SlideShare a Scribd company logo

Amd epyc update_gdep_xilinx_ai_web_seminar_20201028

S
ssuser5b12d1

AMD EPYC processors, launched in October 2020, provide advanced features such as 8 channels of DDR4-3200 memory, 4TB memory capacity, and support for PCIe Gen 4. The processors offer strong performance with high core counts, maximizing memory bandwidth and I/O capabilities compared to competing Intel Xeon processors. Additionally, they feature secure memory encryption and virtualization technologies.

Technology
1 of 31
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Amd epyc update_gdep_xilinx_ai_web_seminar_20201028
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Roadmap Subject to Change
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2015 2016 2017 2018 2019 Competitor Foundry 14nm 14+Fin 10Fin 12nm 7nm 14Fin Performance/Watt 3
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] m14nm I/O DIE I/O DD R DD R DD R DD R ∞ DD R DD R DD R DD R 7nm CPU die 7nm CPU die ∞ ∞ ∞ I/O ∞ ∞ ∞ ∞ 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die Monolithic die Multi-die MCM Chiplet
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ✓ All-in feature set includes: 8 Channels of DDR4-3200* 4TB memory capacity 128 lanes PCIe® 4* SMT & Turbo boost 18G Infinity Fabric™ Secure Memory Encryption Secure Encrypted Virtualization * A MOTHERBOARDDESIGNEDFOR 2ND GEN EPYC PROCESSORS IS REQUIRED TO ENABLE ALL FUNCTIONALITY. SEE ENDNOTE ROM-06 AMD EPYC™ 7F32 8 CORE AMD EPYC™ 7F52 12/16 CORE AMD EPYC™ 7F72 24 CORE 32 CORE 48/64 CORE 先進的な機能を備た、シンプルな製品ラインナップ
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] See Endnote ROM-741. Best published SPECrate®2017_int_base performance as of 9/29/2020. Intel Xeon™ Intel Xeon™ 48C
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Intel Xeon™ See Endnote ROM-742. Both published and estimated SPECrate®2017_int_base performance as of March 2, 2020. 48C
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] a. プロセッサが動作可能な最大シングルコア周波数 b. DDR4-2667 DIMM 4チャネルでの動作に最適化 モデル番号 コア数 スレッド数 ベース周波数 (GHz) 最大ブースト 周波数(GHz) a TDP (W) L3 キャッシュ (MB) DDR チャネル数 最大DDR 周波数 (1DPC) ソケット当たり理論 メモリバンド幅 (GB/s) PCIe Gen 4 レーン数 7H12 64 128 2.6 3.3 280 256 8 3200 204.8 128 7742 64 128 2.25 3.4 225 256 8 3200 204.8 128 7702 7702P 64 128 2.0 3.35 200 256 8 3200 204.8 128 7662 64 128 2.0 3.3 225 256 8 3200 204.8 128 7642 48 96 2.3 3.3 225 256 8 3200 204.8 128 7552 48 96 2.2 3.3 200 192 8 3200 204.8 128 7542 32 64 2.9 3.4 225 128 8 3200 204.8 128 7532 32 64 2.4 3.3 200 256 8 3200 204.8 128 7502 7502P 32 64 2.5 3.35 180 128 8 3200 204.8 128 7452 32 64 2.35 3.35 155 128 8 3200 204.8 128 7F72 24 48 3.2 3.7 240 192 8 3200 204.8 128 7402 7402P 24 48 2.8 3.35 180 128 8 3200 204.8 128 7352 24 48 2.3 3.2 155 128 8 3200 204.8 128 7F52 16 32 3.5 3.9 240 256 8 3200 204.8 128 7302 7302P 16 32 3.0 3.3 155 128 8 3200 204.8 128 7282 16 32 2.8 3.2 120 64 8 3200 85.3b 128 7272 12 24 2.9 3.2 120 64 8 3200 85.3b 128 7F32 8 16 3.7 3.9 180 128 8 3200 204.8 128 7262 8 16 3.2 3.4 155 128 8 3200 204.8 128 7252 8 16 3.1 3.2 120 64 8 3200 85.3b 128 7232P 8 16 3.1 3.2 120 32 8 3200 85.3b 128 New New New
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7742 64 2.25 3.40 256 3200 225 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7542 32 2.90 3.40 128 3200 225 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7532 32 2.40 3.30 256 3200 200 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] P R O D U C T S See endnotes EPYC-18, ROM-169, ROM-557. *EPYC 7702 - see endnote ROM-517. **A motherboard designed for 2nd Gen EPYC processors is required to enable all available functionality
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7F52 16 3.50 3.90 256 3200 240 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7F32 8 3.70 3.90 128 3200 180 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] I/O ▪ 業界初の PCIe® Gen 4 対応 CPU ▪ PCIe Gen 3に対して2倍の双方向データレート ▪ NVMe、GPUやその他のアクセラレータを直接接続 可能な128レーンのPCIe® Gen4インターフェース メモリー ▪ 競合製品に対して33%メモリーチャネル数が多い1 ▪ CPU当たり最大4TBのDDR4メモリー容量 ▪ 最高 3200MHz DDR4 Memoryまでのサポート 4倍 競合製品に対する I/Oバンド幅2 45% 競合に対するメモリ バンド幅の向上3 Note: 1. Each AMD EPYC processor has 8 memory channels. Each Intel Xeon Scalable processor has 6 memory channels. 8 – 6 = 2 ÷ 6 = 0.33 AMD EPYC has 33% more memory channels. Class based on industry-standard pin-based (LGA) X86 processors 2. One EPYC 7002 Series has 433% [or: 4 times] more IO bandwidth than one Intel Scalable processors ROM-21 3. AMD EPYC™ 7002 Series processors have 45% more memory bandwidth than Intel Scalable processors in the same class. ROM-11
