List of AMD graphics processing units

The following is a list that contains general information about GPUs and video cards made by AMD, including those made by ATI Technologies before 2006, based on official specifications in table-form.

The headers in the table listed below describe the following:

• Model – The marketing name for the GPU assigned by AMD/ATI. Note that ATI trademarks have been replaced by AMD trademarks starting with the Radeon HD 6000 series for desktop and AMD FirePro series for professional graphics.
• Codename – The internal engineering codename for the GPU.
• Launch – Date of release for the GPU.
• Architecture – The microarchitecture used by the GPU.
• Fab – Fabrication process. Average feature size of components of the GPU.
• Transistors – Number of transistors on the die.
• Die size – Physical surface area of the die.
• Core config – The layout of the graphics pipeline, in terms of functional units.
• Core clock – The reference base and boost (if available) core clock frequency.
• Fillrate
  • Pixel - The rate at which pixels can be rendered by the raster operators to a display. Measured in pixels/s.
  • Texture - The rate at which textures can be mapped by the texture mapping units onto a polygon mesh. Measured in texels/s.
• Performance
  • Shader operations - How many operations the pixel shaders (or unified shaders in Direct3D 10 and newer GPUs) can perform. Measured in operations/s.
  • Vertex operations - The amount of geometry operations that can be processed on the vertex shaders in one second (only applies to Direct3D 9.0c and older GPUs). Measured in vertices/s.
• Memory
  • Bus type – Type of memory bus utilized.
  • Bus width – Maximum bit width of the memory bus utilized.
  • Size – Size of the graphics memory.
  • Clock – The reference memory clock frequency.
  • Bandwidth – Maximum theoretical memory bandwidth based on bus type and width.
• TDP (Thermal design power) – Maximum amount of heat generated by the GPU chip, measured in Watt.
• TBP (Typical board power) – Typical power drawn by the total board, including power for the GPU chip and peripheral equipment, such as Voltage regulator module, memory, fans, etc., measured in Watt.
• Bus interface – Bus by which the graphics processor is attached to the system (typically an expansion slot, such as PCI, AGP, or PCIe).
• API support – Rendering and computing APIs supported by the GPU and driver.

Due to conventions changing over time, some numerical definitions such as core config, core clock, performance and memory should not be compared one-to-one across generations. The following tables are for reference use only, and do not reflect actual performance.

• R100 – Video Immersion
• R200 – Video Immersion II
• R300 – Video Immersion II + Video Shader
• R410 – Video Shader HD
• R420 – Video Shader HD + DXVA
• R520 – Avivo Video
• R600 – Avivo HD – UVD 1.0
• R700 – UVD 2, UVD 2.2
• Evergreen – UVD 2.2
• Northern Islands – UVD 3 (HD 67xx UVD 2.2)
• Southern Islands – UVD 3.1, VCE 1.0
• Sea Islands – UVD 4.2, VCE 2.0
• Volcanic Islands – UVD 5.0, 6.0, VCE 3.0
• Arctic Islands – UVD 6.3, VCE 3.4
• Vega – UVD 7.0, VCE 4.0 and VCN 1.0 only at AMD Raven Ridge
• Navi 1X – VCN 2.0
• Navi 2X – VCN 3.0
• Navi 3X – VCN 3.0

The following table shows features of AMD/ATI's GPUs.

The following table shows the graphics and compute APIs support across ATI/AMD GPU microarchitectures. Note that a branding series might include older generation chips.

¹ Pixel pipelines : Vertex shaders : Texture mapping units : Render output units
² OpenGL 1.0 (Generic 2D) is provided through software implementations.

• All models include Direct3D 7.0 and OpenGL 1.3
• The R100 cards were originally launched without any numbering as Radeon SDR, DDR, LE and VE; these products were later "rebranded" to their names within the numbered naming scheme, when the Radeon 8000 series was introduced.

¹ Pixel pipelines : Vertex shaders : Texture mapping units : Render output units

^A First number indicates cards with 32 MB of memory. Second number indicates cards with 64 MB of memory.
^B First number indicates OEM cards. Second number indicates Retail cards.

• All models are manufactured with a 180 nm fabrication process
• All models include Direct3D 7.0 and OpenGL 1.3
• Based on the Radeon VE

¹ Pixel pipelines : Vertex shaders : Texture mapping units : Render output units

• All models are manufactured with a 150 nm fabrication process
• All models include Direct3D 8.1 and OpenGL 1.4

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units

• All models are manufactured with a 150 nm fabrication process
• All models include Direct3D 8.1 and OpenGL 1.4
• Based on the Radeon 9200

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units

• All models include Direct3D 9.0 and OpenGL 2.0
• All models use an AGP 8x interface

¹ Pixel shaders : Vertex Shaders : Texture mapping units : Render output units
² The 256-bit version of the 9800 SE when unlocked to 8-pixel pipelines with third party driver modifications should function close to a full 9800 Pro.

