AMD Instinct™ Expands Ecosystem and Delivers Exascale-Class Technology for HPC and AI Applications

– Powered by AMD CDNA™ 2 archi­tec­tu­re and AMD ROCm™5, new AMD Instinct MI210 GPUs acce­le­ra­ting insights and dis­co­very for main­stream users –

SANTA CLARA, Calif., March 22, 2022 (GLOBE NEWSWIRE) — AMD (NASDAQ: AMD) today announ­ced the avai­la­bi­li­ty of the AMD Instinct™ eco­sys­tem with expan­ded sys­tem sup­port from part­ners inclu­ding ASUS, Dell Tech­no­lo­gies, Giga­byte, HPE, Leno­vo and Super­mi­cro, the new AMD Instinct™ MI210 acce­le­ra­tor and the robust capa­bi­li­ties of ROCm™ 5 soft­ware. Altog­e­ther, the AMD Instinct and ROCm eco­sys­tem is offe­ring exas­ca­le-class tech­no­lo­gy to a broad base of HPC and AI cus­to­mers, addres­sing the gro­wing demand for com­pu­te-acce­le­ra­ted data cen­ter workloads and redu­cing the time to insights and discovery.

With twice the plat­forms available com­pared to our pre­vious gene­ra­ti­on acce­le­ra­tors, gro­wing cus­to­mer adop­ti­on across HPC and AI appli­ca­ti­ons, and new sup­port from com­mer­cial ISVs in key workloads, we’re con­ti­nuing to dri­ve adop­ti­on of the AMD Instinct MI200 acce­le­ra­tors and ROCm 5 soft­ware eco­sys­tem,” said Brad McCre­die, cor­po­ra­te vice pre­si­dent, Data Cen­ter GPU and Acce­le­ra­ted Pro­ces­sing, AMD. “Now with the avai­la­bi­li­ty of the AMD Instinct MI210 acce­le­ra­tor to the MI200 fami­ly, our cus­to­mers can choo­se the acce­le­ra­tor that works best for their workloads, whe­ther they need lea­ding-edge acce­le­ra­ted pro­ces­sing for lar­ge sca­le HPC and AI workloads, or if they want access to exas­ca­le-class tech­no­lo­gy in a com­mer­cial format.”

The Lumi super­com­pu­ter powered by AMD EPYC pro­ces­sors and AMD Instinct MI200 acce­le­ra­tors will pro­vi­de a gene­ra­tio­nal leap in per­for­mance for lar­ge-sca­le simu­la­ti­ons and mode­ling as well as AI and deep lear­ning workloads to sol­ve some of the big­gest ques­ti­ons in rese­arch,” said Pek­ka Man­ni­nen, Direc­tor of the LUMI Lea­der­ship and Com­pu­ting Faci­li­ty, CSC. “We’ve uti­li­zed AMD Instinct MI210 acce­le­ra­tors to get hands on expe­ri­ence with the Instinct MI200 fami­ly, pre­pa­ring our sci­en­tists to tack­le the many chal­len­ging and com­plex pro­jects they will run once Lumi is ful­ly deployed.”

Powe­ring The Future of HPC and AI
The AMD Instinct MI200 series acce­le­ra­tors are desi­gned to power dis­co­veries in exas­ca­le sys­tems, enab­ling rese­ar­chers, sci­en­tists and engi­neers to tack­le our most pres­sing chal­lenges, from cli­ma­te chan­ge to vac­ci­ne rese­arch. The AMD Instinct MI210 acce­le­ra­tors spe­ci­fi­cal­ly enable exas­ca­le-class tech­no­lo­gies for cus­to­mers who need fan­ta­stic HPC and AI per­for­mance in a PCIe® for­mat. Powered by the AMD CDNA™ 2 archi­tec­tu­re, AMD Instinct MI210 acce­le­ra­tors extend AMD per­for­mance lea­der­ship in dou­ble pre­cis­i­on (FP64) com­pu­te on PCIe form fac­tor cards1. They also deli­ver a robust solu­ti­on for acce­le­ra­ted deep lear­ning trai­ning offe­ring a broad ran­ge of mixed-pre­cis­i­on capa­bi­li­ties based on the AMD Matrix Core Technology.

Dri­ving the ROCm Adoption
An open soft­ware plat­form that allows rese­ar­chers, sci­en­tists and engi­neers to tap the power of AMD Instinct acce­le­ra­tors to dri­ve sci­en­ti­fic dis­co­veries, the AMD ROCm plat­form is built on the foun­da­ti­on of num­e­rous appli­ca­ti­ons and libra­ri­es powe­ring top HPC and AI applications.

