AMD Radeon™ RX 5700 Series 7nm Energy-Efficient High-Performance GPUs
•Télécharger en tant que PPTX, PDF•
4 j'aime•1,096 vues
Sal Dasgupta,Teja Singh, Ashish Jain, Samuel Naffziger, Deepesh John, Chetan Bisht, Pradeep Jayaraman, Michael Mantor Presented at ISSCC 2020
Recommandé
Recommandé
Contenu connexe
Tendances
Tendances (20)
Similaire à AMD Radeon™ RX 5700 Series 7nm Energy-Efficient High-Performance GPUs
Similaire à AMD Radeon™ RX 5700 Series 7nm Energy-Efficient High-Performance GPUs (20)
Plus de AMD
Plus de AMD (15)
Dernier
Dernier (20)
AMD Radeon™ RX 5700 Series 7nm Energy-Efficient High-Performance GPUs
- 1. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 1 of 38 AMD Radeon™ RX 5700 Series 7nm Energy-Efficient High-Performance GPUs Sal Dasgupta1,Teja Singh2, Ashish Jain2, Samuel Naffziger3, Deepesh John2, Chetan Bisht4, Pradeep Jayaraman1, Michael Mantor4 1AMD, Santa Clara, CA, 2AMD, Austin, TX, 3AMD, Fort Collins, CO, 4AMD, Orlando, FL Presented at ISSCC 2020
- 2. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 2 of 38 Outline • Overview of AMD Radeon™ RX 5700 Series • AMD RDNA Architecture • Power Management features • GDDR6 (G6) PHY • Physical design
- 3. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 3 of 38 AMD Radeon™ RX 5000 series • GPUs are everywhere • GPUs need to service a wide range of form factors and workloads • Fundamental challenge is to get higher and higher performance at lower and lower power PC Gaming Content Creation Console Gaming Cloud Gaming Mobile Devices
- 4. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 4 of 38 Improvements • Up to 1.5x greater performance per watt than its predecessor • Up to 1.25x performance per clock compared to previous 14nm processors • Up to 1.23x higher max frequencies than its predecessor • Up to 1.23x lower power consumption than its predecessor Achieved through • 7nm process • Higher clocks • Focused design for lower dynamic power • Intelligent SOC design with power and performance at the forefront • Improved power management • All new AMD RDNA graphics architecture – higher performance for the same cycles AMD Radeon™ RX 5700 series See endnote RX-327, RX-325 and RX-362
- 5. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 5 of 38 AMD Radeon™ RX 5700 Series Overview
- 6. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 6 of 38 AMD Radeon™ RX 5700 XT PCIE® Gen4 x16 32 GB/s Display DP1.4 HDMI 4K 60fps Multimedia 4K H264 Encode/Decode H265/HEVC Encode/Decode GDDR6 256b 14 Gbps See endnote GD-81
- 7. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 7 of 38 AMD Radeon™ RX 5700 XT Floorplan Graphics BUS Interface Display G6 PHYG6 PHY G6 Control
- 8. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 8 of 38 AMD Radeon RX 5700 XT AMD Radeon RX Vega 64 AMD Radeon RX 580 Compute Units 40 64 36 Texture Units 160 256 144 ROPs 64 64 32 Memory Clocks 14 Gbps GDDR6 1.89 Gbps HBM2 8 Gbps GDDR5 Memory Bus Width 256 bit 2048 bit 256 bit Frame Buffer 8 GB 8 GB 8 GB Boost Clock 1905 Mhz 1546 Mhz 1257 Mhz Typical Board Power 225W 295W 185W Transitor Count 10.3B 12.5B 5.7B Manufacturing Process TSMC 7nm GF 14 nm GF 14 nm Architecture RDNA GCN GCN AMD Radeon™ RX 5700 XT Summary 0% 20% 40% 60% 80% 100% Additional Frequency and Power Improvement 7nm process Performance per Clock Enhancement 0 20 40 60 80 100 120 140 160 Delivered Performance GCN RDNA +50% Same-Power, Same-Configuration Performance Gains Performance Contributors See endnote GD-151, RX-325
