AMD just announced yesterday350$100 million in all stock acquisition of Xilings，Today (1:00 a.m. on October 29) is officiallyReleased the new generation of Radeon RX 6000 series graphics cards，The first batch includes RX 6800、RX 6800 XT、RX 6900 XT three models，All based on RDNA 2 architecture、Navi 21 Core，268Billions of transistors。(Each model specific、Detailed process，Please see the middle of this article。）
David Wang, Senior Vice President of Engineering Research and Development, AMD Radeon Technology Division, said，As a successor to the RX 5000 series RDNA architecture，The mission goal of the RDNA 2 architecture is to improve performance by 1x、Energy efficiency improvement by at least 50%，And fully support DX12 Ultimate and other technological features，And it all needs to be completed within 16 months，Extremely challenging。finally，The development team not only achieves the predetermined goals，Even overcomplete，Actual performance almost exceeded everyone's expectations。
The first thing to emphasize is，RDNA 2 GPU architecture and Zen 3 The CPU architecture is exactly the same，No new manufacturing technology，Still the original TSMC 7nm，It's just that the process itself is more mature，No EUV lithography，All the performance of this actual product、Energy efficiency improvement，All comes from the changes in the architecture itself，Not easy。
AMD says，The RDNA 2 architecture has four highlights，They are high-performance computing units (CUs)、The revolutionary Infinity Cace、Breakthrough high-speed design、Advanced technology features。
Computing units have always been the basic module of AMD GPU architecture，RDNA architecture has undergone drastic reforms，RDNA 2 further adjustment，Including ubiquitous fine clock gate control、Radical assembly line rebalancing、Redesigned data paths，In the end, it brought about 30% energy efficiency improvement。
Infinity Cache is indeed revolutionary，It adds up to 128MB of independent cache in Navi 21 core，Compared with the previous few hundred KB caches, it has made a qualitative leap.，Able to have higher bandwidth、Lower power consumption，Quickly and promptly process data in the GPU。
According to AMD，Infinity Cache plus 256KB G6 cache，Compared with the previous 384KB G6 cache，The bandwidth can be increased by up to 117%，Power consumption can be reduced by 10%。and，This cache design does not require special optimization by game developers to support and exert strength.，Any game can benefit directly，Of course, developers can continue to dig deeper、promote。
High-speed design mainly involves frequency，Under the premise that the process remains unchanged，The maximum frequency of RDNA 2 has increased by 30%，Generally more than 2GHz，The maximum acceleration even exceeds 2.2GHz。by contrast，The acceleration frequency of the RTX 20/30 series is also only in the range of 1.7-1.8GHz，The difference is more than 400MHz。
Comprehensive，The energy efficiency of RDNA 2 (efficiency per watt) is 54% higher than that of RDNA，The performance improvements of the same frequency brought by Infinity Cache respectively、Higher operating frequency、Energy efficiency improvements brought by overall design optimization。
at the same time，RDNA 2 vs. RX 5700 XT also achieved about 2 times performance improvement，Especially with excellent gaming performance at 4K ultra-clear resolution，Games such as Resident Evil 3 have more than 2 times the improvement settings are，Up to 2.2 times!
In addition to strong performance、Energy efficiency，The RDNA 2 architecture has finally been completed in terms of technological characteristics，DX12 Ultimate、DXR Ray Tracking、VRS variable shading rate、Mesh Shaders、Samper Feedbacks are available on competitors，All supported。
certainly，It is absolutely not enough to just follow up，The RDNA 2 architecture also has a large number of its own unique skills，For example, AMD FiedlityFX，It's a collection of game graphics technology，Including contrast adaptive sharpening、noise reduction、Variable coloring、Ambient light occlusion、Screen space reflection、Super resolution, etc.。
in，Super Resolution TechnologyIt's still an experimental feature，New cards will be temporarily missing when they are first released，After the optimization is in place, support will be added through driver updates.。Currently, 35 games support AMD FiedlityFX，Including many recent popular masterpieces，For example, Dust 5、"The Fall of God"、World of Warcraft：Shadow Country、"The Milky Way"、"Far Cry 6"，They will release videos in the near future，Showcase optimization support for AMD Technology。
The RNDA 2 architecture also supports new"Rage Mode" (Rage Mode)，One-click automatic overclocking can be achieved，Release maximum efficiency，And this function does not affect the warranty，Don't worry about it being super bad，Also provides quiet mode、Equalization mode，Suitable for different application scenarios，Maintain the most suitable energy efficiency at any time。
have to be aware of is，Rage Mode switch、Specific settings，Depend on the actual settings of different AIB brands for different models。
As the only current high-performance CPU、GPU、Chipset manufacturer，A unique advantage of AMD is platform optimization。RNDA 2 architecture not only continues to support PCIe 4.0，Can be matched with Ryzen 3000/5000 series processors、400/500Series motherboards perform best，It also supports "Smart Access Memory" technology。
On previous platforms，The processor can only access the graphics card up to 256MB of video memory，Now the RX 6000 series display card is paired with Ryzen 5000 series processor，The latter can access all video memory，And it doesn't require separate optimization of the game。
AMD claims，Furious Mode、With the joint blessing of smart access video memory，4K games can bring an additional performance improvement of up to 13%，This is AMD's unique advantage，No matter how you match N cards, you won't get such free bonuses。
Proposed on the RDNA architectureRadeon Boost Accelerate Technology、Radeon Anti-Lag Anti-LagPromote and develop，Combined with FreeSync synchronous repainting technology，Can greatly reduce the keyboard、Mouse operation，Execute to the graphics card，Delay to monitor output。
