Table of Contents
Summary of information about AMD Radeon
Radeon is a brand specializing in graphics, RAM and SSD that was founded by ATi in 2000. If only in terms of graphics, Radeon is the successor GPU line to the legendary Rage line that once made a splash with ATi. wind in the market. However, after losing business, ATi was forced to sell itself to AMD in 2006 for 5.4 billion USD.
After the acquisition, AMD restructured the Radeon brand, in August 2010 from ATI Radeon to AMD Radeon. Then, on September 11, 2015, AMD’s graphics segment was separated into Radeon Technologies Group. The current manager is Mr. Raja Koduri.
List of cards belonging to the AMD Radeon series
As the title says, I will only mention the Radeon card lines launched after this brand was acquired by AMD in 2006. And also part of the previous lines that spawned many sub-versions, causing confusion for users. At the same time, I will divide it into 3 periods, corresponding to 3 architectures: TeraScale, GCN and RDNA. And to save time and avoid rambling, I will only mention the launch time along with new features (if any) of each line. Finally, I only mentioned the specifications of commercially sold card lines, not counting OEM card lines.
TeraScale Architecture
Developed by ATI Technologies/AMD, it is the second unified shader microarchitecture model after Xenos. TeraScale replaced the old fixed pipeline microarchitecture and competed directly with Nvidia’s first unified shader microarchitecture called Tesla.
HD 2000 series
Launch date: June 28, 2007
Architecture: Radeon R600
Process: 80nm and 65nm
New feature: CrossFireX
Competition: Geforce 8 series
HD 2000 series technical specifications
GPU | Tên mã | Original pulse | Protocol | VRAM | VRAM type |
HD 2350 | RV610 | 525 MHz | PCI Express 1.0 x16 AGP | 256 MB | DDR2 |
HD 2400 PRO | PCI Express 1.0 x16 AGP PCI | 128 MB 256 MB 512 MB | |||
HD 2400 XT | 650 MHZ | PCI Express 1.0 x16 | 256 MB | DDR2 GDDR3 | |
HD 2600 PRO | RV630 | 600 MHz | PCI Express 1.0 x16 AGP | 256 MB 512 MB | |
HD 2600 XT | 800 MHz | GDDR3 GDDR4 | |||
HD 2900 GT | R600 GT | 601 MHz | PCI Express 1.0 x16 | GDDR3 | |
HD 2900 PRO | R600 PRO | 600 MHz | 512 MB 1 GB | GDDR3 GDDR4 | |
HD 2900 XT | R600 XT | 743 MHz |
HD 3000 series
Launch date: October 2007
Architecture: Radeon R600
Process: 65nm and 55nm
New feature:
Supports DirectX 10.0 and 10.1
Supports PCI Express 2.0 protocol
OpenGL 3.3 support
HD 3000 series technical specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
HD 3410 | RV610 | 519 MHz | PCI Express 1.0 x16 | 256 MB | DDR2 |
HD 3450 | RV620 LE | 600 MHz | PCI Express 2.0 x16 PCI AGP 8x | 256 MB 512 MB | |
HD 3470 | RV620 PRO | 800 MHz | PCI Express 2.0 x16 | DDR2 GDDR3 | |
HD 3550 | 594 MHz | 512 MB | DDR2 | ||
HD 3570 | 796 MHz | ||||
HD 3610 | RV630 PRO | 594 MHz | PCI Express 1.0 x16 | 512 MB 1 GB | |
HD 3650 | RV635 PRO | 725 MHz | PCI Express 2.0 x16 AGP 8x | 256 MB 512 MB 1 GB | DDR2 GDDR3 GDDR4 |
HD 3730 | 722 MHz | PCI Express 2.0 x16 | 512 MB | DDR2 | |
HD 3750 | 796 MHz | GDDR3 | |||
HD 3830 | RV670 PRO | 668 MHz | 256 MB | ||
HD 3850 | PCI Express 2.0 x16 AGP 8x | 256 MB 512 MB 1 GB | GDDR3 GDDR4 | ||
HD 3870 | RV670 XT | 777 MHz | 512 MB 1 GB | ||
HD 3850 X2 | 2x RV670 PRO | 668 MHz | PCI Express 2.0 x16 | 512 MB x2 | GDDR3 |
HD 3870 X2 | R680 | 825 MHz | GDDR3 GDDR4 |
HD 4000 series
Launch date: June 16, 2008
Architecture: TeraScale 1
Process: 55nm and 40nm
New feature:
New memory controller cluster. Supports VRAM GDDR5
SIP block UVD 2.0-2.2
The PLX PEX8647 chip appears on X2 codes
Supports OpenCL 1.1, DirectX 10.1
HD 4000 series technical specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
HD 4350 | RV710 | 600 MHz | PCI Express 2.0 x16 PCI Express 2.0 x1 AGP 8x | 256 MB 512 MB 1 GB | DDR2 DDR3 |
HD 4550 | PCI Express 2.0 x16 | ||||
HD 4570 | 650 MHz | 1 GB | DDR2 | ||
HD 4580 | RV635 PRO | 796 MHz | 512 MB | GDDR3 | |
HD 4650 | RV730 PRO | 600 MHz 650 MHz | PCI Express 2.0 x16 AGP 8x | 256 MB 512 MB 1 GB | DDR2 GDDR3 GDDR4 |
HD 4670 | RV730 XT | 750 MHz | 512 MB 1 GB | ||
HD 4730 | RV770 CE | 700 MHz 750 MHz | PCI Express 2.0 x16 | 512 MB | GDDR5 |
HD 4750 | RV740 | 730 MHz | |||
HD 4770 | 750 MHz | ||||
HD 4810 | RV770 CE | 625 MHz 750 MHz | |||
HD 4830 | RV770 LE | 575 MHz | 512 MB 1 GB | GDDR3 GDDR4 | |
HD 4850 | RV770 PRO | 625 MHz | 512 MB 1 GB 2 GB | GDDR3 GDDR4 GDDR5 | |
HD 4860 | RV790 GT | 700 MHz | 512 MB 1 GB | GDDR5 | |
HD 4870 | RV770 XT | 750 MHz | 512 MB 1 GB 2 GB | ||
HD 4890 | RV790 XT | 850 MHz | 1 GB 2 GB | ||
HD 4850 X2 | 2x RV770 PRO | 625 MHz | 2x 512 MB 2x 1 GB | ||
HD 4870 X2 | 2x RV770 XT | 750 MHz | 2x 1 GB |
HD 5000 series
AMD held a press event at the USS Hornet Museum on September 10, 2009 and announced the ATI Eyefinity multi-display technology and specifications of the Radeon HD 5800 series variants. Demand was overwhelming so much so that more than two months after launch, many online retailers are still having difficulty keeping the 5800 and 5900 in stock.
Launch date: September 10, 2009
Architecture: TeraScale 2
Process: 40nm
New features:
Supports 2 CRT monitors on 1 card
AMD Eyefinity launched
Supports DirectX 11.3, OpenGL 4.5 and OpenCL 1.2
HD 5000 series technical specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
HD 5450 | Cedar PRO | 650 MHz | PCI Express 2.1 x16 PCI Express 2.1 x1 PCI | 512 MB 1 GB 2 GB | DDR2 DDR3 |
HD 5550 | Redwood LE | 550 MHz | PCI Express 2.1 x16 | DDR2 GDDR3 GDDR5 | |
HD 5570 | Redwood PRO | 650 MHz | |||
HD 5610 | 1 GB | GDDR3 | |||
HD 5670 | Redwood XT | 775 MHz | 512 MB 1 GB 2 GB | GDDR3 GDDR5 | |
HD 5750 | Juniper PRO | 700 MHz | 512 MB 1 GB | GDDR5 | |
HD 5770 | Juniper XT | 850 MHz | |||
HD 5830 | Cypress LE | 800 MHz | 1 GB | ||
HD 5850 | Cypress PRO | 725 MHz | 1 GB 2 GB | ||
HD 5870 | Cypress XT | 850 MHz | |||
HD 5870 Eyefinity Edition | 2 GB | ||||
HD 5970 | Hemlock XT | 725 MHz | 2x 1 GB 2x 2 GB |
HD 6000 series
This is the first series AMD completely removed the word ATi from the Radeon brand.
Launch date: October 22, 2010
Architecture: TeraScale 2 and TeraScale 3
Process: 40nm
New features: Mainly about the TeraScale 3 architecture
HD 6000 series technical specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
HD 6350 | Cedar | 650 MHz | PCI Express 2.1 x16 | 512 MB | DDR3 |
HD 6450 | Caicos | 625 MHz 750 MHz | 512 MB 1 GB 2 GB | ||
HD 6570 | Turks PRO | 650 MHz | 2 GB 4 GB | DDR3 GDDR5 | |
HD 6670 | Turks XT | 800 MHz | 512 MB 1 GB 2 GB | ||
HD 6750 | Juniper PRO | 700 MHz | 512 MB 1 GB | GDDR5 | |
HD 6770 | Juniper XT | 850 MHz | |||
HD 6790 | Barts LE | 840 MHz | 1 GB | ||
HD 6850 | Barts PRO | 775 MHz | |||
HD 6870 | Barts XT | 900 MHz | 1 GB 2 GB
| ||
HD 6930 | Cayman CE | 750 MHz | |||
HD 6950 | Cayman PRO | 800 MHz | |||
HD 6970 | Cayman XT | 880 MHz | 2 GB | ||
HD 6990 | Antilles XT | 830 MHz | 2x 2 GB |
Here, we end the era of TeraScale architecture.