- 9. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 9 of 38 AMD Radeon™ RX 5700 Series RDNA Graphics Architecture
- 10. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 10 of 38 40 RDNA Compute Units • 80 Scalar Processors • 2560 Stream Processors • 160 64b Bilinear Filter units Multilevel Cache • 4MB L2, 512KB L1, (V$, I$, K$) L0 • 2x V$L0 Load Bandwidth • DCC Everywhere Streamlined Graphics Engine • Geometry Engine (4 Prim shader out, 8 Prim shader in) • 64 Pixel Units • 4 Asynchronous Compute Engines Designed for higher frequencies at lower power AMD Radeon™ RX 5700 XT
- 11. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 11 of 38 AMD Radeon™ RX 5700 XT Floorplan
- 12. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 12 of 38 AMD Radeon™ RX 5700 XT Functional Floorplan InfinityFabric PCIE Gen4 Display Engine MultimediaEngine Geometry Processor ShaderEngine Command Processor HWS DMA 64-bit Memory Controller 64-bit Memory Controller L1 Prim Unit L1 Prim Unit L1 Prim Unit L1 Prim Unit Rasterizer Rasterizer ShaderEngine L2 L2 L2 L2 L2 L2 L2 L2 Compute Units L2 L2 L2 L2 L2 L2 L2 L2 64-bit Memory Controller 64-bit Memory Controller RBs ACE Compute Units RBsRBs RBs Compute Units Compute Units Rasterizer Rasterizer
- 13. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 13 of 38 AMD Radeon™ RX 5700 XT Functional Floorplan InfinityFabric PCIE Gen4 Display Engine MultimediaEngine Geometry Processor ShaderEngine Command Processor HWS DMA 64-bit Memory Controller 64-bit Memory Controller L1 Prim Unit L1 Prim Unit L1 Prim Unit L1 Prim Unit Rasterizer Rasterizer ShaderEngine L2 L2 L2 L2 L2 L2 L2 L2 Compute Units L2 L2 L2 L2 L2 L2 L2 L2 64-bit Memory Controller 64-bit Memory Controller RBs ACE Compute Units RBsRBs RBs Compute Units Compute Units Rasterizer Rasterizer
- 14. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 14 of 38 RDNA – Work Group Processor Up to 20 wave controllers Improved instruction arbitration 10KB scalar register file • 128 32b register per wavefront 128 KB Vector register file ~2X instruction rate vs GCN • Dual SIMD32 Single cycle issue • Wave32 on SIMD32 Bytes Per Flop • 128B Load/Store • 64B Filter Rate Scalar Registers Redraw the graphic so its not a blatant copy of Hot Chips RDNA Workgroup Processor Scalar Units Vector ALUs (SIMD32) Shader Sequencers Texture Mapping Units Vector Registers Texture L0 Cache Scalar Data Cache Local Data Share Shader Instruction Cache 32 wide single and dual half ALU • Full rate 32b FMA, Dual 16b FMA 8 wide transcendental ALU • Single cycle issue • Multi-cycle co-execution SIMD Unit WGP
- 15. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 15 of 38 Schedulers Local Data ShareScalar Data Cache Shader Instruction Cache Texture Mapping Units Texture Filter Units Stream Processors Vector Registers Scalar Registers Scalar Units Scheduler Local Data Share Texture Filter Units L1 CacheVector ALU Texture Fetch Load/Store Units Scalar Registers Scalar Unit Vector Registers Vector Units Branch & Message Unit RDNA Compute Unit GCN Compute Unit RDNA WGP
