The following tables list the ThinkSystem servers that are compatible.

Table 3. Server Support (Part 1 of 3)

Part NumberDescriptionEdge1S Intel V2AMD V3Intel V3
4X67A81102ThinkSystem AMD Instinct MI210 PCIe Gen4 Passive AcceleratorNNNN
4X67A82325ThinkSystem AMD Instinct MI210 4x Infinity Fabric Link Bridge CardNNNN
ZZ33
Z882
ZN4N
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN
ZNNN

Table 4. Server Support (Part 2 of 3)

Part NumberDescriptionDense V32S Intel V2AMD V1Dense V24S V28S
4X67A81102ThinkSystem AMD Instinct MI210 PCIe Gen4 Passive AcceleratorNNNNNN
4X67A82325ThinkSystem AMD Instinct MI210 4x Infinity Fabric Link Bridge CardNNNNNN
ZZ3NNN
Z8NNN
N21NNN
YNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN
NNNNN

1. Supported only with EPYC 7003 "Milan" processors. Not supported with EPYC 7002 "Rome" processors.

Table 5. Server Support (Part 3 of 3)

Part NumberDescription4S V11S Intel V12S Intel V1Dense V1
4X67A81102ThinkSystem AMD Instinct MI210 PCIe Gen4 Passive AcceleratorNNNN
4X67A82325ThinkSystem AMD Instinct MI210 4x Infinity Fabric Link Bridge CardNNNN
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ
ZZZZ

Operating System Support

The following table lists the supported operating systems.

Tip: ? These tables are automatically generated based on data from Lenovo ServerProven.

Table 6. Operating System Support for ThinkSystem AMD Instinct MI210 PCIe Gen4 Passive Accelerator, 4X67A81102

Operating SystemsSR650 V3SR655 V3SR665 V3SR675 V3SR850 V3SR860 V3SR650 V2SR670 V2SR655SR665
Red Hat Enterprise Linux 7.9NNNNNNYYYY
Red Hat Enterprise Linux 8.3NNNNNNYYYY
Red Hat Enterprise Linux 8.4NNNNNNYYYY
Red Hat Enterprise Linux 8.5NNNNNNYYYY
Red Hat Enterprise Linux 8.6NNNNNNYYYY
Red Hat Enterprise Linux 9.0NNNNNNYYYY
SUSE Linux Enterprise Server 15 SP3YYYYYYYYYY
SUSE Linux Enterprise Server 15 SP4YYYYYYYYYY
Ubuntu 18.04.5 LTSNNNNNNNNNN
Ubuntu 20.04 LTSYYYYYYYYYY
Ubuntu 20.04.5 LTSYYYYYYYYYY
Ubuntu 22.04 LTSYYYYYYYYYY

1 HW is not supported with EPYC 7002 processors.

Auxiliary Power Cables

The MI210 Accelerator option part number does not ship with auxiliary power cables.

Table 7. Auxiliary Power Cables for MI210

DescriptionDetails
350mm 8pin (2x4) cableServer support: SR655 (Riser 1 or Riser 2)
Option: 4X97A59853, ThinkSystem SR655 GPU Cable Kit
Feature: B5T5
SBB: SBB7A14640
Base: SC17A50848
FRU: 02JK011
250mm 8pin (2x4) cableServer support: SR655 (Riser 3)
Option: 4X97A59853, ThinkSystem SR655 GPU Cable Kit
Feature: B5TS
SBB: SBB7A10974
Base: SC17A50844
FRU: 02JK010
360mm 8pin (2x4) cableAuxiliary power cables needed with the SR650 V3, SR655 V3, SR665 V3, SR650 V2, SR665
Option*:
  • SR665: 4M17A80478 / 4M17A11759, ThinkSystem SR665 GPU Thermal Option Kit v2
  • SR650 V2: 4H47A38666 / 4H47A80491, ThinkSystem SR650 V2 GPU Full Length Thermal Option Kit v2
  • SR650 V3: 4X67A82883, ThinkSystem SR650 V3 GPU Full Length Thermal Option Kit
  • SR655 V3: 4X67A86438, ThinkSystem SR655 V3 GPU Enablement Kit
  • SR665 V3: 4X67A85856, ThinkSystem SR665 V3 GPU Full Length Thermal Option Kit
Feature: BAD8
SBB: SBB7A49792 or SBB7A21691
Base: SC17A95312 or SC17A59596
FRU: 03HA297 or 02YE420
215mm 8pin (2x4) cableAuxiliary power cable needed with the SR670 V2
Option: 4X97A69627, ThinkSystem SR670 V2 GPU Option Power Cable
Feature: BFNL
SBB: SBB7A40281
Base: SC17A10990
FRU: 02YE945
235mm 8pin (2x4) cableAuxiliary power cable needed with the SR675 V3
Option: 4X97A84509, ThinkSystem SR675 V3 GPU Power Cable Kit
Feature: BRGU
SBB: SBB7A65297
Base: SC17B39302
FRU: 03LE555
200mm 8pin (2x4) cableAuxiliary power cable needed with the SR850 V3 or SR860 V3
Option: 4X97A88017, ThinkSystem SR850 V3/SR860 V3 A100/A6000/MI210 GPU Power Cable Option Kit
Feature: BTPB
SBB: SBB7A72760
Base: SC17B40606
FRU: 03LF917