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Intel® Xeon® 2nd Gen Scalable (“Cascade Lake SP”) AMD EPYC™ 7001 Series AMD EPYC™ 7002 Series Max Theoretical Memory BW per 2P server (1 DR DPC) 12 x DDR4-2933 = 282GB/s 16 x DDR4-2666 = 340GB/s 16 x DDR4-3200* = 410GB/s AMD EPYC™ Advantage +21% +45% DIMMDIMM A B C D E F G H 広帯域メモリーがパフォーマンスのスケーリングに貢献します *AMD POR memory speeds in a one DIMM per channel system implementation AMD EPYC™ 7002 シリーズ DDR4 1DIMM per Channel Server* DR RDIMM: 3200 MHz LRDIMM: 3200 MHz Not supported on all motherboards – see endnotes ROM-06 for details.
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] EPYCがサポートする 広帯域 PCIE GEN4 はパフォーマンスの スケーリングに貢献します  8 x16 レーン 全てにおいて、 PCIe® Gen4 対応 − 1レーン当たり 64GB/s 双方向バンド幅, ソケット当たり512GB/s  8 x16 レーン全てにおいて、IOMMU サポート  レーン分割のサポート、 x16あたり最大8個の PCIeデバイス  フル PCIe P2P サポート、ソケット内, ソケット間ともに − 最大P2P ペイロードサイズ: 256B − 最大 DMA ペイロードサイズ: 512B  I/O ダイ上の AMD Infinity Fabric ™ は DMA and P2P に対して最適化さ れています  1 ソケットプラットフォームにおいては、全てのレーンをI/Oとして利 用可能  2 ソケットプラットフォームにおいては、最大162レーンをI/Oとして 利用可能 Not supported on all motherboards – see endnotes ROM-06 x16 x16 x8 x8 x4 x4x4 x4 x2 x2x2 x2x2 x2x2 x2 x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1 x16 x8 x8 x4 x4x4 x4 x2 x2x2 x2x2 x2x2 x2 x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1 x1 x1x1 x1x1 x1x1 x1
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] See EndnotesROM-173, ROM-174, ROM-175, ROM-176, ROM-177 and ROM-178
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] AMD EPYC™ Tech Docs and White Papers 検索サイト： https://www.amd.com/en/processors/server-tech-docs/search。
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Source: https://www.amd.com/system/files/documents/AMD-EPYC-7002-AnsysFluent-Scalability-Performance.pdf
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Refer from : https://www.amd.com/system/files/documents/amd-epyc-7Fx2-ansys-fluent.pdf
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Rome とAlveo は共にPCIe Gen4で通信し、HPC, ML, VM, Database, Networking NFV and 5G Domains のUse case においてKeyとなる高スループットを期待した測定。 XDMA-type data (large block size): Host Kernel, FPGA local Memory 間のデータ転送。 QDMA-type data(smaller block packet sizes): DPDK利用で、Host user space, FPGA local Memory間のPacket転送。 共に広帯域を確認。 参照先： https://www.amd.com/system/files/documents/epyc-7fx2-processors-network-performance-using-dpdk.pdf ※日本AMDで独自にまとめたスライドです
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] NVIDIA
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] AMD Developer Central (No NDAの技術サイト) https://developer.amd.com/ AMD EPYC™ Tech Docs and White Papers (各種 ホワイトペーパー、技術資料のページ) https://www.amd.com/en/processors/server-tech-docs/search GPUOpen (AMDのGPUに関するページ) https://gpuopen.com/ AMD EPYC™ & HPE (AMDによるHPE製品に関するページ) https://www.amd.com/ja/campaigns/amd-and-hpe
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ▪ EPYC-07 - Based on June 8, 2018 AMD internal testing of same-architectureproduct ported from 14 to 7 nm technology with similar implementationflow/methodology, using performancefrom SGEMM. ▪ ROM-04 Based on standard calculation method for determiningFLOPS. ROM-04 ▪ ROM-06 - Some supported features and functionalityof 2nd Gen AMD EPYC™ processors require a BIOS update from your server manufacturer when used with a motherboard designed for the 1st Gen AMD EPYC series processor. A motherboard designed for 2nd Gen EPYC processors is required to enable all available functionality. ▪ ROM-07 - Motherboards designed for 1st Gen EPYC processors may not be compatiblewith 2nd Gen AMD EPYC processors with a TDP greater than 200 watts. Contact the server manufacturer to confirm compatibility. ▪ ROM-342 1-n, 2-socket 2nd Gen EPYC™ 7742 powered server (http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20001.pdf) scoring 701 SPECrate®2017_int_base has up to 2.3x the SPECrate® 2017 Integer (Base) performanceof the highest previous generation score of 304 SPECrate®2017_int_base by a 1-n, 2-socket 1st Gen EPYC™ 7601 powered server result (http://www.spec.org/cpu2017/results/res2019q2/cpu2017-20190411-11817.html ) as of 1/26/20. ROM-342 ▪ ROM-741: Best published SPECrate®2017_int_base performanceas of 9/29/2020. AMD EPYC 2P processor scores reflect highest scoring processor at each availablecore level (core count is per processor): AMD EPYC 7742 (64C) score of 701 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20001.pdf AMD EPYC 7642 (48C) score of 586 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200622-23009.html AMD EPYC 7542 (32C) score of 437 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200706-23309.html AMD EPYC 7302 (16C) score of 246 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191220-20441.pdf AMD EPYC 7262 (8C) score of 135 http://spec.org/cpu2017/results/res2020q2/cpu2017-20200608-22777.html Intel® Xeon® Scalable 2P processors scores using the highest competitivepart in the Bronze, Silver and Gold metal classes: Intel Xeon Gold 6258R (28C) score of 397 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200915-23981.html Intel Xeon Gold 5220R (24C) score of 319 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200608-22594.html Intel Xeon Silver 4216 (16C) score of 193 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200804-23653.html Intel Xeon Bronze 3206R (8C) score of 60.1 http://spec.org/cpu2017/results/res2020q2/cpu2017-20200609-22881.html SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard Performance EvaluationCorporation. See w ww.spec.orgfor more informationROM-741
