Engineering:RDNA 2
Template:Infobox GPU microarchitecture RDNA 2 is a GPU microarchitecture designed by AMD, released with the Radeon RX 6000 series on November 18, 2020. Alongside powering the RX 6000 series, RDNA 2 is also featured in the SoCs designed by AMD for the PlayStation 5, Xbox Series X/S, and Steam Deck consoles.
Background
On July 7, 2019, AMD released the first iteration of the RDNA microarchitecture, a new graphics architecture designed specifically for gaming that replaced the aging Graphics Core Next (GCN) microarchitecture. With RDNA, AMD sought to reduce latency and improve power efficiency over their previous Vega series based on GCN 5th gen and Nvidia's competing Turing microarchitecture.[1]
RDNA 2 was first publicly announced in January 2020 with AMD initially calling RDNA 2 a "refresh" of the original RDNA architecture from the previous year.[2] At AMD's Financial Analysts Day held on March 5, 2020, AMD showed a client GPU roadmap that gave details on RDNA's successor, RDNA 2, that it would again be built using TSMC's 7 nm process and would be coming in 2020.[3] AMD told their investors that they were targeting a 50% uplift in performance-per-watt and increased IPC with the RDNA 2 microarchitecture.[4]
On October 28, 2020, AMD held an online unveiling event for the RDNA 2 architecture and Radeon RX 6000 series.[5][6] The event came 20 days after AMD's unveiling event for Ryzen 5000 series processors built on the Zen 3 microarchitecture.
Architectural details
Compute Unit
RDNA 2 contains a significant increase in the number of Compute Units (CUs) with a maximum of 80, a doubling from the maximum of 40 in the Radeon RX 5700 XT.[1] Each Compute Unit contains 64 shader cores.[7] CUs are organized into groups of two named Work Group Processors with 32 KB of shared L0 cache per WGP. Each CU contains two sets of an SIMD32 vector unit, an SISD scalar unit, textures units, and a stack of various caches.[8] New low precision data types like INT4 and INT8 are new supported data types for RDNA 2 CUs.[8]
The RDNA 2 graphics pipeline has been reconfigured and reordered for greater performance-per-watt and more efficient rendering by moving the caches closer to the shader engines. A new mesh shaders model allows shader rendering to be done in parallel using smaller batches of primitives called "meshlets". As a result, the mesh shaders feature enables greater control of the GPU geometry pipeline.[9]
Ray tracing
Real-time hardware accelerated ray tracing is a new feature for RDNA 2 which is handled by a dedicated ray accelerator inside each CU.[10] Ray tracing on RDNA 2 relies on the more open DirectX Raytracing protocol rather than the Nvidia RTX protocol.[11]
In February 2023, it was reported that driver updates had boosted ray tracing performance by up to 40% using DirectX Raytracing.[12]
Clock speeds
With RDNA 2 using the same 7 nm node as RDNA, AMD claims that RDNA 2 achieves a 30% frequency increase over its predecessor while using the same power.[13]
Cache and memory subsystem
In addition to the traditional L1 and L2 caches that GPUs possess, RDNA 2 adds a new global L3 cache that AMD calls "Infinity Cache".[14] AMD claims that RDNA 2's 128 MB of on-die Infinity Cache "dramatically reduces latency and power consumption".[15] The GPU having access to a large L2 or L3 cache allows it to more quickly access necessary data compared to accessing VRAM or system RAM. The Infinity Cache is made up of two sets of 64 MB cache that can run on its own clock rate independent from the GPU cores. The Infinity Cache has a peak internal transfer bandwidth of 1986.6 GB/s and results in less reliance being placed on the GPU's GDDR6 controllers.[8] Each Shader Engine now has two sets of L1 caches. The large cache of RDNA 2 GPUs give them a higher overall memory bandwidth compared to Nvidia's GeForce RTX 30 series GPUs.