*The option part numbers are for thermal kits and include other components needed to install the GPU. See the server product guide for details.

Regulatory Approvals

The MI210 Accelerator has the following regulatory approvals:

Electromagnetic Compliance

  • Australia and New Zealand: CISPR 32: 2015 +COR1: 2016, Class A
  • Canada ICES-003, Issue 7, Class A
  • European Countries: EN 55032: 2015 + A11: 2020 Class B, EN 55024: 2010, EN 55035: 2017
  • Japan VCCI-CISPR32:2016, VCCI 32-1: 2016 Class A
  • Korea KN32, Class A, KN35, RRA Public Notification 2019-32
  • Taiwan CNS 13438: 2016, C6357, Class A
  • USA FCC 47 CFR Part 15, Subpart B, Class A

Product Safety Compliance

  • UL 62368-1, 2nd Edition, 2014-12
  • CSA-C22.2 No. 62368-1, 2nd Edition, 2014-12
  • EN 62368-1, 2nd Edition, 2014 + A1: 2017
  • IEC 62368-1, 2nd Edition, 2014
  • RoHS Compliance: EU RoHS Directive (EU) 2015/863 Amendment to EU RoHS 2 (Directive 2011/65/EU)
  • REACH Compliance
  • Halogen Free: IEC 61249-2-21:2003 standard

Operating Environment

The MI210 Accelerator has the following operating characteristics:

Warranty

One year limited warranty. When installed in a Lenovo server, the MI210 Accelerator assumes the server's base warranty and any warranty upgrades.

Related Publications

For more information, refer to these documents:

Related Product Families

Product families related to this document are the following:

Notices

Lenovo may not offer the products, services, or features discussed in this document in all countries. Consult your local Lenovo representative for information on the products and services currently available in your area. Any reference to a Lenovo product, program, or service is not intended to state or imply that only that Lenovo product, program, or service may be used. Any functionally equivalent product, program, or service that does not infringe any Lenovo intellectual property right may be used instead. However, it is the user's responsibility to evaluate and verify the operation of any other product, program, or service. Lenovo may have patents or pending patent applications covering subject matter described in this document. The furnishing of this document does not give you any license to these patents. You can send license inquiries, in writing, to:

Lenovo (United States), Inc.
8001 Development Drive
Morrisville, NC 27560
U.S.A.
Attention: Lenovo Director of Licensing

LENOVO PROVIDES THIS PUBLICATION "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some jurisdictions do not allow disclaimer of express or implied warranties in certain transactions, therefore, this statement may not apply to you.

This information could include technical inaccuracies or typographical errors. Changes are periodically made to the information herein; these changes will be incorporated in new editions of the publication. Lenovo may make improvements and/or changes in the product(s) and/or the program(s) described in this publication at any time without notice.

The products described in this document are not intended for use in implantation or other life support applications where malfunction may result in injury or death to persons. The information contained in this document does not affect or change Lenovo product specifications or warranties. Nothing in this document shall operate as an express or implied license or indemnity under the intellectual property rights of Lenovo or third parties. All information contained in this document was obtained in specific environments and is presented as an illustration. The result obtained in other operating environments may vary. Lenovo may use or distribute any of the information you supply in any way it believes appropriate without incurring any obligation to you.

Any references in this publication to non-Lenovo Web sites are provided for convenience only and do not in any manner serve as an endorsement of those Web sites. The materials at those Web sites are not part of the materials for this Lenovo product, and use of those Web sites is at your own risk. Any performance data contained herein was determined in a controlled environment. Therefore, the result obtained in other operating environments may vary significantly. Some measurements may have been made on development-level systems and there is no guarantee that these measurements will be the same on generally available systems. Furthermore, some measurements may have been estimated through extrapolation. Actual results may vary. Users of this document should verify the applicable data for their specific environment.

