GeForce 20 series: Difference between revisions
m →Chipset table: It most likely won't be released in 2018, but announced at CES 2019 and released later in the year |
mNo edit summary |
||
| Line 37: | Line 37: | ||
== Architecture == |
== Architecture == |
||
{{Main|Turing (microarchitecture)}} |
{{Main|Turing (microarchitecture)}} |
||
The RTX 20 series is based on the [[Turing (microarchitecture)|Turing microarchitecture]] and features real-time [[ray tracing (graphics)|ray tracing]].<ref>{{Cite news|url=https://www.theverge.com/2018/8/20/17758724/nvidia-geforce-rtx-2080-specs-pricing-release-date-features|title=Nvidia announces RTX 2000 GPU series with ‘6 times more performance’ and ray-tracing|work=The Verge|access-date=2018-08-20}}</ref> The cards are manufactured on an |
The RTX 20 series is based on the [[Turing (microarchitecture)|Turing microarchitecture]] and features real-time [[ray tracing (graphics)|ray tracing]].<ref>{{Cite news|url=https://www.theverge.com/2018/8/20/17758724/nvidia-geforce-rtx-2080-specs-pricing-release-date-features|title=Nvidia announces RTX 2000 GPU series with ‘6 times more performance’ and ray-tracing|work=The Verge|access-date=2018-08-20}}</ref> The cards are manufactured on an optimized [[16nm]] node from [[TSMC]], named 12nm [[FinFET]] NVIDIA (FFN).<ref>{{Cite news|url=https://www.anandtech.com/show/13249/nvidia-announces-geforce-rtx-20-series-rtx-2080-ti-2080-2070|title=NVIDIA Announces the GeForce RTX 20 Series: RTX 2080 Ti & 2080 on Sept. 20th, RTX 2070 in October|work=Anandtech|access-date=2018-12-06}}</ref> This real-time ray tracing is accelerated by the use of new RT cores, which are designed to process [[quadtree]]s and spherical hierarchies, and speed up collision tests with individual triangles. |
||
The ray tracing performed by the RT cores can be used to produce reflections, refractions and shadows, replacing traditional raster techniques such as [[cube mapping|cube maps]] and [[depth map|depth maps]]. Instead of replacing rasterization entirely, however, the information gathered from ray tracing can be used to augment the shading with information that is much more [[photorealism|photo-realistic]], especially in regards to off-camera action. This can only be utilized on games that support ray-tracing, which led to some controversy at launch because few games supported the feature at the time. |
The ray tracing performed by the RT cores can be used to produce reflections, refractions and shadows, replacing traditional raster techniques such as [[cube mapping|cube maps]] and [[depth map|depth maps]]. Instead of replacing rasterization entirely, however, the information gathered from ray tracing can be used to augment the shading with information that is much more [[photorealism|photo-realistic]], especially in regards to off-camera action. This can only be utilized on games that support ray-tracing, which led to some controversy at launch because few games supported the feature at the time. |
||
| Line 135: | Line 135: | ||
| |
| |
||
| |
| |
||
| |
|No |
||
| |
| |
||
| |
| |
||
Revision as of 18:56, 21 December 2018
 | |
| Release date | September 20, 2018 |
|---|---|
| Codename | TU10x |
| Architecture | Turing |
| Models | GeForce RTX series |
| Transistors |
|
| Cards | |
| Mid-range |
|
| High-end |
|
| Enthusiast | GeForce RTX 2080 Ti[3] Titan RTX |
| API support | |
| DirectX | Direct3D 12.0 (feature level 12_1) |
| OpenCL | OpenCL 1.2 |
| OpenGL | OpenGL 4.6 |
| Vulkan | Vulkan 1.1 |
| History | |
| Predecessor | GeForce 10 series |
| Successor | TBA |
The GeForce RTX 20 Series is a family of graphics processing units developed by Nvidia, and was announced at Gamescom on August 20, 2018.[4] It is the successor to the GeForce 10 series,[5] and started shipping on September 20, 2018.[6]
Architecture
The RTX 20 series is based on the Turing microarchitecture and features real-time ray tracing.[7] The cards are manufactured on an optimized 16nm node from TSMC, named 12nm FinFET NVIDIA (FFN).[8] This real-time ray tracing is accelerated by the use of new RT cores, which are designed to process quadtrees and spherical hierarchies, and speed up collision tests with individual triangles.