GCN architectural period
HD 7000 series
Launch date: January 9, 2012
Architecture: Graphics Core Next (GCN) generation 1. There is 1 code that uses GCN generation 2 (HD 7790)
Process: 28nm
New features:
New GCN architecture
Supports PCI Express 3.0 protocol
Supports Vulkan 1.0 API
Video Coding Engine support
HD 7000 series technical specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
HD 7730 | Cape Verde LE | 800 MHz | PCI Express 3.0 x16 | 1 GB 2 GB | DDR3 GDDR5 |
HD 7750 | Cape Verde PRO | 800 MHz 900 MHz | 1 GB 2 GB 4 GB | ||
HD 7770 GHz Edition | Cape Verde XT | 1000 MHz | 1 GB 2 GB
| GDDR5 | |
HD 7790 | Bonaire XT | ||||
HD 7850 | Pitcairn PRO | 860 MHz | |||
HD 7870 GHz Edition | Pitcairn XT | 1000 MHz | 2 GB | ||
HD 7870 XT | Tahiti LE | 925 MHZ 975 MHz | |||
HD 7950 | Tahiti PRO | 800 MHz | 3 GB | ||
HD 7950 Boost | Tahiti PRO2 | 850 MHz 925 MHz | |||
HD 7970 | Tahiti XT | 925 MHz | 3 GB 6 GB | ||
HD 7970 GHz Edition | Tahiti XT2 | 1000 MHz 1050 MHz | |||
HD 7990 | New Zealand | 950 MHz 1000 MHz | 2x 3 GB |
I will skip the HD 8000-series because that line is just a rename version of the HD 7000 series and is only sold through OEM.
Emerging thanks to “coin”: R7/R9 200 series
At this time, bitcoin began to gain some attention and AMD cards with outstanding mining performance quickly became scarce. This is also the first time gamers have witnessed a stir in the market of coins in general and bitcoin in particular with the price of cards being increased by 164% compared to the suggested retail price and always out of stock around Q3/2013 to Q1/2014.
Launch date: October 8, 2013
Architecture: Graphics Core Next (GCN) generation 2. 1 code uses GCN generation 3 (R9 285)
Process: 28nm
New features:
AMD TrueAudio
Crossfire support with all previous card generations
DirectX 12 support
Technical specifications R7/R9 200 series
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
R7 240 | Oland PRO | 730 MHz | PCI Express 3.0 x16 | 2 GB 4 GB | DDR3 GDDR5 |
R7 250 | Oland XT | 1000 MHz | 1 GB 2 GB | ||
R7 250E | Cape Verde PRO | 800 MHz | GDDR5 | ||
R7 250X | Cape Verde XT | 1000 MHz | |||
R7 260 | Bonaire | 1 GB | |||
R7 260X | Bonaire XTX | 1100 MHz | 1 GB 2 GB | ||
R7 265 | Pitcairn PRO | 900 MHZ | 2 GB | ||
R9 270 | Pitcairn XT | ||||
R9 270X | 1000 MHz | 2 GB 4 GB | |||
R9 280 | Tahiti PRO | 827 MHZ | 3 GB | ||
R9 280X | Tahiti XTL | 850 MHz | |||
R9 285 | Tonga PRO | 918 MHz | 2 GB | ||
R9 290 | Hawaii PRO | 947 MHz | 4 GB | ||
R9 290X | Hawaii XT | 1000 MHz | 4 GB 8 GB | ||
R9 295X2 | Vesuvius | 1018 MHz | 2x 4 GB |
R7/R9 300 series
Launch date: June 16, 2015
Architecture: Graphics Core Next (GCN) generation 1 and 2. Fury series uses GCN generation 3
Process: 28nm
New features:
Frame limiter. Helps avoid wasting rendering performance at too high a number of fps, leading to overheating of the card.
Virtual super resolution support allows images to be rendered at a higher resolution and then downsampled, giving smooth images without the need for anti-aliasing.