- 16. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 16 of 38 Cycle N Lanes 15 - 0 Cycle N+1 Lanes 31 – 16 Cycle N+2 Lanes 47 - 32 Cycle N+3 Lanes 63 - 48 WAVE64-4CYCLEISSUE Operand gathering 4 cycle issue VGPRVGPR VGPR VGPR Operand gathering 4 cycle issue Operand gathering 4 cycle issue Operand gathering 4 cycle issue SIMD0 SIMD 1 SIMD 2 SIMD 3 VGPRVGPR VGPR VGPR VGPRVGPR VGPR VGPRVGPRVGPR VGPR VGPR S Cycle0 SIMD0 Waves Cycle1 SIMD1 Waves Cycle2 SIMD2 Waves Cycle3 SIMD3 Waves SHARED SCALAR Cycle0 SIMD0 Waves Cycle1 SIMD1 Waves Cycle2 SIMD2 Waves Cycle3 SIMD3 Waves All work-items of a wave64 have an opportunity to do work once every 4 clocks due to hardware interleaving Special Function Unit alternate execution unit running at ¼ rate A wave from a SIMD has an opportunity to accomplish a scalar instruction once every 4 clocks GCN Instruction Issue
- 17. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 17 of 38 SIMD 0 SIMD 1S S Operand gathering 1 cycle issue VGPRVGPR VGPR VGPR VGPRVGPR VGPR VGPR Operand gathering 1 cycle issue SIMD0 Wave32 – every cycle issue Vector Instruction Issue any cycle Or SFU Issue once every 4 cycles SIMD0 Wave32 Every cycle issue SIMD1 Wave32 – every cycle issue Vector Instruction Issue any cycle Or SFU Issue once every 4 cycles SIMD1 Wave32 Every cycle issue Vector Units - All work-items of one wave32 have an opportunity to do work every clock Special Function Unit uses 1 issue cycle and then executes in parallel Each SIMD equipped with a scalar unit for an instruction execution every cycle RDNA Instruction Issue
- 18. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 18 of 38 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 s_add_i32 s0, s1, s2 … … … v_mul_f32 v0, v1, s0 … (simd busy 4 cycles) … … v_add_f32 v5, v4, v3 … … … v_sub_f32 v6, v7, v0 … … … s_add_i32 s0, s1, s2 v_mul_f32 v0, v1, s0 v_add_f32 v5, v4, v3 v_sub_f32 v6, v7, v0 s_add_i32 s0, s1, s2 … (salu dependency stall on S0) v_mul_f32 v0, v1, s0 v_add_f32 v5, v4, v3 … (valu dependency stall on V0) … … v_sub_f32 v6, v7, v0 s_add_i32 s0, s1, s2 … (salu dependency stall on S0) v_mul_f32 v0, v1, s0 (lo) v_mul_f32 v0, v1, s0 (hi) v_add_f32 v5, v4, v3 (lo) v_add_f32 v5, v4, v3 (hi) … (valu dependency stall on V0 lo) v_sub_f32 v6, v7, v0 (lo) v_sub_f32 v6, v7, v0 (hi) SHORTEST WAVE ISSUE LATENCY 44% REDUCTION IN ISSUE CYCLES RDNA Instruction Issue Example
- 19. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 19 of 38 PCIe® 4.0 Async Compute L2 L1 Texture Geometry Rasterizer & Render Backends PCIe® 4.0 SOC Fabric GDDR6 Command Interfaces Shader Complex RDNA Redesigned Cache Hierarchy New L1 Cache Hierarchy
- 20. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 20 of 38 RDNA Cache Hierarchy • Doubled the Load Bandwidth from L0 to ALU • Improved BW Amplification • Reduced Latency and Power • Reduced Congestion at L2 Level • Reduced Data Movement 4x64B/C 16x32/C 32B/CLK 32B/CLK 128B/CLK 64B/CLK 128B/CLK 64B/CLK 2X 2X Relative Cache Latency -24%-21% -7% See RX-329 in Endnotes.