The ray tracing performed by the RT cores can be used to produce reflections, refractions and shadows, replacing traditional raster techniques such as cube maps and depth maps. Instead of replacing rasterization entirely, however, the information gathered from ray tracing can be used to augment the shading with information that is much more photo-realistic, especially in regards to off-camera action. This can only be utilized on games that support ray-tracing, which led to some controversy at launch because few games supported the feature at the time.
The second-generation Tensor cores (succeeding Volta's) work in cooperation with the RT cores, and are used to fill in the blanks in a partially ray traced image, a technique known as de-noising. Another of the applications of the Tensor cores is a new method to replace anti-aliasing called DLSS (deep learning super-sampling).[9] The Tensor cores perform the result of deep learning on supercomputers to codify how to, for example, increase the resolution of images. In the Tensor cores' primary usage, a problem to be solved is analyzed on a supercomputer, which is taught by example what results are desired, and the supercomputer determines a method to use to achieve those results, which is then done with the consumer's Tensor cores. These methods are delivered "over the air" to consumers.
New features in Turing:
- CUDA Compute Capability 7.5
- Ray tracing (RT) cores – bounding volume hierarchy acceleration[10]
- Tensor (AI) cores – neural net artificial intelligence – large matrix operations
- New memory controller with GDDR6 support
- DisplayPort 1.4a with Display Stream Compression (DSC) 1.2
- PureVideo Feature Set J hardware video decoding
- GPU Boost 4
- NVLink Bridge
- VirtualLink VR
- Improved NVENC codec
Nvidia segregates the GPU dies for Turing into A and non-A variants, which is appended or excluded on the hundreds part of the GPU code name. Non-A variants are not allowed to be factory overclocked, whilst A variants are.[11]
Development platform
Nvidia RTX is the development platform introduced with the GeForce 20 series. RTX uses Microsoft's DXR, Nvidia's OptiX, and Vulkan for access to ray tracing.[12] The ray tracing technology used in the RTX Turing GPUs was in development at Nvidia for 10 years.[13]
Chipset table
| Model | Launch | Code name(s) | Fab (nm) | Transistors (billion) | Die size (mm2) | Bus interface | Shader processors | Texture mapping units | Render output units | Ray Tracing Cores | Tensor Cores[a] | SM Count[b] | L2 Cache (MB) | Clock speeds | Fillrate | Memory | Processing power (GFLOPS) | Ray tracing performance | TDP (watts) | NVLink Support | Launch price (USD) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Base core clock (MHz) | Boost core clock (MHz) | Memory (MT/s) | Pixel (GP/s)[c] | Texture (GT/s)[d] | Size (GiB) | Bandwidth (GB/s) | Bus type | Bus width (bit) | Single precision (Boost) | Double precision (Boost) | Half precision (boost) | Rays/s (Billions) | RTX-OPS (Trillions) | Tensor FLOPS (Trillions) | MSRP | Founders Edition | ||||||||||||||||
| GeForce RTX 2060[14] | January 2019 | TU106 | 12 | PCIe 3.0 x16 | 1920 | 120 | 48 | 30 | 240 | 30 | 1320 | 1620 | 14000 | 158.4 | 6 | 336 | GDDR6 | 192 | 5069 (6221) | 5 | No | |||||||||||
| GeForce RTX 2070[15] | October 17, 2018 | TU106-400-A1 | 10.8 | 445 | 2304 | 144 | 64 | 36 | 288 | 36 | 4 | 1410 | 1620 | 90.24 | 203.4 | 8 | 448 | 256 | 6497 (7465) | 203 (233) | 12994 (14930) | 6 | 45 | 59.7 | 175 | No | $499 | — | ||||
| TU106-400A-A1 | 1620+ | 90.24+ | 203.4+ | 6497 (7465+) | 203 (233+) | 12994 (14930+) | $499+ | $599 | ||||||||||||||||||||||||
| GeForce RTX 2080[16] | September 20, 2018 | TU104-400-A1 | 13.6 | 545 | 2944 | 184 | 64 | 46 | 368 | 46 | 1515 | 1710 | 96.96 | 278.76 | 8920 (10068) | 279 (315) | 17840 (20137) | 8 | 60 | 80.5 | 215 | 2-way NVLink | $699 | — | ||||||||
| TU104-400A-A1 | 1710+ | 96.96+ | 278.76+ | 8920 (10068+) | 279 (315+) | 17840 (20137+) | $699+ | $799 | ||||||||||||||||||||||||
| GeForce RTX 2080 Ti[17] | September 27, 2018 | TU102-300-K1-A1 | 18.6 | 754 | 4352 | 272 | 88 | 68 | 544 | 68 | 6 | 1350 | 1545 | 118.8 | 367.2 | 11 | 616 | 352 | 11750 (13448) | 367 (421) | 23500 (26896) | 10 | 78 | 107.6 | 250 | $999 | — | |||||
| TU102-300A-K1-A1 | 1545+ | 118.8+ | 367.2+ | 11750 (13448+) | 367 (421+) | 23500 (26896+) | $999+ | $1,199 | ||||||||||||||||||||||||
| NVIDIA TITAN RTX[18] | December 18, 2018 | TU102-400-A1 | 4608 | 288 | 96 | 72 | 576 | 72 | 1770 | 129.6 | 388.8 | 24 | 672 | 384 | 12442 (16312) | 389 (510) | 24884 (32625) | 11 | 84 | 130 | 280 | $2,499 | ||||||||||
- ^ A Tensor core is a mixed-precision FPU specifically designed for matrix arithmetic.