Power efficiency
AMD claims that RDNA 2 achieves up to a 54% increase in performance-per-watt over the first RDNA microarchitecture.[15] 21% of that 54% improvement is attributed to performance-per-clock enhancements, in part due to the addition of Infinity Cache.[16]
Media engine
RDNA 2 uses the VCN 3.0, VCN 3.1, and VCN 3.1.2 video decoding blocks in its media engine.[17][18] It adds support for AV1 decoding at up to 8K resolution, though AV1 hardware encoding support would not come until RDNA 3 in 2022.[19][20] However, the low-end Navi 24 die and iGPUs based on RDNA 2.0 do not contain any media encoders and cannot decode AV1 as a result.[17]
| Navi 21 | Navi 22 | Navi 23 | Navi 24 | ||
|---|---|---|---|---|---|
| Launch | 2020 | 2021 | 2021 | 2022 | |
| Codename | Sienna Cichlid | Navy Flounder | Dimgrey Cavefish | Beige Goby | |
| Compute units (Stream processors) [FP32 cores] |
80 (5120) [10240] |
40 (2560) [5120] |
32 (2048) [4096] |
16 (1024) [2048] | |
| Process | TSMC N7 | TSMC N6 | |||
| Transistors | 26.8B | 17.2B | 11.06B | 5.4B | |
| Transistor density | 51.5 MTr/mm2 | 51.3 MTr/mm2 | 46.7 MTr/mm2 | 50.5 MTr/mm2 | |
| Die size | 520 mm2 | 335 mm2 | 237 mm2 | 107 mm2 | |
| Max TDP | 400 W | 250 W | 176 W | 107 W | |
| Package socket | BGA 2425 | BGA 1701 | BGA 1269 | Unknown | |
| Products | Desktop |
|
|
|
|
| Mobile | N/A |
|
|
| |
| Workstation (desktop) |
|
N/A |
|
| |
| Workstation (mobile) |
N/A |
|
| ||
Products
Desktop
Mobile
Workstation
Desktop Workstation
Mobile Workstation
Integrated graphics processing units (iGPUs)
| Model | Launch | Codename | Architecture & fab |
Die size[lower-alpha 1] |
Core | Fillrate[lower-alpha 2][lower-alpha 3][lower-alpha 4] | Processing power[lower-alpha 2][lower-alpha 5] (GFLOPS) |
Cache | TDP[lower-alpha 1] | CPUs/APUs | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Config[lower-alpha 6] | Clock[lower-alpha 2] (MHz) |
Texture (GT/s) |
Pixel (GP/s) |
Half | Single | Double | L0 | L1 | L2 | |||||||
| Radeon Graphics | 2022 | Raphael | RDNA 2 TSMC N6 |
122 mm2 | 2 CU 128:8:4:2 |
400 2200 |
3.2 17.6 |
1.6 8.8 |
204.8 1126.4 |
102.4 563.2 |
6.4 35.2 |
32 KB | 128 KB | 2 MB | 65–170 W | Ryzen 7000 series |
| Radeon 610M | 2022 | Mendocino Dragon Range |
100 mm2 | 1500 1900 |
12.0 15.2 |
6.0 7.6 |
768 972.8 |
384 486.4 |
24 30.4 |
15–55 W | Ryzen 7020 series Ryzen 7045 series | |||||
| Radeon 660M | 2022 | Rembrandt | 208 mm2 | 6 CU 384:24:16:6 |
1500 1900 |
36.0 45.6 |
24.0 30.4 |
2304 2918.4 |
1152 1459.2 |
72 91.2 |
96 KB | 384 KB | 28–54 W | Ryzen 6000 series Ryzen 7035 series | ||
| Radeon 680M | 12 CU 768:48:32:12 |
2000 2200 |
96.0 105.6 |
64.0 70.4 |
6144 6758.4 |
3072 3379.2 |
192 211.2 |
192 KB | 768 KB | 15–54 W | ||||||
- ↑ 1.0 1.1 Full chip, not just the iGP
- ↑ 2.0 2.1 2.2 Boost values (if available) are stated below the base value in italic.
