Northern Islands (GPU family)
Encyclopedia
The Northern Islands series is a family of GPUs
Graphics processing unit
A graphics processing unit or GPU is a specialized circuit designed to rapidly manipulate and alter memory in such a way so as to accelerate the building of images in a frame buffer intended for output to a display...

 developed by Advanced Micro Devices
Advanced Micro Devices
Advanced Micro Devices, Inc. or AMD is an American multinational semiconductor company based in Sunnyvale, California, that develops computer processors and related technologies for commercial and consumer markets...

 (AMD) for its Radeon
Radeon
Radeon is a brand of graphics processing units and random access memory produced by Advanced Micro Devices , first launched in 2000 by ATI Technologies, which was acquired by AMD in 2006. Radeon is the successor to the Rage line. There are four different groups, which can be differentiated by...

 line, based on the 40 nm process.

Starting with this family, the former ATI
ATI Technologies
ATI Technologies Inc. was a semiconductor technology corporation based in Markham, Ontario, Canada, that specialized in the development of graphics processing units and chipsets. Founded in 1985 as Array Technologies Inc., the company was listed publicly in 1993 and was acquired by Advanced Micro...

 brand was officially discontinued in favor of making a correlation between the graphics products and the AMD branding for computing platforms (the CPUs and chipsets). Therefore, the AMD brand was used as the replacement. The logo for graphics products and technologies also received a minor makeover (using design elements of the 2010 AMD VISION
AMD Vision
AMD Vision is a marketing term from AMD. The 'AMD Vision' branding segments personal computers and laptops which contain the company's Fusion APUs into four categories, with increasing capability, each aimed at a different target audience....

 logo).

Release

The original design was for 32 nm
32 nanometer
The 32 nm process is the step following the 45 nanometer process in CMOS semiconductor device fabrication. 32 nanometer refers to the average half-pitch of a memory cell at this technology level...

, but TSMC
TSMC
Taiwan Semiconductor Manufacturing Company, Limited or TSMC is the world's largest dedicated independent semiconductor foundry, with its headquarters and main operations located in the Hsinchu Science Park in Hsinchu, Taiwan.-Overview:...

 cancelled its 32 nm process in favor of 28 nm development. AMD was forced to adapt its designs as second generation 40 nm products. Even before canceling 32 nm process, TSMC priced their 32 nm process higher than 40 nm. AMD calculated that for mid-range chips, the new VLIW4 architecture (at 32 nm), would not be as cost effective as an updated VLIW5 on 40 nm.

Due to a major delay caused by the 32 nm cancellation, the top parts, Cayman and Antilles (dual Cayman), were delayed causing the mid-range part (Barts) of the series, as AMD moved the smaller chips (Barts, Turks, and Caicos) to 40 nm early, to be launched first. Also, AMD added stereoscopic 3D support to this Radeon lineup along with HDMI 1.4a and DisplayPort 1.2.

Radeon HD 6400

Codenamed Caicos, this entry-level GPU was released along with Turks on February 7, 2011. The sole Caicos product is the Radeon HD 6450, meant to replace the HD 5450. Compared to the 5450, it has double the stream processors, GDDR5
GDDR5
GDDR5 SDRAM is a type of high performance DRAM graphics card memory designed for computer applications requiring high bandwidth...

 support, along with new Northern Island technologies.

Radeon HD 6600/6500

Codenamed Turks, these entry-level GPUs were released on February 7, 2011. The Turks family includes Turks PRO and Turks XT which are marketed as HD 6570 and HD 6670 respectively. They were originally released to OEMs only, but later released to retail.

The Radeon HD 6570 and 6670 are minor upgrades of their Evergreen counterparts, the HD 5570 and 5670. Turks GPUs contain 80 more stream processors and 4 more texture units. They have also been upgraded to support the new technologies found in the Northern Islands GPUs such as HDMI 1.4a, UVD3, and stereoscopic 3D.

Radeon HD 6700

Codenamed Barts LE, the mid-range GPU Radeon HD 6790 was released on April 5, 2011. There is one retail product available, the Radeon HD 6790. Barts uses shaders of the same 5-way VLIW
Very long instruction word
Very long instruction word or VLIW refers to a CPU architecture designed to take advantage of instruction level parallelism . A processor that executes every instruction one after the other may use processor resources inefficiently, potentially leading to poor performance...

 architecture as HD 5000 series.
  • HD 6790 has 800 stream processors at 840MHz, a 256-bit memory interface and 1 GB GDDR5 DRAM at 1GHz with maximum power draw of 150W. Performance is superior to the NVIDIA GTX 550 Ti and Radeon HD 5770, less powerful than the Radeon HD 6850 and close to the GTX 460 768MB and Radeon HD 5830.