- 21. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 21 of 38 AMD Radeon™ RX 5700 Series Power Management
- 22. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 22 of 38 AMD Radeon™ RX 5700 Series Power Management • Main functions • Manage clocks/voltages to maximize performance during active workloads • Draw minimum power during low activity conditions • Challenges • Highly variable GPU workloads • Parallel work leads to high power/current demand • Ever increasing needs for memory bandwidth • Power features include • AVFS to choose the most optimal per-part voltage • DVFS to choose the best operating point given current environment • Voltage droop mitigation and impact reduction • Agile responses to the current draw demands of the moment • Graceful throttling of the graphics core at thermal limits • Aggressive clock and power gating
- 23. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 23 of 38 Power Management: Fine-grained DPM Previous Generation: Coarse-grained DPM • On previous architectures, operation was limited to a few (typically 8) pre-defined frequency values • Limited choice for Power Management Controller to choose from • Power/Thermal constrained effective frequency determined by dithering between neighboring coarse states AMD RadeonTM RX 5700 Series: Fine-grained DPM • Much finer-grained DPM state selection across the V/F curve • Improved perf/W efficiency by up to 5% as compared to previous generation by staying on the optimal curve • More accurate frequency selection between what the workload needs and what it gets DPM7 DPM6 DPM5 DPM4 DPM3 DPM2 DPM1 DPM0 Coarse- grained DPM Fine- grained DPM Fmax Fidle Fmin *Based on AMD internal data.
- 24. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 24 of 38 Power Management – Per Part Fmax Previous generations • Fmax determined by slowest part of distribution • Lower Cac workloads may leave power on the table for a large population of parts AMD RadeonTM RX 5700 Series • Each individual part allowed to achieve max potential (up to 15% higher) by selecting its own Vmax-limited Fmax based on the speed of the part • Enables applications with lower Cac to sustain higher clocks rather than be limited to artificially low limits set by slowest parts Based on AMD internal data
- 25. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 25 of 38 AMD Radeon™ RX 5700 Series GDDR6 Interface
- 26. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 26 of 38 • 14Gbp/s 256b, 448 GB/s • Up to 75% BW per pin improvement over GDDR5 • Up to 60% performance/Watt GDDR6 Memory Based on AMD internal data
- 27. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 27 of 38 G6 PHY Read Data Strobe (RDQS) mode to save power when high memory bandwidth is not required T-coil provides bandwidth enhancement and improved return loss enabling the high data rates on a single-ended interface • up to 16% in height • up to 26% in width 40.2 0.300 50.8 0.349 Based on AMD internal data
- 28. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 28 of 38 AMD Radeon™ RX 5700 Series Physical Design
- 29. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 29 of 38 AMD Radeon™ RX 5700 Physical Design Physical Design challenges • New architecture • Frequency uplift beyond natural process uplift • Large design (10.3B transistors) with wide busses and complex crossbar structures • Decreasing dynamic switching capacitance while providing that frequency uplift • Operational logistics of managing such a large design Approach • High-performance clock distribution • Intelligent SRAM generation • Automated place and route while offering hooks for customization • Exploiting the benefits of the technology while compensating for the new challenges it brings • Careful use of mixed VTH cells to close timing gaps while maintaining power requirements • Balancing resource constraints and the desire for physical reuse against area and performance targets • Power aware floorplanning and bus planning working in conjunction with logic design teams
- 30. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 30 of 38 Global Clock Distribution Optimizations Global clock distribution adopts low skew multiple mesh design style built of highly optimized configurable clock cells • Smaller mesh regions and optimized driver design reduces global distribution skew and variability costs by 30% for most synchronous paths Clock mesh wire power reduced up to 40% (normalized to area) by optimizing high level metal usage and reducing parasitic capacitance in clock drivers MESH1 MESH2 MESH5 MESH7 MESH6 MESH4 MESH3 S P I N E SPINE S P I N E S P I N E
- 31. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 31 of 38 Local Clock Distribution Optimizations Configurable mixed-depth structured clock tree adopted for local clock distribution • Reduces median clock insertion by up to 50% which helps reduce jitter and PVT variability • Multiple levels of clock gating provides both coarse and fine control Bottom up expansion of clock tree adopted instead of region-based cloning • Local clock tree CAC decreased by up to 10% with load-based cloning S P I N E S P I N E S P I N E
- 32. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 32 of 38 Bus planning Bus widths • Large busses (upto 2048b) and complex crossbar structures • Very large number of physical partitions need to be managed in an operational cadence • 60 unique designs with ~1-2M instance count, despite considerable reuse. • Requires prototyping and proving of achieving performance targets well in advance of netlist drops
- 33. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 33 of 38 AMD Radeon™ RX 5700 Series Conclusion
- 34. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 34 of 38 AMD Radeon™ RX 5700 XT – RDNA Performance Improvements Based on internal testing. See endnote RX-363 56CU vs. 40CU 300W TBP vs. 225W TBP
- 35. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 35 of 38 Conclusion AMD RadeonTM RX 5700 Series increased frequency while lowering active power and improving performance per clock Enabled by • Performance and power efficient next-generation AMD RDNA architecture • Increased memory bandwidth while maintaining power envelope and keeping costs low • Advanced power management techniques that allowed residency in the optimum power states while not limiting performance to the worst of the population • Attaining timing closure through reducing skew and jitter, improved bus planning, judicious use of Vt cells, and innovative floorplanning See endnote RX-327, RX-325 and RX-362
- 36. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 36 of 38 Acknowledgment We would like to thank our talented AMD design teams across Austin, Bangalore, Boston, Fort Collins, Hyderabad, Markham, Santa Clara, and Shanghai.