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ▪ ROM-742: Both published and estimated SPECrate®2017_int_base performanceas of March 2, 2020. AMD EPYC 7742 (64C) score of 355 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191223-20460.pdf AMD EPYC 7642 (48C) score of 306 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200827-23775.html AMD EPYC 7542 (32C) score of 223 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191028-19322.pdf AMD EPYC 7402P (24C) score of 177 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191028-19326.pdf AMD EPYC 7302P (16C) score of 124 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20005.pdf AMD EPYC 7262 (8C) score of 68.3 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191220-20437.pdf Intel® Xeon® Scalable 2P processors scores using the highest competitivepart in the Bronze, Silver and Gold metal classes. Estimates are based on highest score 2P processor result and divided by 1.99x for conservative estimate: Intel Xeon Gold 6258R (28C) estimated score of 199 (derived from 2P score of 397/1.99=199) https://spec.org/cpu2017/results/res2020q3/cpu2017-20200915-23981.html Intel Xeon Gold 5220R (24C) estimated score of 160 (derived from 2P score of 319/1.99=160) http://spec.org/cpu2017/results/res2020q3/cpu2017-20200608-22594.html Intel Xeon Silver 4216 (16C) score of 87.3 http://spec.org/cpu2017/results/res2019q2/cpu2017-20190524-14372.pdf Intel Xeon Bronze 3206R (8C) estimated score of 30.2 (derived from 2P score of 60.1/1.99= 30.2) http://spec.org/cpu2017/results/res2020q2/cpu2017-20200609-22881.html Results may vary with production system publications. The average of all published Intel results that have a 1P and 2P corresponding number is 2.08x with the maximum being 2.3x and the minimumbeing 1.95x so conservatively AMD estimated using 1.99x scaling. AMD expects Intel published numbersin the near future will likely be higher. SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard PerformanceEvaluation Corporation. See www.spec.orgfor more information. ROM-742 ▪ ROM-42 - Based on AMD internal testing of ANSYS FLUENT 19.1, lm6000_16m benchmark, as of July 17, 2019 of a 2P EPYC 7742 powered reference server versus a 2P Intel Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-49 - Based on AMD internal testing of LSTC LS-DYNA R9.3.0, neon benchmark, as of July 17, 2019 of a 2P EPYC 7742 powered referenceserver versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-56 - Based on AMD internal testing of Altair RADIOSS 2018, T10M benchmark, as of July 17, 2019 using a 2P EPYC 7742 powered reference server versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-63 - Based on AMD internal testing of ESI VPS 2018.0, NEON4m benchmark, as of July 17, 2019 using a 2P EPYC 7742 powered reference server versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-113 AMD Internal testing as of 30July2019 of a 2P AMD EPYC 7742 powered reference platform versus a 2P Intel Platinum 8280 powered production server, on GROMACS version 2019.3 benchmark. Results may vary. ROM-113
| AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] DISCLAIMER The information contained herein is for informational purposes only, and is subject to change without notice. While every precaution has been taken in the preparation of this document, it may contain technical inaccuracies, omissions and typographical errors, and AMD is under no obligation to update or otherwise correct this information. Advanced Micro Devices, Inc. makes no representations or warranties with respect to the accuracy or completeness of the contents of this document, and assumes no liability of any kind, including the implied warranties of noninfringement, merchantability or fitness for particular purposes, with respect to the operation or use of AMD hardware, software or other products described herein. No license, including implied or arising by estoppel, to any intellectual property rights is granted by this document. Terms and limitations applicable to the purchase or use of AMD’s products are as set forth in a signed agreement between the parties or in AMD's Standard Terms and Conditions of Sale. GD-18 ©2019 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, [insert all other AMD trademarks used in the material here per AMD’s Checklist for Trademark Attribution] and combinations thereof are trademarks of Advanced Micro Devices, Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.