- ↑ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
- ↑ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
- ↑ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
- ↑ Unified shaders : Texture mapping units : Render output units : Ray accelerators and Compute units (CU)
Consoles
| Model | Launch | Codename | Architecture | Fab | rowspan="2" Script error: No such module "Vertical header".[lower-alpha 1] | Die size[lower-alpha 1] |
Core | Fillrate[lower-alpha 2][lower-alpha 3][lower-alpha 4] | Processing power[lower-alpha 2][lower-alpha 5] (TFLOPS) |
Cache | Memory | TDP[lower-alpha 1] | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Config[lower-alpha 6] | Clock[lower-alpha 2] (MHz) |
Texture (GT/s) |
Pixel (GP/s) |
Half | Single | Double | L0 | L1 | L2 | L3 | Type | Size | Bandwidth (GB/s) |
Bus width |
Memory Clock (Gb/s) | ||||||||
| Steam Deck | 2022 | Aerith | RDNA 2 | TSMC N7 | 2.4 | 163 mm2 | 8 CU 512:32:16:8 |
1000 1600 |
32.0 51.2 |
16.0 25.6 |
2 3.2 |
1 1.6 |
0.063 0.1 |
128 KB | 512 KB | 1 MB | 8 MB | LPDDR5 | 16 GB | 88 | 128-bit | 5.5 | 15 W |
| 2023 | Sephiroth | TSMC N6 | Unknown | Unknown | 102.4 | 6.4 | |||||||||||||||||
| Xbox Series S | 2020 | Project Lockhart | TSMC N7 | 8.0 | 197 mm2 | 20 CU 1280:80:32:20 |
1565 | 125.2 | 50.08 | 8.013 |
4.006 |
0.25 |
320 KB | 1.25 MB | 4 MB | N/A | GDDR6 | 8 GB | 224 | 14.0 | 100 W | ||
| Xbox Series X | Project Scarlett | 15.3 | 360 mm2 | 52 CU 3328:208:64:52 |
1825 | 379.6 | 116.8 | 24.294 |
12.147 |
0.759 |
832 KB | 3.25 MB | 5 MB | 10 GB | 560 | 320-bit | 200 W | ||||||
| PlayStation 5 | 2020 | Oberon | 10.6 | 308 mm2 | 36 CU 2304:144:64:36 |
2233 | 321.552 | 142.912 | 20.579 |
10.29 |
0.643 |
576 KB | 2.25 MB | 4 MB | 16 GB | 448 | 256-bit | 180 W | |||||
| 2022 | Oberon Plus | TSMC N6 | 264 mm2 | ||||||||||||||||||||
- ↑ 1.0 1.1 1.2 Full chip, not just the GPU
- ↑ 2.0 2.1 2.2 Boost values (if available) are stated below the base value in italic.
- ↑ Texture fillrate is calculated as the number of Texture Mapping Units multiplied by the base (or boost) core clock speed.
- ↑ Pixel fillrate is calculated as the number of Render Output Units multiplied by the base (or boost) core clock speed.
- ↑ Precision performance is calculated from the base (or boost) core clock speed based on a FMA operation.
- ↑ Unified shaders : Texture mapping units : Render output units : Ray accelerators and Compute units (CU)
See also
- Ampere - competing Nvidia microarchitecture released in a similar time-frame with the competing GeForce RTX 30 series
References
- ↑ 1.0 1.1 James, Dave (October 31, 2019). "AMD Navi RDNA architecture – a GPU designed purely for PC gamers" (in en-US). https://www.pcgamesn.com/amd/navi-rdna-architecture-release-date-specs-performance.
- ↑ Alcorn, Paul (January 29, 2020). "AMD to Introduce New Next-Gen RDNA GPUs in 2020, Not a Typical 'Refresh' of Navi" (in en-US). https://www.tomshardware.com/news/amds-navi-to-be-refreshed-with-next-gen-rdna-architecture-in-2020.
- ↑ Smith, Ryan (March 5, 2020). "AMD's 2020-2022 Client GPU Roadmap: RDNA 3 & Navi 3X On the Horizon With More Perf & Efficiency" (in en-US). https://www.anandtech.com/show/15592/amds-2020-client-gpu-roadmap-rdna-3-on-the-horizon.
- ↑ Smith, Ryan (March 5, 2020). "AMD's RDNA 2 Gets A Codename: "Navi 2X" Comes This Year With 50% Improved Perf-Per-Watt" (in en-US). https://www.anandtech.com/show/15591/amds-rdna-2-gets-a-codename-navi-2x-comes-this-year-with-50-improved-perfperwatt.
- ↑ Thubron, Rob (September 9, 2020). "AMD confirms Radeon 6000 reveal and Zen 3 launch for October" (in en-AU). https://www.techspot.com/news/86683-big-radeon-6000-reveal-could-arrive-tomorrow.html.