AMD has confirmed that the HD 6700 cards use the Juniper XT and Juniper Pro cores (same cores as the HD 5700 series).
Thus 6770 and 6750 cards are essentially rebranded 5770 and 5750 respectively, with label being the main difference
  • The HD 6750 comes in at 720 stream processors clocked at 700MHz. The memory is packed in the same amount but clocked a tad slower at 1.15GHz (4.6GHz effective), resulting in 73.66GB/s bandwidth, which lines up perfectly with the outgoing HD 5750.

  • In previous HD 6000-series cards, AMD’s Universal Video Decoder was upgraded to version 3.0 which supported BluRay 3D codecs, hardware decoding for DivX / xVid and a list of other improvements. The HD 6700-series however carries over Juniper’s UVD 2.0 so none of these new formats are processed through the GPU.

  • According to AMD, these cards have been upgraded to support HDMI 1.4a but without the 3D features brought forward by UVD 3.0.

Radeon HD 6800

Codenamed Barts, the performance (high-end) GPU Radeon HD 6800 series was released on October 22, 2010. Products include Radeon HD 6850 and Radeon HD 6870. Barts uses shaders of the same 5-way VLIW
Very long instruction word
Very long instruction word or VLIW refers to a CPU architecture designed to take advantage of instruction level parallelism . A processor that executes every instruction one after the other may use processor resources inefficiently, potentially leading to poor performance...

 architecture as HD 5000 series.
  • HD 6850 has 960 stream processors at 775MHz, a 256-bit memory interface and 1 GB GDDR5 DRAM at 1GHz with maximum power draw of 127 W. Compared to competitors of the GTX 460 series, performance falls in line with the 1 GB cards of the NVIDIA GTX 460 series and superior to the 768MB GeForce GTX 460. Compared to predecessor graphics of the Radeon 5800 series, the 6850 is significantly faster than the Radeon HD 5830 and close to the performance of the Radeon HD 5850. A single 6-pin PCIe power connector requirement makes it suitable for most power supplies.

  • HD 6870 has 1120 stream processors at 900MHz, a 256-bit memory interface and 1 GB GDDR5 DRAM at 1.05GHz (4.2GHz effective) with a maximum power draw of 151 W. Performance is superior to the GeForce GTX 460, comparable to the GeForce GTX 560, and less than the GeForce GTX 560 Ti. Compared to predecessor graphics cards of the Radeon 5800 series, the 6870 is faster than the HD 5850 and close to the performance of the Radeon HD 5870.

Radeon HD 6900

This family includes three different high-end products.

Codenamed Cayman, the enthusiast GPU Radeon HD 6900 series was expected to be released on November 12, 2010. These release dates were pushed further back and Cayman was released on December 15, 2010. Products include Radeon HD 6950 and Radeon HD 6970. Cayman is based on new 4-way VLIW architecture, which was chosen over AMD's older VLIW5 in order to reduce complexity in the design of AMD's stream processors
Stream processors
There are multiple articles regarding the term stream processors:*Stream Processing - a technique used to accelerate the processing of many types of video and image computations....

. Studies showed that few applications fully leveraged the extra stage in a VLIW5 SP. Reducing the stream processors to VLIW4 allows AMD to save on transistors for each individual SP and add more overall in the future.
  • In games, the performance of HD 6970 is comparable to the NVIDIA GeForce GTX 570 and GTX 480. The Radeon HD 6950 is slightly slower than the 6970, comparable to slightly faster than the GTX 560 Ti and faster than the HD 5870. The HD 6950 was further discovered to be nearly identical to the 6970 in core design, though the 6950 has lower rated GDDR5 memory. Other than that, the two only differed in BIOS flashed software. As such, a BIOS flash would essentially upgrade the 6950 to a 6970. This was later addressed by AMD and its partners by laser cutting the extra cores (rather than simply disabling them in BIOS), and/or using non-reference card designs that would not work with a 6970 BIOS. Some 6950s can still be "unlocked", but it is much more difficult, requiring careful card selection and custom BIOS.