More Related Content

PDF
Product Roadmap iEi 2017
Andrei Teleanu
 
PDF
한컴MDS_NVIDIA Jetson Platform
HANCOM MDS
 
PDF
AMD Ryzen Threadripper in Techday
Low Hong Chuan
 
PDF
6th gen Intel Core and Intel Xeon
Overclockers.ua
 
PDF
Qvl
Juan Carlos Silva Diaz
 
PDF
numPYNQ: accelerating NumPy on PYNQ
NECST Lab @ Politecnico di Milano
 
PDF
Sun fire x2250 technical training presentation
xKinAnx
 
PDF
Radeon RX Vega
Low Hong Chuan
 
Product Roadmap iEi 2017
Andrei Teleanu
 
한컴MDS_NVIDIA Jetson Platform
HANCOM MDS
 
AMD Ryzen Threadripper in Techday
Low Hong Chuan
 
6th gen Intel Core and Intel Xeon
Overclockers.ua
 
Qvl
Juan Carlos Silva Diaz
 
numPYNQ: accelerating NumPy on PYNQ
NECST Lab @ Politecnico di Milano
 
Sun fire x2250 technical training presentation
xKinAnx
 
Radeon RX Vega
Low Hong Chuan
 

What's hot (15)

PDF
CORSAIR VENGEANCE A4100 GAMING PC REVIEW
Dharmendra Rama
 
ODS
Comparativa cpu
tuxman82
 
PDF
Hardware Discovery Commands
Kevin OBrien
 
PDF
Inside ASUS ZENBOOK
JJ Wu
 
PDF
PGCon 2014 - What Do You Mean my Database Server Core Dumped? - How to Inspec...
Faisal Akber
 
PPT
Cheap 3d pc project
Jorge Viloria
 
PDF
Computex 2017 - AMD Press Conference
Low Hong Chuan
 
PDF
Android Development Tools
Dominik Helleberg
 
PDF
Yeni Nesil Sunucular ile Veritabanınız
MSHOWTO Bilisim Toplulugu
 
PDF
15 mei 2013
Syamsul Arifin
 
PDF
Dell Technologies Dell EMC POWERMAX Storage On One Single Page - POSTER - v1a...
Smarter.World
 
PDF
Nsrit37
sai sekhar
 
PDF
Dell Technologies Dell EMC Data Protection Solutions On One Single Page - POS...
Smarter.World
 
PDF
Android Development Tools
Dominik Helleberg
 
PDF
Vigor Ex
rwachsman
 
CORSAIR VENGEANCE A4100 GAMING PC REVIEW
Dharmendra Rama
 
Comparativa cpu
tuxman82
 
Hardware Discovery Commands
Kevin OBrien
 
Inside ASUS ZENBOOK
JJ Wu
 
PGCon 2014 - What Do You Mean my Database Server Core Dumped? - How to Inspec...
Faisal Akber
 
Cheap 3d pc project
Jorge Viloria
 
Computex 2017 - AMD Press Conference
Low Hong Chuan
 
Android Development Tools
Dominik Helleberg
 
Yeni Nesil Sunucular ile Veritabanınız
MSHOWTO Bilisim Toplulugu
 
15 mei 2013
Syamsul Arifin
 
Dell Technologies Dell EMC POWERMAX Storage On One Single Page - POSTER - v1a...
Smarter.World
 
Nsrit37
sai sekhar
 
Dell Technologies Dell EMC Data Protection Solutions On One Single Page - POS...
Smarter.World
 
Android Development Tools
Dominik Helleberg
 
Vigor Ex
rwachsman
 
Ad

Similar to Amd epyc update_gdep_xilinx_ai_web_seminar_20201028 (20)

PDF
AMD EPYC™ Microprocessor Architecture
AMD
 
PPTX
Mirabilis Design- NoC Webinar- 15th-Oct 2024
Deepak Shankar
 
PDF
Deep learning: Hardware Landscape
Grigory Sapunov
 
PDF
PCCC22：日本AMD株式会社 テーマ1「第4世代AMD EPYC™ プロセッサー (Genoa) の概要」
PC Cluster Consortium
 
PPTX
AMD EPYC 7002 Launch World Records
AMD
 
PDF
Mark Papermaster Next Horizon Presentation
Low Hong Chuan
 
PDF
AMD Embedded Solutions Guide
AMD
 
PPTX
Azure Local AMD Solutions Offering 20250131.pptx
Urielrodzmig
 
PDF
NextHorizonOnTheCaseWirgThePresnrentions
BigDog29
 
PPT
O que há de novo na plataforma x86 para High Performance por Jefferson de A S...
Joao Galdino Mello de Souza
 
PPTX
AMD Chiplet Architecture for High-Performance Server and Desktop Products
AMD
 
PPTX
AMD EPYC 7002 World Records
AMD
 
PPTX
AI Hardware Landscape 2021
Grigory Sapunov
 
PDF
Deep Learning Training at Scale: Spring Crest Deep Learning Accelerator
inside-BigData.com
 
PDF
PCCC23：日本AMD株式会社 テーマ２「AMD EPYC™ プロセッサーを用いたAIソリューション」
PC Cluster Consortium
 
PDF
“Programming Vision Pipelines on AMD’s AI Engines,” a Presentation from AMD
Edge AI and Vision Alliance
 
PPTX
Lesson 2.pptxwertfyguhijkokjhgfdsasdfghjk
shilumbua00
 
PDF
L09.pdf
TRNHONGLINHBCHCM
 
PDF
A CGRA-based Approach for Accelerating Convolutional Neural Networks
Shinya Takamaeda-Y
 