- ↑ Lyles, Taylor (September 9, 2020). "AMD's next-generation Zen 3 CPUs and Radeon RX 6000 'Big Navi' GPU will be revealed next month" (in en-US). The Verge. https://www.theverge.com/2020/9/9/21429127/amd-zen-3-cpu-big-navi-gpu-events-october.
- ↑ Walton, Jarred (November 3, 2020). "AMD Unveils Big Navi: RX 6900 XT, RX 6800 XT and RX 6800 Take On Ampere" (in en-US). https://www.tomshardware.com/news/amd-rx-6000-rdna-2-big-navi-gpus-revealed.
- ↑ 8.0 8.1 8.2 Evanson, Nick (December 6, 2020). "Nvidia Ampere vs. AMD RDNA 2: Battle of the Architectures" (in en-AU). https://www.techspot.com/article/2151-nvidia-ampere-vs-amd-rdna2/.
- ↑ Riley, Colin (2020). "Sampler Feedback & Mesh Shaders" (in en-US). https://gpuopen.com/wp-content/uploads/slides/AMD_RDNA2_DirectX12_Ultimate_SamplerFeedbackMeshShaders.pdf.
- ↑ Evenden, Ian (November 17, 2020). "AMD teases 'mind-blowing' RDNA 2 ray tracing in its new tech demo" (in en-US). https://www.pcgamer.com/amd-has-all-the-shinies/.
- ↑ Salter, Jim (October 28, 2020). "AMD's newest graphics cards: RDNA2 power from $579 to $999" (in en-US). https://arstechnica.com/gaming/2020/10/amds-newest-graphics-cards-rdna2-power-from-579-999/.
- ↑ Mujtaba, Hassan (February 16, 2023). "AMD Radeon RX 6000 "RDNA 2" GPUs Get Up To 40% Ray Tracing Performance Boost In 3DMark DXR With Latest Drivers" (in en-US). https://wccftech.com/amd-radeon-rx-6000-rdna-2-gpus-40-percent-ray-tracing-performance-boost-3dmark-dxr-latest-drivers/.
- ↑ Altavilla, Dave; Chiapetta, Marco (October 28, 2020). "AMD Unveils Three Powerful Radeon RX 6000 Series Big Navi Cards To Topple NVIDIA's Best" (in en-US). https://hothardware.com/news/amd-unveils-radeon-rx-6800-radeon-rx-6800-xt-and-radeon-rx-6900-xt.
- ↑ Chacos, Brad (November 18, 2020). "RDNA 2 deep-dive: What's inside AMD's Radeon RX 6000 graphics cards" (in en-US). https://www.pcworld.com/article/393733/rdna-2-deep-dive-inside-amd-radeon-rx-6000-graphics-cards.html.
- ↑ 15.0 15.1 "AMD Unveils Next-Generation PC Gaming with AMD Radeon™ RX 6000 Series – Bringing Leadership 4K Resolution Performance to AAA Gaming". AMD (Press release). Santa Clara, CA. October 28, 2020. Retrieved April 10, 2023.
- ↑ Frausto-Robledo, Anthony (October 18, 2021). "AMD's RDNA 2 Architecture—Behind the Radeon Pro W6600’s Stellar Performance" (in en-US). https://architosh.com/2021/10/amds-rdna-2-architecture-behind-the-radeon-pro-w6600s-stellar-performance/.
- ↑ 17.0 17.1 Shilov, Anton (May 4, 2022). "First Details About AMD's Next Generation Video Engine Revealed" (in en-US). https://www.tomshardware.com/news/next-amd-video-engine-may-lack-av1.
- ↑ Larabel, Michael (September 15, 2020). "AMD Radeon Navi 2 / VCN 3.0 Supports AV1 Video Decoding" (in en-US). https://www.phoronix.com/news/AV1-Decode-For-AMD-VCN-3.0.
- ↑ "AMD Navi 2X GPUs (RDNA2) to support AV1 decoding" (in en-US). September 15, 2020. https://videocardz.com/newz/amd-navi-2x-gpus-rdna2-to-support-av1-decoding.
- ↑ "AMD RDNA2 Graphics Architecture Features AV1 Decode Hardware-Acceleration" (in en-US). October 12, 2020. https://www.techpowerup.com/273243/amd-rdna2-graphics-architecture-features-av1-decode-hardware-acceleration.
 |  |