- ^ The number of Streaming multi-processors on the GPU.
- ^ Pixel fillrate is calculated as the lowest of three numbers: number of ROPs multiplied by the base core clock speed, number of rasterizers multiplied by the number of fragments they can generate per rasterizer multiplied by the base core clock speed, and the number of streaming multiprocessors multiplied by the number of fragments per clock that they can output multiplied by the base clock rate.
- ^ Texture fillrate is calculated as the number of TMUs multiplied by the base core clock speed.
See also
References
- ^ "Introducing NVIDIA GeForce RTX 2070 Graphics Card". NVIDIA. Retrieved 2018-08-20.
- ^ "NVIDIA GeForce RTX 2080 Founders Edition Graphics Card". NVIDIA. Retrieved 2018-08-20.
- ^ "Graphics Reinvented: NVIDIA GeForce RTX 2080 Ti Graphics Card". NVIDIA. Retrieved 2018-08-20.
- ^ "GeForce RTX 2080 launch live blog: Nvidia's Gamescom press conference as it happens". TechRadar. Retrieved 2018-08-21.
- ^ "Nvidia unveils powerful new RTX 2070, RTX 2080, RTX 2080 Ti graphics cards". Polygon. Retrieved 2018-08-20.
- ^ "Nvidia's new RTX 2080, 2080 Ti video cards shipped on Sept 20, starting at $799". Ars Technica. Retrieved 2018-08-20.
- ^ "Nvidia announces RTX 2000 GPU series with '6 times more performance' and ray-tracing". The Verge. Retrieved 2018-08-20.
- ^ "NVIDIA Announces the GeForce RTX 20 Series: RTX 2080 Ti & 2080 on Sept. 20th, RTX 2070 in October". Anandtech. Retrieved 2018-12-06.
- ^ "NVIDIA Deep Learning Super-Sampling (DLSS) Shown To Press". www.legitreviews.com. Retrieved 2018-09-14.
{{cite web}}: Cite has empty unknown parameter:|1=(help) - ^ "The NVIDIA Turing GPU Architecture Deep Dive: Prelude to GeForce RTX". AnandTech.
- ^ "NVIDIA Segregates Turing GPUs; Factory Overclocking Forbidden on the Cheaper Variant". TechPowerUP. Retrieved 2018-12-07.
- ^ "NVIDIA RTX™ platform". Nvidia.
- ^ https://www.youtube.com/watch?v=02h_dNl-C-s
- ^ "Nvidia Geforce RTX 2060". VideoCardz.
- ^ "Introducing NVIDIA GeForce RTX 2070 Graphics Card". NVIDIA.
- ^ "NVIDIA GeForce RTX 2080 Founders Edition Graphics Card". NVIDIA.
- ^ "Graphics Reinvented: NVIDIA GeForce RTX 2080 Ti Graphics Card". NVIDIA.
- ^ "NVIDIA TITAN RTX". NVIDIA. Retrieved 2018-12-18.
Further reading
- "Nvidia Turing GPU Architecture: Graphics Reinvented" (PDF). Nvidia. Nvidia Corporation.