PDF
AIDC India - AI on IA
Intel® Software
 
AMD EPYC™ Microprocessor Architecture
AMD
 
Mirabilis Design- NoC Webinar- 15th-Oct 2024
Deepak Shankar
 
Deep learning: Hardware Landscape
Grigory Sapunov
 
PCCC22：日本AMD株式会社 テーマ1「第4世代AMD EPYC™ プロセッサー (Genoa) の概要」
PC Cluster Consortium
 
AMD EPYC 7002 Launch World Records
AMD
 
Mark Papermaster Next Horizon Presentation
Low Hong Chuan
 
AMD Embedded Solutions Guide
AMD
 
Azure Local AMD Solutions Offering 20250131.pptx
Urielrodzmig
 
NextHorizonOnTheCaseWirgThePresnrentions
BigDog29
 
O que há de novo na plataforma x86 para High Performance por Jefferson de A S...
Joao Galdino Mello de Souza
 
AMD Chiplet Architecture for High-Performance Server and Desktop Products
AMD
 
AMD EPYC 7002 World Records
AMD
 
AI Hardware Landscape 2021
Grigory Sapunov
 
Deep Learning Training at Scale: Spring Crest Deep Learning Accelerator
inside-BigData.com
 
PCCC23：日本AMD株式会社 テーマ２「AMD EPYC™ プロセッサーを用いたAIソリューション」
PC Cluster Consortium
 
“Programming Vision Pipelines on AMD’s AI Engines,” a Presentation from AMD
Edge AI and Vision Alliance
 
Lesson 2.pptxwertfyguhijkokjhgfdsasdfghjk
shilumbua00
 
L09.pdf
TRNHONGLINHBCHCM
 
A CGRA-based Approach for Accelerating Convolutional Neural Networks
Shinya Takamaeda-Y
 
AIDC India - AI on IA
Intel® Software
 
Ad

Recently uploaded (20)

PDF
Building Integrated photovoltaic BIPV_UPV.pdf
SandeepSingh286037
 
PDF
The Rise and Fall of 3GPP – Time for a Sabbatical?
3G4G
 
PDF
Advanced methodologies resolving dimensionality complications for autism neur...
IAESIJAI
 
PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
SOFTTECHHUB
 
PDF
Review of recent advances in non-invasive hemoglobin estimation
IAESIJAI
 
PDF
Assigned Numbers - 2025 - Bluetooth® Document
Bloombase
 
PDF
Machine learning based COVID-19 study performance prediction
IAESIJAI
 
PDF
Per capita expenditure prediction using model stacking based on satellite ima...
IAESIJAI
 
PDF
Approach and Philosophy of On baking technology
30anthuong17
 
PDF
Reach Out and Touch Someone: Haptics and Empathic Computing
Mark Billinghurst
 
PDF
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
Jian-Hong Pan
 
PPTX
20250228 LYD VKU AI Blended-Learning.pptx
DatVoNgoc
 
PPTX
Big Data Technologies - Introduction.pptx
kuthubussaman1
 
PDF
Network Security Unit 5.pdf for BCA BBA.
Serpent6
 
PDF
Mobile App Security Testing_ A Comprehensive Guide.pdf
flufftailshop
 
PDF
MIND Revenue Release Quarter 2 2025 Press Release
MIND CTI
 
PDF
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
SOFTTECHHUB
 
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
Dropbox
 
PDF
Empathic Computing: Creating Shared Understanding
Mark Billinghurst
 
PPTX
A Presentation on Artificial Intelligence
memembruh336
 
Building Integrated photovoltaic BIPV_UPV.pdf
SandeepSingh286037
 
The Rise and Fall of 3GPP – Time for a Sabbatical?
3G4G
 
Advanced methodologies resolving dimensionality complications for autism neur...
IAESIJAI
 
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
SOFTTECHHUB
 
Review of recent advances in non-invasive hemoglobin estimation
IAESIJAI
 
Assigned Numbers - 2025 - Bluetooth® Document
Bloombase
 
Machine learning based COVID-19 study performance prediction
IAESIJAI
 
Per capita expenditure prediction using model stacking based on satellite ima...
IAESIJAI
 