With ROCm 5, AMD extends its soft­ware plat­form by adding new hard­ware sup­port for the AMD Instinct MI200 series acce­le­ra­tors and the AMD Rade­on™ PRO W6800 pro­fes­sio­nal gra­phics card, plus Red Hat® Enter­pri­se Linux® 8.5 sup­port, incre­asing acces­si­bi­li­ty of ROCm for deve­lo­pers and enab­ling out­stan­ding per­for­mance across key workloads.

Addi­tio­nal­ly, through the AMD Infi­ni­ty Hub, the cen­tral loca­ti­on for open-source appli­ca­ti­ons that are por­ted and opti­mi­zed on AMD GPUs, end-users can easi­ly find, down­load and install con­tai­ne­ri­zed HPC apps and ML frame­works. AMD Infi­ni­ty Hub appli­ca­ti­on con­tai­ners are desi­gned to redu­ce the tra­di­tio­nal­ly dif­fi­cult issue of obtai­ning and instal­ling soft­ware releases while allo­wing users to learn based on shared expe­ri­en­ces and pro­blem-sol­ving opportunities.

Expan­ding Part­ner and Cus­to­mer Ecosystem
As more pur­po­se-built appli­ca­ti­ons are opti­mi­zed to work with ROCm and AMD Instinct acce­le­ra­tors, AMD con­ti­nues to grow its soft­ware eco­sys­tem with the addi­ti­on of com­mer­cial ISVs, inclu­ding Ansys®, Cas­ca­de Tech­no­lo­gies, and Tem­po­Quest. The­se ISVs pro­vi­de appli­ca­ti­ons for acce­le­ra­ted workloads inclu­ding Com­pu­ta­tio­nal Flu­id Dyna­mics (CFD), wea­ther fore­cas­ting, Com­pu­ter Aided Engi­nee­ring (CAE) and more. The­se updates are on top of exis­ting appli­ca­ti­on sup­port in ROCm which includes HPC, AI and Machi­ne Lear­ning appli­ca­ti­ons, AMBER, Chro­ma, CP2K, GRID, GRO­MACs, LAAMPS, MILC, Mini-HAAC, NAMD, NAMD 3.0, ONNX-RT, OpenMM, PyTorch, RELION, SPECFEM3D Car­te­si­an, SPECFEM3D Glo­be, and TensorFlow.

AMD is also enab­ling part­ners like ASUS, Dell Tech­no­lo­gies, Giga­byte, HPE, Leno­vo, Super­mi­cro, and Sys­tem Inte­gra­tors inclu­ding Col­fax, Exxact, KOI Com­pu­ters, Nor-Tech, Pen­gu­in and Sym­me­tric to offer dif­fe­ren­tia­ted solu­ti­ons to address next gene­ra­ti­on com­pu­ting chal­lenges. Super­com­pu­ting cus­to­mers are alre­a­dy taking advan­ta­ge of the bene­fits offe­red via the­se new cus­to­mer wins inclu­ding the Fron­tier instal­la­ti­on at Oak Ridge Natio­nal Labo­ra­to­ryKTH/DardelCSC/LUMI and Cines/Adastra.

Enab­ling Access for Cus­to­mers and Partners
The AMD Acce­le­ra­tor Cloud offers cus­to­mers an envi­ron­ment to remo­te­ly access and eva­lua­te AMD Instinct acce­le­ra­tors and AMD ROCm soft­ware. Whe­ther it’s port­ing lega­cy code, bench­mar­king an appli­ca­ti­on or test­ing mul­ti-GPU or mul­ti-node sca­ling, the AMD Acce­le­ra­tor Cloud gives pro­s­pec­ti­ve cus­to­mers and part­ners quick and easy access to the latest GPUs and soft­ware. The AMD Acce­le­ra­tor Cloud is also used to power various events such as hacka­thons and ROCm trai­ning ses­si­ons offe­red to both exis­ting and pro­s­pec­ti­ve cus­to­mers, allo­wing deve­lo­pers to hone their skills and learn how to get the most out of AMD Instinct accelerators.