• All models include Direct3D 9.0 and OpenGL 2.0
• All models use a PCIe ×16 interface

¹ Pixel shaders : Vertex Shaders : Texture mapping units : Render output units

• All models include Direct3D 9.0 and OpenGL 2.0
• Based on the Radeon X300

¹Pixel shaders : Vertex Shaders : Texture mapping units : Render output units

• All models include AGP 8×
• All models include Direct3D 9.0b and OpenGL 2.0

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units

• All models include PCIe ×16
• All models include Direct3D 9.0b and OpenGL 2.0

¹ Pixel shaders : Vertex Shaders : Texture mapping units : Render output units

• All models include Direct3D 9.0b and OpenGL 2.0
• Based on Radeon X700

Note that ATI X1000 series cards (e.g. X1900) do not have Vertex Texture Fetch, hence they do not fully comply with the VS 3.0 model. Instead, they offer a feature called "Render to Vertex Buffer (R2VB)" that provides functionality that is an alternative Vertex Texture Fetch.

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units

• All Radeon HD 3000 IGP models include Direct3D 10.0 and OpenGL 3.3

¹ Unified shaders : Texture mapping units : Render output units
² The clock frequencies may vary in different usage scenarios, as AMD PowerPlay technology is implemented. The clock frequencies listed here refer to the officially announced clock specifications.
³ The sideport is a dedicated memory bus. It is preferably used for a frame buffer.

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units
² Unified shaders : Texture mapping units : Render output units

¹ Unified shaders : Texture mapping units : Render output units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.
³ The TDP is reference design TDP values from AMD. Different non-reference board designs from vendors may lead to slight variations in actual TDP.
⁴ All models feature UVD2 and PowerPlay.

• All Radeon HD 4000 IGP models include Direct3D 10.1 and OpenGL 2.0

¹ Unified shaders : Texture mapping units : Render output units
² The clock frequencies may vary in different usage scenarios, as ATI PowerPlay technology is implemented. The clock frequencies listed here refer to the officially announced clock specifications.
³ The sideport is a dedicated memory bus. It preferably used for frame buffer.

• The HD5000 series is the last series of AMD GPUs which supports two analog CRT-monitors with a single graphics card (i.e. with two RAM-DACs).
• AMD Eyefinity introduced.

• The Radeon HD 6000 series has a new tesselation engine which is said to double the performance when working with tesselation compared to the previous HD 5000 series.

• All models feature the UNB/MC Bus interface
• All models lack double-precision FP
• With driver Update OpenGL 4.4 available (Last Catalyst 15.12). OpenGL 4.5 available with Crimson Beta (driver version 15.30 or higher).
• All models feature Angle independent anisotropic filtering, UVD3, and AMD Eyefinity capabilities, with up to three outputs.
• All models feature 3D Blu-ray Disc acceleration.
• Embedded GPUs as part of AMD's Lynx platform APUs.

• All models feature the UNB/MC Bus interface
• All models do not support double-precision FP
• TeraScale 2 (VLIW5) based APUs feature angle independent anisotropic filtering, UVD3, and Eyefinity capabilities, with up to three outputs.
• TeraScale 3 (VLIW4) based APUs feature angle independent anisotropic filtering, UVD3.2, and Eyefinity capabilities, with up to four outputs.

These GPUs are either integrated into the mainboard or occupy a Mobile PCI Express Module (MXM).

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units.

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units.

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units.

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units.

OpenGL 3.3 is possible with latest drivers for all RV6xx.

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units.
² Unified Shaderss : Texture mapping units : Render output units

¹ Unified Shaders : Texture mapping units : Render output units

¹ Unified shaders : Texture mapping units : Render output units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.

¹ Unified shaders : Texture mapping units : Render output units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.

• All models feature the UNB/MC Bus interface
• All models lack double-precision FP
• All models feature Angle independent anisotropic filtering, UVD3, and Eyefinity capabilities, with up to three outputs.