Approach and Philosophy of On baking technology
30anthuong17
 
Reach Out and Touch Someone: Haptics and Empathic Computing
Mark Billinghurst
 
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
Jian-Hong Pan
 
20250228 LYD VKU AI Blended-Learning.pptx
DatVoNgoc
 
Big Data Technologies - Introduction.pptx
kuthubussaman1
 
Network Security Unit 5.pdf for BCA BBA.
Serpent6
 
Mobile App Security Testing_ A Comprehensive Guide.pdf
flufftailshop
 
MIND Revenue Release Quarter 2 2025 Press Release
MIND CTI
 
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
SOFTTECHHUB
 
Dropbox Q2 2025 Financial Results & Investor Presentation
Dropbox
 
Empathic Computing: Creating Shared Understanding
Mark Billinghurst
 
A Presentation on Artificial Intelligence
memembruh336
 

Amd epyc update_gdep_xilinx_ai_web_seminar_20201028

  • 2. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Roadmap Subject to Change
  • 3. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2015 2016 2017 2018 2019 Competitor Foundry 14nm 14+Fin 10Fin 12nm 7nm 14Fin Performance/Watt 3
  • 4. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] m14nm I/O DIE I/O DD R DD R DD R DD R ∞ DD R DD R DD R DD R 7nm CPU die 7nm CPU die ∞ ∞ ∞ I/O ∞ ∞ ∞ ∞ 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die 7nm CPU die Monolithic die Multi-die MCM Chiplet
  • 5. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
  • 6. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ✓ All-in feature set includes: 8 Channels of DDR4-3200* 4TB memory capacity 128 lanes PCIe® 4* SMT & Turbo boost 18G Infinity Fabric™ Secure Memory Encryption Secure Encrypted Virtualization * A MOTHERBOARDDESIGNEDFOR 2ND GEN EPYC PROCESSORS IS REQUIRED TO ENABLE ALL FUNCTIONALITY. SEE ENDNOTE ROM-06 AMD EPYC™ 7F32 8 CORE AMD EPYC™ 7F52 12/16 CORE AMD EPYC™ 7F72 24 CORE 32 CORE 48/64 CORE 先進的な機能を備た、シンプルな製品ラインナップ
  • 7. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] See Endnote ROM-741. Best published SPECrate®2017_int_base performance as of 9/29/2020. Intel Xeon™ Intel Xeon™ 48C
  • 8. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Intel Xeon™ See Endnote ROM-742. Both published and estimated SPECrate®2017_int_base performance as of March 2, 2020. 48C
  • 9. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] a. プロセッサが動作可能な最大シングルコア周波数 b. DDR4-2667 DIMM 4チャネルでの動作に最適化 モデル番号 コア数 スレッド数 ベース周波数 (GHz) 最大ブースト 周波数(GHz) a TDP (W) L3 キャッシュ (MB) DDR チャネル数 最大DDR 周波数 (1DPC) ソケット当たり理論 メモリバンド幅 (GB/s) PCIe Gen 4 レーン数 7H12 64 128 2.6 3.3 280 256 8 3200 204.8 128 7742 64 128 2.25 3.4 225 256 8 3200 204.8 128 7702 7702P 64 128 2.0 3.35 200 256 8 3200 204.8 128 7662 64 128 2.0 3.3 225 256 8 3200 204.8 128 7642 48 96 2.3 3.3 225 256 8 3200 204.8 128 7552 48 96 2.2 3.3 200 192 8 3200 204.8 128 7542 32 64 2.9 3.4 225 128 8 3200 204.8 128 7532 32 64 2.4 3.3 200 256 8 3200 204.8 128 7502 7502P 32 64 2.5 3.35 180 128 8 3200 204.8 128 7452 32 64 2.35 3.35 155 128 8 3200 204.8 128 7F72 24 48 3.2 3.7 240 192 8 3200 204.8 128 7402 7402P 24 48 2.8 3.35 180 128 8 3200 204.8 128 7352 24 48 2.3 3.2 155 128 8 3200 204.8 128 7F52 16 32 3.5 3.9 240 256 8 3200 204.8 128 7302 7302P 16 32 3.0 3.3 155 128 8 3200 204.8 128 7282 16 32 2.8 3.2 120 64 8 3200 85.3b 128 7272 12 24 2.9 3.2 120 64 8 3200 85.3b 128 7F32 8 16 3.7 3.9 180 128 8 3200 204.8 128 7262 8 16 3.2 3.4 155 128 8 3200 204.8 128 7252 8 16 3.1 3.2 120 64 8 3200 85.3b 128 7232P 8 16 3.1 3.2 120 32 8 3200 85.3b 128 New New New
  • 10. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7742 64 2.25 3.40 256 3200 225 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 ※日本AMDで独自にまとめたスライドです
  • 11. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7542 32 2.90 3.40 128 3200 225 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 16MB L3 Z2 L2 Z2L2 Z2 L2 Z2L2 ※日本AMDで独自にまとめたスライドです
  • 12. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7532 32 2.40 3.30 256 3200 200 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 16MB L3 Z2L2 Z2L2 16MB L3 Z2 L2 Z2 L2 ※日本AMDで独自にまとめたスライドです
  • 13. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] P R O D U C T S See endnotes EPYC-18, ROM-169, ROM-557. *EPYC 7702 - see endnote ROM-517. **A motherboard designed for 2nd Gen EPYC processors is required to enable all available functionality
  • 14. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7F52 16 3.50 3.90 256 3200 240 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 ※日本AMDで独自にまとめたスライドです
  • 15. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] 2 x DDR Memory / IO Die 2 x DDR 2 x DDR 2 x DDR 7F32 8 3.70 3.90 128 3200 180 CORES FREQ BOOST L3 DDR4 TDP 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 16MB L3 Z2L2 16MB L3Z2 L2 ※日本AMDで独自にまとめたスライドです
  • 16. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
  • 17. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] I/O ▪ 業界初の PCIe® Gen 4 対応 CPU ▪ PCIe Gen 3に対して2倍の双方向データレート ▪ NVMe、GPUやその他のアクセラレータを直接接続 可能な128レーンのPCIe® Gen4インターフェース メモリー ▪ 競合製品に対して33%メモリーチャネル数が多い1 ▪ CPU当たり最大4TBのDDR4メモリー容量 ▪ 最高 3200MHz DDR4 Memoryまでのサポート 4倍 競合製品に対する I/Oバンド幅2 45% 競合に対するメモリ バンド幅の向上3 Note: 1. Each AMD EPYC processor has 8 memory channels. Each Intel Xeon Scalable processor has 6 memory channels. 8 – 6 = 2 ÷ 6 = 0.33 AMD EPYC has 33% more memory channels. Class based on industry-standard pin-based (LGA) X86 processors 2. One EPYC 7002 Series has 433% [or: 4 times] more IO bandwidth than one Intel Scalable processors ROM-21 3. AMD EPYC™ 7002 Series processors have 45% more memory bandwidth than Intel Scalable processors in the same class. ROM-11