  • Codenamed Antilles, the Enthusiast dual-GPU (dual-6970) Radeon HD 6990 was launched on March 9, 2011. It features an 830 MHz reference engine clock speed, 3072 stream processors, 5.1 TFLOPS computing performance, 192 texture units, 4 GB of GDDR5 frame buffer (DRAM), and 375 W maximum board power.

  • The AMD Radeon HD 6990 (As with some other 6000 Series AMD Cards) comes with a dual BIOS switch. This enables what some claim to be a hidden 'AMD Uber Mode', however it is used most commonly as a backup when flashing the BIOS. (The same method used to flash the HD 6950 to appear as a HD 6970)

ASIC table

Northern Islands (HD 6xxx) series overview
  • All models manufactured in a 40 nm fabrication process.
  • All models support DirectX
    DirectX
    Microsoft DirectX is a collection of application programming interfaces for handling tasks related to multimedia, especially game programming and video, on Microsoft platforms. Originally, the names of these APIs all began with Direct, such as Direct3D, DirectDraw, DirectMusic, DirectPlay,...

     11, OpenGL
    OpenGL
    OpenGL is a standard specification defining a cross-language, cross-platform API for writing applications that produce 2D and 3D computer graphics. The interface consists of over 250 different function calls which can be used to draw complex three-dimensional scenes from simple primitives. OpenGL...

     4.2 and OpenCL
    OpenCL
    OpenCL is a framework for writing programs that execute across heterogeneous platforms consisting of CPUs, GPUs, and other processors. OpenCL includes a language for writing kernels , plus APIs that are used to define and then control the platforms...

     1.1

Model Year Code name
Code name
A code name or cryptonym is a word or name used clandestinely to refer to another name or word. Code names are often used for military purposes, or in espionage...

Transistors (Million) Die Size (mm2) Bus
Computer bus
In computer architecture, a bus is a subsystem that transfers data between components inside a computer, or between computers.Early computer buses were literally parallel electrical wires with multiple connections, but the term is now used for any physical arrangement that provides the same...

 interface
Memory (MiB
Mebibyte
The mebibyte is a multiple of the unit byte for digital information. The binary prefix mebi means 220, therefore 1 mebibyte is . The unit symbol for the mebibyte is MiB. The unit was established by the International Electrotechnical Commission in 2000 and has been accepted for use by all major...

)
Clock rate Config core1 Fillrate
Fillrate
The term fillrate usually refers to the number of pixels a video card can render and write to video memory in a second. In this case, fillrates are given in megapixels per second or in gigapixels per second , and they are obtained by multiplying the number of raster operations by the clock...

Memory GFLOPS TDP3 (W) Double-precision FP Features / Notes Release Price (USD)
Core (MHz
Hertz
The hertz is the SI unit of frequency defined as the number of cycles per second of a periodic phenomenon. One of its most common uses is the description of the sine wave, particularly those used in radio and audio applications....

)
Memory (MHz
Hertz
The hertz is the SI unit of frequency defined as the number of cycles per second of a periodic phenomenon. One of its most common uses is the description of the sine wave, particularly those used in radio and audio applications....

)
Pixel (GP
Pixel
In digital imaging, a pixel, or pel, is a single point in a raster image, or the smallest addressable screen element in a display device; it is the smallest unit of picture that can be represented or controlled....

/s)
Texture (GT
Texel (graphics)
A texel, or texture element is the fundamental unit of texture space, used in computer graphics. Textures are represented by arrays of texels, just as pictures are represented by arrays of pixels....

/s)
Bandwidth (GB
Gigabyte
The gigabyte is a multiple of the unit byte for digital information storage. The prefix giga means 109 in the International System of Units , therefore 1 gigabyte is...

/s)
Bus type2 Bus width (bit
Bit
A bit is the basic unit of information in computing and telecommunications; it is the amount of information stored by a digital device or other physical system that exists in one of two possible distinct states...