• All models include Direct3D 11, OpenGL 4.4 and OpenCL 1.2
• All models feature the UNB/MC Bus interface
• All models lack double-precision FP
• All models feature angle independent anisotropic filtering, UVD3 and Eyefinity capabilities, with up to three outputs.
• All models feature VLIW5

¹ Unified shaders : Texture mapping units : Render output units : Compute units
² TDP specified for AMD reference designs, includes CPU power consumption. Actual TDP of retail products may vary.

¹ Vertex shaders : Pixel shaders : Texture mapping units : Render output units
² Unified shaders : Texture mapping units : Render output units : Compute Units

¹ Vertex shaders : Pixel shaders : Texture mapping unit : Render output units
² Unified shaders : Texture mapping unit : Render output units

¹ Unified shaders : Texture mapping units : Render output units : Compute Units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory
³ Windows 7, 8.1, 10 Support for Fire Pro Cards with Terascale 2 and later by firepro driver 15.301.2601

¹ Unified shaders : Texture mapping units : Render output units : Compute Units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.
³ Support for Windows 7, 8.1 for OpenGL 4.4 and OpenCL 2.0, when Hardware is prepared with firepro driver 14.502.1045

• Vulkan 1.0 and OpenGL 4.5 possible for GCN with Driver Update FirePro equal to Radeon Crimson 16.3 or higher.
• Vulkan 1.1 possible for GCN with Radeon Pro Software 18.Q1.1 or higher. It might not fully apply to GCN 1.0 or 1.1 GPUs.

¹ Unified shaders : Texture mapping units : Render output units : Compute Units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.
³ OpenGL 4.4: support with AMD FirePro driver release 14.301.000 or later, in footnotes of specs

In 2014, AMD released the D-Series specifically for Mac Pro workstations.

¹ Unified shaders : Texture mapping units : Render output units : compute units

• Vulkan 1.0 and OpenGL 4.5 possible for GCN with Driver Update FirePro equal to Radeon Crimson 16.3 or higher. OpenCL 2.1 and 2.2 possible for all OpenCL 2.0-Cards with Driver Update in Future (Khronos). Linux Support for OpenCL is limited with AMDGPU Driver 16.60 actual to Version 1.2.
• Vulkan 1.1 possible for GCN with Radeon Pro Software 18.Q1.1 or higher. It might not fully apply to GCN 1.0 or 1.1 GPUs.

¹ Unified shaders : Texture mapping units : Render output units : compute units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.
³ OpenGL 4.4: support with AMD FirePro driver release 14.301.000 or later, in footnotes of specs

• Vulkan 1.1 possible for GCN with Radeon Pro Software 18.Q1.1 or higher. It might not fully apply to GCN 1.0 or 1.1 GPUs.

• Vulkan 1.1 possible for GCN with Radeon Pro software 18.Q1.1 or higher.

• Half precision power (FP16) is equal to single precision power (FP32) in 4th GCN generation (in 5th Gen: half precision (FP16) = 2× SP (FP32))

¹ Unified shaders : Texture mapping units : Render output units : compute units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.

• Vulkan 1.0 and OpenGL 4.5 possible for GCN with Driver Update FirePro equal to Radeon Crimson 16.3 or higher. OpenGL 4.5 was only in Windows available. Actual Linux Driver support OpenGL 4.5 and Vulkan 1.0, but only OpenCL 1.2 by AMDGPU Driver 16.60.
• Vulkan 1.1 possible for GCN with Radeon Pro Software 18.Q1.1 or higher. It might not fully apply to GCN 1.0 or 1.1 GPUs.

¹ Unified shaders : Texture mapping units : Render output units: Compute units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.
³ OpenGL 4.4: support with AMD FirePro driver release 14.301.000 or later, in footnotes of specs

¹ Unified shaders : Texture mapping units : Render output units : compute units
² The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.

¹ Unified shaders : Texture mapping units : Render output units
² CU = Compute units
³ The effective data transfer rate of GDDR5 is quadruple its nominal clock, instead of double as it is with other DDR memory.

¹ Pixel shaders : Vertex shaders : Texture mapping units : Render output units
² Unified shaders : Texture mapping units : Render output units
³ Unified shaders : Texture mapping units : Render output units : RT Cores
⁴ The Latte looks similar to the RV730 used in the Radeon HD4650/4670.
⁵ In most cases, especially in games, half of this data can be considered.

• List of Nvidia graphics processing units
• List of Intel graphics processing units
• List of AMD processors with 3D graphics
• Apple M1
• Video Coding Engine, AMD's equivalent SIP core till 2017
• Video Core Next, AMD's video core which combines the functionality of Video Coding Engine and Unified Video Decoder

• TechPowerUp! GPU Database