Liquid VR support
VRAM HBM first appeared (on R9 Fury X)
R7/R9 300-series specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
R7 350 | Cape Verde XTL | 925 MHz | PCI Express 3.0 x16 | 2 GB | GDDR5 |
R7 360 | Bonaire PRO | 1050 MHz | |||
R7 370 | Pitcairn PRO | 975 MHZ | 2 GB 4 GB
| ||
R9 370X | Pitcairn XT | 1000 MHz | |||
R9 380 | Tonga PRO | 970 MHz | |||
R9 380X | Tonga XT | 4 GB | |||
R9 390 | Hawaii PRO | 1000 MHz | 8 GB | ||
R9 390X | Hawaii XT | 1050 MHz | |||
R9 Fury | Fiji PRO | 1000 MHz | 4 GB | HBM | |
R9 Nano | Fiji XT | ||||
R9 Fury X | 1050 MHz | ||||
Radeon Pro Duo Fiji | 1000 MHz | 2x 4 GB |
Card line drawing up to 162W via PCIe slot: RX 400 series
The Polaris chips on the RX 400-series cards are manufactured on the 14 nm FinFET process, developed by Samsung Electronics and licensed to GlobalFoundries. The Polaris 30 chip is manufactured on the 12 nm FinFET process, developed by Samsung and GlobalFoundries.
Some reviewers have discovered that the AMD Radeon RX 480 violates PCI Express power consumption specifications, which only allow a maximum of 75 watts to be drawn from the motherboard’s PCI Express slot. Tom’s Hardware’s Chris Angelini found that during stress testing, the card could draw an average of 90 watts from the slot and 86 watts while gaming. The highest power draw can be up to 162 watts.
TechPowerUp corroborated these results, and found that the card also draws 166 watts from the 6-pin power supply, which also only supplies 75W.
PC Perspective’s Ryan Shrout also discovered that the card he used for testing drew 80-84 watts from the motherboard and dropped the voltage in other PCIe slots to 11.5 volts on the Asus ROG Rampage V Extreme motherboard.
AMD has released a driver that reprograms the voltage regulation module to draw less power from the motherboard. While this does aggravate the overdraw condition on the 6-pin power supply, it is less of a concern because these connectors have a safety margin well above the on-paper rating of 75W. Of course this also reduces the performance of the card. In many later versions, the power supply was changed to an 8-pin type to ensure there was no short circuit.
Launch date: June 29, 2016
Architecture: Graphics Core Next (GCN) generation 4, codename Polaris
Process: 14nm
New features:
New screen controller
4th generation GCN architecture
Can decode HEVC at 4K 60FPS with 10-bit color channel
Supports DolbyVision and HDR10
RX 400-series specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
RX 460 | Baffin | 1090 MHz | PCI Express 3.0 x16 | 2 GB 4 GB | GDDR5 |
RX 470 | Ellesmere PRO | 926 MHz | 4 GB 8 GB | ||
RX 470D | Ellesmere | 4 GB | |||
RX 480 | Ellesmere XT | 1120 MHz | 4 GB 8 GB |
The next generation of “national” cards: RX 500 series
Fundamentally, the GPUs on the RX 500-series cards are manufactured on the same process and architecture as the RX 400 cards, but thanks to improvements in the manufacturing process, the new chips can push clock speeds up higher with the same power source.
However, that is not the reason why this card line became “national”, the reason is somewhat familiar “coin” and AMD’s slow response. Specifically, at the time of launch, AMD’s cards had terrible mining performance, but now there are specialized bitcoin miners on the market and the performance of mining speed/power consumption of they go far beyond graphics cards.
But that doesn’t make the coin less hot, the market witnessed the birth of a series of alt coins that miners call trash coins. People will rush to mine new coins, convert them to bitcoin and find another newly released junk coin to mine. Many Vietnamese “miners” hit the jackpot when they bought houses and cars, creating a local fever.
Because the card fever is only local in Vietnam and China, AMD responded very superficially and was slow in increasing production and launching specialized cards for mining. And so when they launched a series of new mining cards, the bitcoin bubble collapsed and “strong” cards were sold all over the street. This means that the price of new cards must also be reduced. Around 2018-2019 is a paradise for gamers when both old and new cards have dirt cheap prices.