)
Idle Max.
Radeon HD 6450 Feb 7, 2011 (OEM)
Apr 7, 2011 (retail)
Caicos 370 67 PCIe 2.1 x16 512
1024
625-750 533-800
800-900
160:8:4 2.5-3 5-6 8.5-12.8
25.6-28.8
DDR3 GDDR5 64 200-240 9 18
27
$55 (GDDR5)
Radeon HD 6570 Feb 7, 2011 (OEM)
Apr 19, 2011 (retail)
Turks 716 118 PCIe 2.1 x16 1024
2048
650 900
1000
480:24:8 5.2 15.6 28.8
64
DDR3
GDDR5
128 624 10
11
44
60
rowspan=2 $79 (GDDR5)
Radeon HD 6670 512
1024
800 1000 6.4 19.2 64 GDDR5 768 12 66 $99
Radeon HD 6750 Jan 21, 2011 (OEM)
Apr 28, 2011 (retail)
Juniper PRO 1040 170 PCIe 2.1 x16 512
1024
700 1150 720:36:16 11.2 25.2 73.6 GDDR5 128 1008 16 86 rowspan=2
Radeon HD 6770 Juniper XT 850 1200 800:40:16 13.6 34 76.8 1360 18 108
Radeon HD 6790 Apr 4, 2011 Barts LE 1700 255 PCIe 2.1 x16 1024 840 1050 800:40:16 13.4 33.6 134.4 GDDR5 256 1344 19 150 rowspan=3 HD3D UVD3 HDMI 1.4a $149
Radeon HD 6850 Oct 22, 2010 Barts Pro 775 1000 960:48:32 24.8 37.2 128 1488 127 HD3D UVD3 HDMI 1.4a DP 1.2
DisplayPort
DisplayPort is a digital display interface standard produced by the Video Electronics Standards Association . The specification defines a royalty-free digital interconnect for audio and video. The interface is primarily used to connect a video source to a display device such as a computer monitor...

DisplayPort rev.1.2 (DP1.2) provides 21.6Gbps / 17.28Gbps dedicated video bandwidth (VBW) (2x DP1.1 bandwidth -> 10.8Gbps / 8.64Gbps VBW; or 2x VBW as a DL-DVI) that supports higher resolutions up to 4096x2160@50Hz or few lower resolutions with 30-bit color depth, 48Mbps for bitstreaming audio enables unrestricted LPCM (e.g. 8ch LPCM 192khz/24b) along with support for Dolby and DTS lossless audio formats, and 3D stereoscopy (AMD's HD3D). Also, DP1.2 supports Multi-Stream Transport (MST) which allows transmitting several video streams as interleaved packets several through single port.
$179
Radeon HD 6870 Barts XT 900 1050 1120:56:32 28.8 50.4 134.4 2016 151 $239
Radeon HD 6950 Dec 15, 2010 Cayman Pro 2640 389 PCIe 2.1 x16 1024 2048 800 1250 1408:88:32 25.6 70.4 160 GDDR5 256 2253 20 200 HD3D UVD3 HDMI 1.4a DP 1.2
Dual BIOS option
$259
$299
Radeon HD 6970 Cayman XT 2048 880 1375 1536:96:32 28.2 84.5 176 2703 250 $369
Radeon HD 6990 Mar 8, 2011 Antilles 2640 ×2 389 ×2 PCIe 2.1 x16 2048 ×2 830 1250 1536:96:32 ×2 2× 26.5 2× 79.6 2× 160 GDDR5 2× 256 5099 37 375 HD3D UVD3 HDMI 1.4a DP 1.2
Dual BIOS option
$699

  • 1 Unified Shaders
    GPGPU
    General-purpose computing on graphics processing units is the technique of using a GPU, which typically handles computation only for computer graphics, to perform computation in applications traditionally handled by the CPU...

     (Vertex shader/Geometry shader/Pixel shader) : Texture mapping unit
    Texture mapping unit
    A texture mapping unit is a component in modern graphics processing units , historically it is a separate physical processor. A TMU is able to rotate and resize a bitmap to be placed onto an arbitrary plane of a given 3D object as a texture...

     : Render Output unit
    Render Output unit
    The Render Output Unit, often abbreviated as "ROP", and sometimes called Raster Operations Pipeline, is one of the final steps in the rendering process of modern 3D accelerator boards. The pixel pipelines take pixel and texel information and process it, via specific matrix and vector operations,...

  • 2 The effective data transfer rate of GDDR5
    GDDR5
    GDDR5 SDRAM is a type of high performance DRAM graphics card memory designed for computer applications requiring high bandwidth...

    is quadruple its nominal clock, instead of double as 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.

External links

The source of this article is wikipedia, the free encyclopedia.  The text of this article is licensed under the GFDL.
 
x
OK