- 37. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 37 of 38 Notes • GD-81 HEVC (H.265), H.264, and VP9 acceleration are subject to and not operable without inclusion/installation of compatible HEVC players. GD-81 • GD-151: Boost Clock Frequency is the maximum frequency achievable on the GPU running a bursty workload. Boost clock achievability, frequency, and sustainability will vary based on several factors, including but not limited to: thermal conditions and variation in applications and workloads. • RX-325: Testing done by AMD performance labs 5/23/19, using the Division 2 @ 25x14 Ultra settings. Performance may vary based on use of latest drivers. RX-325 • RX-327: Testing done by AMD performance labs 5/23/19, showing a geomean of 1.25x per/clock across 30 different games @ 4K Ultra, 4xAA settings. Performance may vary based on use of latest drivers. RX- 327 • RX-329 Testing conducted by AMD Performance Labs as of 05/30/2019 on Radeon RX 5700XT with AMD Driver 19.10 (1902270946) on Intel i7-6900k, and on Radeon Vega Frontier Edition with AMD Driver 19.30 (1904231814) on Intel i7-5960k. Both systems used 2x8GB DDR4 2133Mhz RAM, Asus ROG Rampage V Edition Motherboard, and Windows 10 Enterprise. Performance may vary. RX-329. • RX-362: Testing done by AMD performance labs on June 4, 2019. Systems were tested with: Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz (6 core) with 16GB DDR4 @ 2133 MHz using a Asus X99-E Motherboard running Windows 10 Enterprise 64-bit (Ver. 1809, build 17763.053). Using the following graphics cards: Navi 10 (Driver 19.30_1905161434 (CL# 1784070)) with 40 compute units, versus a Vega 64 (Driver 19.4.1) with 40 compute units enabled. Breakdown based on AMD internal data June 4, 2019. Performance may vary. RX-362 • RX-363 Testing done by AMD performance labs 5/30/2019 on Core i9-9900K (3.6 GHz), 16GB DDR4-3200MHz, GIGABYTE Z390 AORUS ELITE, Win 10 64-bit, AMD Driver 19.30 for RX5700, and 19.10-190502a for Vega 56. Measuring FPS using: Dirt Rally 2, Sid Meier's Civilization 6, Metro Exodus, Tom Clancy's Ghost Recon Wildlands, Shadow of the Tomb Raider Battlefield 5, Assassin's Creed Odyssey, Call of Duty: Black Ops 4 The Division 2, Far Cry New Dawn. All at max settings. PC manufacturers may vary configurations yielding different results.. Performance may vary based on use of latest drivers. RX-363
- 38. 8.4: Radeon RX 5700 Series : The AMD 7nm Energy-Efficient High-Performance GPUs© 2020 IEEE International Solid-State Circuits Conference 38 of 38 Disclaimer and Endnotes 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 All rights reserved. AMD, the AMD Arrow logo, 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.