  • 18. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Intel® Xeon® 2nd Gen Scalable (“Cascade Lake SP”) AMD EPYC™ 7001 Series AMD EPYC™ 7002 Series Max Theoretical Memory BW per 2P server (1 DR DPC) 12 x DDR4-2933 = 282GB/s 16 x DDR4-2666 = 340GB/s 16 x DDR4-3200* = 410GB/s AMD EPYC™ Advantage +21% +45% DIMMDIMM A B C D E F G H 広帯域メモリーがパフォーマンスのスケーリングに貢献します *AMD POR memory speeds in a one DIMM per channel system implementation AMD EPYC™ 7002 シリーズ DDR4 1DIMM per Channel Server* DR RDIMM: 3200 MHz LRDIMM: 3200 MHz Not supported on all motherboards – see endnotes ROM-06 for details.
  • 19. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] EPYCがサポートする 広帯域 PCIE GEN4 はパフォーマンスの スケーリングに貢献します  8 x16 レーン 全てにおいて、 PCIe® Gen4 対応 − 1レーン当たり 64GB/s 双方向バンド幅, ソケット当たり512GB/s  8 x16 レーン全てにおいて、IOMMU サポート  レーン分割のサポート、 x16あたり最大8個の PCIeデバイス  フル PCIe P2P サポート、ソケット内, ソケット間ともに − 最大P2P ペイロードサイズ: 256B − 最大 DMA ペイロードサイズ: 512B  I/O ダイ上の AMD Infinity Fabric ™ は DMA and P2P に対して最適化さ れています  1 ソケットプラットフォームにおいては、全てのレーンをI/Oとして利 用可能  2 ソケットプラットフォームにおいては、最大162レーンをI/Oとして 利用可能 Not supported on all motherboards – see endnotes ROM-06 x16 x16 x8 x8 x4 x4x4 x4 x2 x2x2 x2x2 x2x2 x2 x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1 x16 x8 x8 x4 x4x4 x4 x2 x2x2 x2x2 x2x2 x2 x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1x1 x1 x1 x1x1 x1x1 x1x1 x1
  • 20. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] See EndnotesROM-173, ROM-174, ROM-175, ROM-176, ROM-177 and ROM-178
  • 21. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] AMD EPYC™ Tech Docs and White Papers 検索サイト： https://www.amd.com/en/processors/server-tech-docs/search。
  • 22. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
  • 23. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Source: https://www.amd.com/system/files/documents/AMD-EPYC-7002-AnsysFluent-Scalability-Performance.pdf
  • 24. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Refer from : https://www.amd.com/system/files/documents/amd-epyc-7Fx2-ansys-fluent.pdf
  • 25. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] Rome とAlveo は共にPCIe Gen4で通信し、HPC, ML, VM, Database, Networking NFV and 5G Domains のUse case においてKeyとなる高スループットを期待した測定。 XDMA-type data (large block size): Host Kernel, FPGA local Memory 間のデータ転送。 QDMA-type data(smaller block packet sizes): DPDK利用で、Host user space, FPGA local Memory間のPacket転送。 共に広帯域を確認。 参照先： https://www.amd.com/system/files/documents/epyc-7fx2-processors-network-performance-using-dpdk.pdf ※日本AMDで独自にまとめたスライドです
  • 26. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] NVIDIA
  • 27. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use]
  • 28. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] AMD Developer Central (No NDAの技術サイト) https://developer.amd.com/ AMD EPYC™ Tech Docs and White Papers (各種 ホワイトペーパー、技術資料のページ) https://www.amd.com/en/processors/server-tech-docs/search GPUOpen (AMDのGPUに関するページ) https://gpuopen.com/ AMD EPYC™ & HPE (AMDによるHPE製品に関するページ) https://www.amd.com/ja/campaigns/amd-and-hpe
  • 29. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ▪ EPYC-07 - Based on June 8, 2018 AMD internal testing of same-architectureproduct ported from 14 to 7 nm technology with similar implementationflow/methodology, using performancefrom SGEMM. ▪ ROM-04 Based on standard calculation method for determiningFLOPS. ROM-04 ▪ ROM-06 - Some supported features and functionalityof 2nd Gen AMD EPYC™ processors require a BIOS update from your server manufacturer when used with a motherboard designed for the 1st Gen AMD EPYC series processor. A motherboard designed for 2nd Gen EPYC processors is required to enable all available functionality. ▪ ROM-07 - Motherboards designed for 1st Gen EPYC processors may not be compatiblewith 2nd Gen AMD EPYC processors with a TDP greater than 200 watts. Contact the server manufacturer to confirm compatibility. ▪ ROM-342 1-n, 2-socket 2nd Gen EPYC™ 7742 powered server (http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20001.pdf) scoring 701 SPECrate®2017_int_base has up to 2.3x the SPECrate® 2017 Integer (Base) performanceof the highest previous generation score of 304 SPECrate®2017_int_base by a 1-n, 2-socket 1st Gen EPYC™ 7601 powered server result (http://www.spec.org/cpu2017/results/res2019q2/cpu2017-20190411-11817.html ) as of 1/26/20. ROM-342 ▪ ROM-741: Best published SPECrate®2017_int_base performanceas of 9/29/2020. AMD EPYC 2P processor scores reflect highest scoring processor at each availablecore level (core count is per processor): AMD EPYC 7742 (64C) score of 701 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20001.pdf AMD EPYC 7642 (48C) score of 586 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200622-23009.html AMD EPYC 7542 (32C) score of 437 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200706-23309.html AMD EPYC 7302 (16C) score of 246 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191220-20441.pdf AMD EPYC 7262 (8C) score of 135 http://spec.org/cpu2017/results/res2020q2/cpu2017-20200608-22777.html Intel® Xeon® Scalable 2P processors scores using the highest competitivepart in the Bronze, Silver and Gold metal classes: Intel Xeon Gold 6258R (28C) score of 397 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200915-23981.html Intel Xeon Gold 5220R (24C) score of 319 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200608-22594.html Intel Xeon Silver 4216 (16C) score of 193 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200804-23653.html Intel Xeon Bronze 3206R (8C) score of 60.1 http://spec.org/cpu2017/results/res2020q2/cpu2017-20200609-22881.html SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard Performance EvaluationCorporation. See w ww.spec.orgfor more informationROM-741