MI200 Series Specifications

Model Com­pu­te Units Stream Pro­ces­sors FP64 | FP32 Vec­tor (Peak) FP64 | FP32 Matrix (Peak) FP16 | bf16
(Peak)
INT8
(Peak)
HBM2e
ECC
Memory
Memo­ry Bandwidth Form Fac­tor
AMD Instinct MI210 104 6,656 Up to 22.6 TF Up to 45.3 TF Up to 181.0 TF Up to 181.0 TOPS 64GB Up to 1.6 TB/sec PCIe®
AMD Instinct MI250 208 13,312 Up to 45.3 TF Up to 90.5 TF Up to 362.1 TF Up to 362.1 TOPS 128GB 3.2 TB/sec OCP Acce­le­ra­tor Modu­le (OAM)
AMD Instinct MI250x 220 14,080 Up to 47.9 TF Up to 95.7 TF Up To 383.0 TF Up to 383.0 TOPS 128GB 3.2 TB/sec OCP Acce­le­ra­tor Modu­le (OAM)

Sup­port­ing Resources

About AMD

For more than 50 years AMD has dri­ven inno­va­ti­on in high-per­for­mance com­pu­ting, gra­phics and visua­liza­ti­on tech­no­lo­gies. Bil­li­ons of peo­p­le, lea­ding For­tu­ne 500 busi­nesses and cut­ting-edge sci­en­ti­fic rese­arch insti­tu­ti­ons around the world rely on AMD tech­no­lo­gy dai­ly to impro­ve how they live, work and play. AMD employees are focu­sed on buil­ding lea­der­ship high-per­for­mance and adap­ti­ve pro­ducts that push the boun­da­ries of what is pos­si­ble. For more infor­ma­ti­on about how AMD is enab­ling today and inspi­ring tomor­row, visit the AMD (NASDAQ: AMDweb­siteblogLin­ke­dIn and Twit­ter pages.

©2022 Advan­ced Micro Devices, Inc. All rights reser­ved. AMD, the AMD Arrow logo, AMD CDNA, AMD Instinct, Rade­on, ROCm and com­bi­na­ti­ons the­reof are trade­marks of Advan­ced Micro Devices, Inc. PCIe is a regis­tered trade­mark of PCI-SIG Cor­po­ra­ti­on. Red Hat, Red Hat Enter­pri­se Linux, and the Red Hat logo, are trade­marks or regis­tered trade­marks of Red Hat, Inc. or its sub­si­dia­ries in the U.S. and other count­ries. Other pro­duct names used in this publi­ca­ti­on are for iden­ti­fi­ca­ti­on pur­po­ses only and may be trade­marks of their respec­ti­ve companies.

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

___________________________
1 MI200-41 — Cal­cu­la­ti­ons con­duc­ted by AMD Per­for­mance Labs as of Jan 14, 2022, for the AMD Instinct™ MI210 (64GB HBM2e PCIe® card) acce­le­ra­tor at 1,700 MHz peak boost engi­ne clock resul­ted in 45.3 TFLOPS peak theo­re­ti­cal dou­ble pre­cis­i­on (FP64 Matrix), 22.6 TFLOPS peak theo­re­ti­cal dou­ble pre­cis­i­on (FP64), and 181.0 TFLOPS peak theo­re­ti­cal Bfloat16 for­mat pre­cis­i­on (BF16), floa­ting-point performance.

Cal­cu­la­ti­ons con­duc­ted by AMD Per­for­mance Labs as of Sep 18, 2020 for the AMD Instinct™ MI100 (32GB HBM2 PCIe® card) acce­le­ra­tor at 1,502 MHz peak boost engi­ne clock resul­ted in 11.54 TFLOPS peak theo­re­ti­cal dou­ble pre­cis­i­on (FP64), and 184.6 TFLOPS peak theo­re­ti­cal half pre­cis­i­on (FP16), floa­ting-point performance.

Published results on the NVi­dia Ampere A100 (80GB) GPU acce­le­ra­tor, boost engi­ne clock of 1410 MHz, resul­ted in 19.5 TFLOPS peak dou­ble pre­cis­i­on ten­sor cores (FP64 Ten­sor Core), 9.7 TFLOPS peak dou­ble pre­cis­i­on (FP64) and 39 TFLOPS peak Bfloat16 for­mat pre­cis­i­on (BF16), theo­re­ti­cal floa­ting-point per­for­mance. The TF32 data for­mat is not IEEE com­pli­ant and not included in this comparison.
https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/nvidia-ampere-architecture-whitepaper.pdf, page 15, Table 1.