Take "Fortnite" at 4K resolution as an example，The above technology can reduce the delay from 63ms to 40ms，Save 37%，The result is that the game is more responsive、More smooth operation experience。
AMD also mentioned its own advantages，That is to adopt standard design，Better versatility，Including the host shell、powered by、Compatibility of heat dissipation。
The power supply is obviously a 12-pin power supply dielectric specially designed for the RTX 30 series.，Or with adapter cable，Or redesigned power supply，No non-public version is used。
Next, let’s talk about the product process in detail
Radeon RX 6800 XT。72Group computing unit，4608stream processors，Core game frequency 2015MHz、Acceleration frequency 2250MHz，That is, it can run at almost 2GHz or above，Integrated 128MB Infinity Cache，20.74 GPU performance of teraflops，With 256-bit 16GB GDDR6 graphics memory，Equivalent frequency 16GHz，Bandwidth 512GB/s，The power consumption of the whole card is 300W，Dual 8-pin auxiliary power supply。
RX 6800 Compared with the competitor NVIDIA RTX, which consumes 20W more power 3080，It's in 4K、2The game performance at K resolution is basically the same，Many games can still lead slightly。
Even if the difference between official and actual data is taken into account、Different optimizations for different games，RX 6800 XT's comprehensive strength can also be basically in the same level as RTX 3080，And don't forget that A-cards have always been fighting the future，Driver updates will continuously improve performance、New features (such as super-resolution)。
The violent mode I just mentioned、Smart access to video memory，It's even more icing on the cake，Different games can bring up to 13% additional performance improvements，In this way, it will be no big problem to stabilize the RTX 3080。
Radeon RX 6800。60Group computing unit，3840stream processors，Core game frequency 1815MHz、Maximum acceleration frequency 2105MHz，Infinity Cache remains complete with 128MB，16.17 GPU performance of teraflops，Video memory or 256-bit 16GB GDDR6，The power consumption of the entire card has been reduced to 250W。
The competitor of the RX 6800 is NVIDIA RTX 2080 Of，Of course, it is better to say that it is an RTX with almost completely equal performance than its 3070，With the blessing of smart access to video memory，Leading in performance，According to official data, the highest rate reached an astonishing 34%.，It is estimated that the overall strength is 10% ahead of the situation.。
AMD also directly attacked Nvidia's competitors on the spot，Said Radeon RX 6800 has passed 4K and 1440p tests，When the number of games is limited，It can beat or fight against Nvidia's old graphics cards。
Radeon RX 6900 XT。Flagship-class Radeon RX 6900 XT It has a complete 80 sets of computing units、5120stream processors，Core game frequency 2015MHz、Acceleration frequency up to 2250MHz，23.04 GPU performance of teraflops，Same 128MB Infinity Cache、256-bit 16GB GDDR6，And the power consumption of the entire card is actually as good as RX 6800 XT is also controlled at 300W。
so that，RX 6900 XT's energy efficiency is more prominent，Energy efficiency improvement reached a terrifying 65%。The game performance has reached a level basically the same as the competitor's flagship RTX 3090，Some games are still ahead，The power consumption is 50W lower!
Although Nvidia targets RTX 3090 in 8K gaming，But its 24GB large memory makes it more likely to be used by the creative industry as a Titan alternative。AMD has not chosen to add new memory on the same model，This RX 6900 XT maintains 16GB of GDDR6 memory，And offer a competitive price。
All three new Radeon RX 6000 series cards also feature a traditional 2.5-slot design，Three-fan system with conventional PCIe connection and cooling card。But AMD didn't do any fancy work like Nvidia to narrow down its PCB，No new 12-pin power connector is launched。
AMD uses 128MB cache memory，This is based on the company's Zen 3 L3 CPU cache design。It is optimized for graphical use，And promises to provide twice the bandwidth。All this means，These latest Radeon cards can transfer data to the graphics pipeline more efficiently，and improve performance without significantly increasing power consumption。AMD's goal is to bring performance from previous RX 5700 XT card doubles，However, the power consumption has increased slightly。
Although AMD promises to go hand in hand with Nvidia in 4K gaming and more，But one of the biggest regrets of this competition is the lack of products comparable to Nvidia's DLSS。Nvidia's AI-powered supersampling technology has transformed games that support the technology，Simply switch game settings to bring excellent picture quality and higher frame rates。
AMD and Nvidia、Intel's rivalry has never stopped，But a few years ago，AMD seems to have lost the battle in this video card war。but，With the latest release of Radeon RX 6000 series，AMD's performance and price have launched considerable challenges to Nvidia's RTX 30 series cards。
In fact, Nvidia released the RTX30 series graphics card last month has caused a huge sensation in the industry，Not only doubling its effectiveness，The price is only half of the previous Titan。
On this basis，AMD claims its new RDNA 2 technology offers twice as much power as the previous generation，But the power consumption is roughly the same (and lower than similar products of Nvidia)。AMD's RX 6900 XT and RTX3090 have similar performance，The price is much lower than the official public version of 3090，This price is of course quite competitive。
The next stop for AMD graphics cards will be the RDNA 3 architecture，Currently under design，Advanced node will be adopted，As for whether it is 7nm EVU or 5nm，Maybe it's a killer，The release time is expected to be at the beginning of the year after。
in addition，AMD is also pushing for high-performance computing、CDNA architecture for deep learning，Will be used for Radeon Instinct Computing Card Product Line。
You may also be interested in the following：