  • 30. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] ▪ ROM-742: Both published and estimated SPECrate®2017_int_base performanceas of March 2, 2020. AMD EPYC 7742 (64C) score of 355 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191223-20460.pdf AMD EPYC 7642 (48C) score of 306 http://spec.org/cpu2017/results/res2020q3/cpu2017-20200827-23775.html AMD EPYC 7542 (32C) score of 223 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191028-19322.pdf AMD EPYC 7402P (24C) score of 177 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191028-19326.pdf AMD EPYC 7302P (16C) score of 124 http://spec.org/cpu2017/results/res2019q4/cpu2017-20191125-20005.pdf AMD EPYC 7262 (8C) score of 68.3 http://spec.org/cpu2017/results/res2020q1/cpu2017-20191220-20437.pdf Intel® Xeon® Scalable 2P processors scores using the highest competitivepart in the Bronze, Silver and Gold metal classes. Estimates are based on highest score 2P processor result and divided by 1.99x for conservative estimate: Intel Xeon Gold 6258R (28C) estimated score of 199 (derived from 2P score of 397/1.99=199) https://spec.org/cpu2017/results/res2020q3/cpu2017-20200915-23981.html Intel Xeon Gold 5220R (24C) estimated score of 160 (derived from 2P score of 319/1.99=160) http://spec.org/cpu2017/results/res2020q3/cpu2017-20200608-22594.html Intel Xeon Silver 4216 (16C) score of 87.3 http://spec.org/cpu2017/results/res2019q2/cpu2017-20190524-14372.pdf Intel Xeon Bronze 3206R (8C) estimated score of 30.2 (derived from 2P score of 60.1/1.99= 30.2) http://spec.org/cpu2017/results/res2020q2/cpu2017-20200609-22881.html Results may vary with production system publications. The average of all published Intel results that have a 1P and 2P corresponding number is 2.08x with the maximum being 2.3x and the minimumbeing 1.95x so conservatively AMD estimated using 1.99x scaling. AMD expects Intel published numbersin the near future will likely be higher. SPEC®, SPECrate® and SPEC CPU® are registered trademarks of the Standard PerformanceEvaluation Corporation. See www.spec.orgfor more information. ROM-742 ▪ ROM-42 - Based on AMD internal testing of ANSYS FLUENT 19.1, lm6000_16m benchmark, as of July 17, 2019 of a 2P EPYC 7742 powered reference server versus a 2P Intel Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-49 - Based on AMD internal testing of LSTC LS-DYNA R9.3.0, neon benchmark, as of July 17, 2019 of a 2P EPYC 7742 powered referenceserver versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-56 - Based on AMD internal testing of Altair RADIOSS 2018, T10M benchmark, as of July 17, 2019 using a 2P EPYC 7742 powered reference server versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-63 - Based on AMD internal testing of ESI VPS 2018.0, NEON4m benchmark, as of July 17, 2019 using a 2P EPYC 7742 powered reference server versus a 2P Xeon Platinum 8280 powered server. Results may vary. ▪ ROM-113 AMD Internal testing as of 30July2019 of a 2P AMD EPYC 7742 powered reference platform versus a 2P Intel Platinum 8280 powered production server, on GROMACS version 2019.3 benchmark. Results may vary. ROM-113
  • 31. | AMD EPYC | OCT 2020 | univ AI inferenceウェビナー  [AMD Public Use] DISCLAIMER The information contained herein is for informational purposes only, and is subject to change without notice. While every precaution has been taken in the preparation of this document, it may contain technical inaccuracies, omissions and typographical errors, and AMD is under no obligation to update or otherwise correct this information. Advanced Micro Devices, Inc. makes no representations or warranties with respect to the accuracy or completeness of the contents of this document, and assumes no liability of any kind, including the implied warranties of noninfringement, merchantability or fitness for particular purposes, with respect to the operation or use of AMD hardware, software or other products described herein. No license, including implied or arising by estoppel, to any intellectual property rights is granted by this document. Terms and limitations applicable to the purchase or use of AMD’s products are as set forth in a signed agreement between the parties or in AMD's Standard Terms and Conditions of Sale. GD-18 ©2019 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, [insert all other AMD trademarks used in the material here per AMD’s Checklist for Trademark Attribution] and combinations thereof are trademarks of Advanced Micro Devices, Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.