© Copyright Lenovo 2023. All rights reserved.

This document, LP1862, was created or updated on December 6, 2023.

Send us your comments in one of the following ways:

This document is available online at https://lenovopress.lenovo.com/LP1862.

Trademarks

Lenovo and the Lenovo logo are trademarks or registered trademarks of Lenovo in the United States, other countries, or both. A current list of Lenovo trademarks is available on the Web at https://www.lenovo.com/us/en/legal/copytrade/.

The following terms are trademarks of Lenovo in the United States, other countries, or both: Lenovo®, ServerProven®, ThinkAgile®, ThinkSystem®

The following terms are trademarks of other companies:

Intel® is a trademark of Intel Corporation or its subsidiaries.

Linux® is the trademark of Linus Torvalds in the U.S. and other countries.

Other company, product, or service names may be trademarks or service marks of others.

Models: 4X67A81102, 4X67A82325, 4X67A81102 AMD Instinct MI210 Accelerator, 4X67A81102, AMD Instinct MI210 Accelerator, Instinct MI210 Accelerator, Accelerator

File Info : application/pdf, 11 Pages, 323.81KB

PDF preview unavailable. Download the PDF instead.

lp1862

References

wkhtmltopdf 0.12.4-dev-4fa8338 Qt 4.8.7

Related Documents

PreviewAMD Radeon Instinct MI50 Server Accelerator for Deep Learning and HPC
Discover the AMD Radeon Instinct MI50 server accelerator, a 7nm FinFET technology powerhouse designed for large-scale deep learning and HPC. Featuring up to 26.5 TFLOPS FP16 performance, 16GB/32GB HBM2 memory, and PCIe Gen 4 support, it offers exceptional compute capabilities for training complex neural networks and demanding HPC workloads.
PreviewAMD AOCL User Guide: Optimize Performance on Zen Processors
The AMD Optimizing CPU Libraries (AOCL) User Guide provides comprehensive instructions for installing, using, and tuning a suite of high-performance numerical libraries optimized for AMD 'Zen'-based processors. It covers AOCL-BLIS, AOCL-libFLAME, AOCL-FFTW, AOCL-LibM, AOCL-ScaLAPACK, AOCL-RNG, AOCL-SecureRNG, AOCL-Sparse, AOCL-LibMem, AOCL-Cryptography, and AOCL-Compression, along with integration with HPL and MUMPS, and performance tuning guidelines.
PreviewAMD Radeon™ Pro GPUs: Professional Graphics Card Benefits
Explore the advantages of AMD Radeon™ Pro GPUs for professional workstations, focusing on performance, reliability, multitasking, software certifications, and advanced features for demanding workflows.
PreviewAMD Ryzen 7 9800X3D Processor - Technical Specifications and Features
Detailed technical specifications for the AMD Ryzen 7 9800X3D processor, including core count, clock speeds, cache, socket type, manufacturing process, and key features like AMD 3D V-Cache technology.
PreviewLeveraging the IRON AI Engine API to Program AMD Ryzen AI NPUs
A technical guide for developers on programming AI Engines in AMD Ryzen processors using the IRON AI Engine API and MLIR-based compilers, covering spatial compute, data movement, performance analysis, and vectorization.
PreviewAMD Ryzen 9 7900 Processor Specifications and Support
Comprehensive technical specifications and support information for the AMD Ryzen 9 7900 desktop processor, detailing core count, clock speeds, memory support, connectivity, graphics capabilities, and product IDs.
PreviewAI PCs Revolutionizing Manufacturing: Enhancing Efficiency and Productivity
Explore how AI PCs, powered by AMD Ryzen AI PRO processors, are transforming the manufacturing sector by overcoming limitations of traditional PCs, enabling real-time analytics, predictive maintenance, and improved operational efficiency.
PreviewOptimizing AMD Ryzen™ Threadripper™ PRO Processors with Ansys HFSS for Enhanced Performance
This white paper explores the significant performance gains achieved by integrating AMD Optimizing CPU Libraries (AOCL) into Ansys HFSS simulation software on Lenovo ThinkStation P620 workstations powered by AMD Ryzen Threadripper PRO processors. It details testing methodologies and highlights the benefits for engineers and designers.