Launch date: April 18, 2017
Architecture: Graphics Core Next (GCN) generation 4
Process: 14nm and 12nm
New features:
Supports DisplayPort 1.4 HBR, HDMI 2.0b output standards
Supports HDR 10 color standard
RX 500-series specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
RX 550 | Lexa | 1100 MHz | PCI Express 3.0 x8 | 2 GB 4 GB | GDDR5 |
RX 550X | |||||
RX 560 | Baffin | 1175 MHz | |||
RX 570 | Polaris 20 XL | 1168 MHz | PCI Express 3.0 x16 | 4 GB 8 GB | |
RX 580 | Polaris 20 XT | 1257 MHz | |||
RX 590 | Polaris 30 XT | 1469 MHz | 8 GB |
Come and leave silently: RX Vega series
Yes, these are the same GPUs integrated into many of AMD’s APUs on the market around 2019-2021. However, you may not know, there is a line of discrete cards called RX Vega that includes 4 card models, but no one cares because the RX 500-series is too cheap and the RX Vega is expensive + hot + consumes a lot more power. No one was interested in buying an RX Vega and this card line quietly withdrew from the market.
Launch date: August 14, 2017
Architecture: Graphics Core Next (GCN) generation 5
Process: 14nm and 7nm
New features:
New VRAM standard: HBM2
New generation CU
DirectX 12.1 support
RX Vega series specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
RX Vega 56 | Vega 10 XL | 1156 MHz | PCI Express 3.0 x16 | 8 GB | HBM2 |
RX Vega 64 | Vega 10 XT | 1247 MHz | |||
RX Vega 64 Liquid | 1406 MHz | ||||
Radeon VII | Vega 20 | 1400 MHz | 16 GB |
This is the end of the era of GCN architecture.
RDNA architecture period
The world’s first 7nm GPU card line: RX 5000 series
After many years of being behind rival Nvidia, AMD finally rose up when it announced the world’s first line of cards with 7nm GPU – RX 5000. The prefix “50” is speculated to mean AMD celebrates its 50th anniversary in 2019. And with the miniaturization of transistors, AMD’s cards are no longer hot and power hungry – two things often associated with AMD’s name throughout its development history.
Launch date: July 7, 2019
Architecture: Radeon DNA (RDNA) generation 1
Process: TSMC 7nm FinFET
New features:
New VRAM standard: GDDR6
New generation CU, helps improve IPC compared to the old generation
New architecture: RDNA.
Supports PCI Express 4.0 x16 protocol
RX 5000 series specifications
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
Radeon RX 5300 | Navi 14 XL | 1327 MHz | PCI Express 4.0 x8 | 3GB | GDDR6 |
Radeon RX 5300 XT | Navi 14 XL | 1670 MHz | 4GB | GDDR5 | |
RX 5500 | Navi 14 XT | 4GB | GDDR6 | ||
RX 5500 XT | Navi 14 XTX | 1717 MHz | 8GB | ||
RX 5600 | Navi 10 XE | 1375 MHz | PCI Express 4.0 x16 | 6GB | |
RX 5600 XT | Navi 10 XLE | ||||
RX 5700 | Navi 10 XL | 1465 MHz | 8GB | ||
RX 5700 XT | Navi 10 XT | 1605 MHz | |||
RX 5700 XT 50th Anniversary | Navi 10 XTX | 1680 MHz |
Outpace the competition: RX 6000 series
Under the leadership of Dr. Lisa Su, AMD has done the unthinkable: launched a line of cards that consume less power than rival Nvidia’s RTX 3000 series of cards and led the way in technology that Nvidia had to “copy”. Not to mention, the RX 6000’s computing performance is generally better. Unfortunately, Nvidia is still strong thanks to two factors: Ray Tracing and DLSS. However, this gap can be bridged through software and driver development, which AMD is still completing every day.
Launch date: October 28, 2020
Architecture: RDNA 2
Process: 7nm
New features:
Performance doubled compared to the previous generation
Rage Mode: Automatically OC with just 1 click
Infinity Cache, SAM
Supports API DX12 Ultimate, Ray Tracing,
Supports Super Resolution technology (similar to Nvidia’s DLSS)
Pulse boost to over 2000 MHz.
GPU | Codename | Original pulse | Protocol | VRAM | VRAM type |
RX 6700 XT | Navi 22 | 2321/2581 MHz | PCI Express 4.0 x16 | 12GB | GDDR6 |
RX 6800 | Navi 21 | 1815/2105 MHz | 16GB | ||
RX 6800 XT | 2015/2250 MHz | ||||
RX 6900 XT |
Epilogue
So I have gone through the history of the AMD Radeon graphics card brand. Hopefully the above article has given you a look at the development history of one of the two largest graphics card manufacturers in the world and how AMD graphics cards have developed over the past 12 years. Most importantly, the thing that AMD cards have always been accused of being hot and consuming electricity has disappeared since the RX 6000 generation, please take note.