Amd radeon

The first version of RDNA was a huge step up over AMD’s previous GCN architecture. AMD aimed for the 7nm RDNA 2 to trounce again the performance-per-watt of RDNA, and it has pulled it off with an actual 54% increase. If that wasn’t impressive enough, the mighty RX 6900 XT boosts the performance-per-watt increase to a whopping 65%.

AMD’s RX 7800 XT doesn’t radically outperform its predecessor, but it does come in $150 lower at launch. At current prices, however, it’s effectively a lateral move with some architectural and feature upgrades to sweeten the deal.

TODAY’S BEST DEALS

• + Good 1440p and 1080p performance
• + 16GB is plenty of VRAM
• + Strong rasterization performance
• + DP2.1, AV1, and other architectural improvements

What’s New

New Game Support

• Frostpunk 2
• God of War Ragnarök
• Warhammer 40,000: Space Marine 2
• The Sims 4 DirectX 11 Update

AMD Fluid Motion Frames (AFMF) 2

• A major advancement in frame generation technology for AMD HYPR-RX.
• Lower Latency and Higher Performance
• AFMF 2 enhances fast-paced gaming by significantly reducing frame generation latency and improving performance scaling through new modes.

Fast Motion Optimization

• Enjoy smoother gameplay and higher FPS with improved frame generation consistency during fast motion.

Improved Borderless-Fullscreen Support

• Expanded display mode support for RDNA 3 series graphics products ensures compatibility with virtually all borderless fullscreen games.

Expanded API Support

• AFMF 2 can now be enabled for any OpenGLNEW, VulkanNEW, DirectX 11, and DirectX 12 titles.

Radeon Chill Interoperability

• AFMF 2 now supports Radeon Chill, providing a low-latency FPS capping option.

Optimized AMD Ryzen AI 300 Series Support

• AFMF 2 is optimized for an extensive list of AMD products, including AMD Ryzen AI 300 series processors.

Radeon Anti-Lag 2

• Vulkan Support for Counter-Strike 2
• Added support for Vulkan API, offering additional responsive gaming options.

Geometric Downscaling for Video

• Improved image quality by reducing artifacts during downscaled video playback.
• Geometric Downscaling is supported on AMD Radeon 800M integrated graphics, as well as AMD Radeon RX 7000 series desktop and mobile discrete graphics cards.

Expanded AMD Radeon Boost Support

• FINAL FANTASY XVI

Expanded HYPR-Tune Support

• HYPR-Tune support allows HYPR-RX to enable in-game technologies like AMD FidelityFX Super Resolution and AMD Radeon Anti-Lag 2.

Support has been added to automatically configure AMD FidelityFX Super Resolution 3 with frame generation in:

• Black Myth: Wukong
• Creatures of Ava
• Frostpunk 2
• God of War Ragnarök

Support has been added to automatically configure AMD Radeon Anti-Lag 2 in:

• Ghost of Tsushima DIRECTOR’S CUT

Expanded Vulkan Extensions Support

• VK_AMD_anti_lag
• VK_KHR_maintenance7
• VK_KHR_pipeline_binary
• VK_EXT_shader_object

Mesh Nodes in Work Graphs

• Microsoft DirectX 12 Work Graphs with Mesh Nodes support for AMD Radeon RX 7000 Series graphics cards.

Fixed Issues and Improvements

• Intermittent driver timeout or application crash while playing Warhammer 40,000: Space Marine 2.
• Intermittent driver timeout or application crash during certain cutscenes while playing FINAL FANTASY XVI on some AMD Graphics Products, such as the Radeon RX 6600 XT.
• Overly dark shadows or desaturated colors may be observed while playing Black Myth: Wukong when Global Illumination is to Medium or higher.
• Intermittent in-game corruption may be observed while playing Ghost of Tsushima DIRECTOR’S CUT with AMD Software: Adrenalin Edition Record & Streaming and HDR enabled.
• AFMF may become inactive after enabling certain on-screen overlays.
• AMD Software: Adrenalin Edition may unexpectedly initiate upon system wake from sleep mode.
• Audio and video may intermittently become out of sync while recording using the AV1 codec in AMD Software: Adrenalin Edition.

• Intermittent performance when entering certain areas while playing DayZ. [Resolution targeted for 24.10.1]
• Intermittent driver timeout or crash may be observed while playing Warhammer 40,000: Space Marine 2 on some AMD Graphics Products, such as the AMD Ryzen AI 9 HX 370.
• Users experiencing this issue can enable Variable Graphics Memory in AMD Software: Adrenalin Edition as a temporary measure (AMD Software: Adrenalin Edition -> Performance -> Tuning -> Variable Graphics Memory).

Here you can find similar drivers with different versions or releases for different operating systems:

• August 29, 2024
• Windows 10/11
• 717 MB

• July 19, 2024
• Windows 10/11
• 717 MB

(Image credit: Future / John Loeffler)

1. AMD Radeon RX 7900 GRE

The best graphics card overall

Our expert review:

Specifications

Core Clock: 1,500MHz
Shaders: 5,376
Ray Processors: 84
AI Processors: 336
Memory: 16GB GDDR6
Memory Clock: 2,500MHz
Outputs: 3 x DisplayPort 2.1, 1 x HDMI 2.1
Power Connectors: 2 x 8-pin
TGP: 300W
Today’s Best Deals
(Black)

Reasons to buy

Best-in-class performance
16GB VRAM
Costs the same as the RTX 4070

Reasons to avoid

Still expensive
No AMD reference card
Still loses to Nvidia cards in 3D modeling

The AMD Radeon RX7900 GRE began its life as an available-in-China-only exclusive. AMD then surprised the world by following up the Chinese launch with a global roll-out, that set the cat among the pigeons in the midrange GPU market.

Undercutting the price of Nvidia’s RTX 4070 (its closest rival) by approximately 10%, it offers a performance that stands tall against its competitor. With 16GB of VRAM under the hood, it provides a best-in-class performance at 1440p, with a very respectable 4K gaming performance to boot.

Admittedly, it still loses out when it comes to things like native ray tracing, which is where Nvidia’s more expensive cards really shine, and it can’t compete on an equal footing with 3D modeling either — Nvidia’s trump card, thanks to their CUDA software. However, for graphic design and photo/video editing, it’s a veritable beast.

With a comparative performance at a more affordable price, the RX7900 GRE has asserted its dominance in its own category. It looks the part, too, with its cool Hellhound logo, and practical features, such as an extra hole in the backplate to improve the cooling efficiency of its heatsink, keeping it below 53 degrees Celsius.

For gamers in particular, the RX7900 GRE should be firmly on your radar. Boasting a flawless performance at 1080p and 1440p, and a 4K performance that stands up well against that of the RTX 4070, this is one of the best midrange graphics cards you can buy. Whatever reasons AMD had for rolling out this former exclusive to the world, we can all be glad that they did.

Cons

• — Chiplets don’t actually improve performance (and may hurt it)
• — 7900 XT costs too much relative to the 7900 XTX
• — Weaker ray tracing and AI performance

Why you can trust Tom’s Hardware

Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.

Go to page:

AMD RDNA 3 / Navi 3x at a Glance

• Up to 96 CUs / 12,288 shaders
• 50% better performance per watt than RDNA 2
• Double the ALU count per CU
• GPU chiplet architecture
• Up to 96MB of Infinity Cache

AMD RX 7000-series specifications

With all the AMD RDNA 3 GPUs now launched and available worldwide, here are the core specifications.

Swipe to scroll horizontally

The previous generation RDNA 2 and RX 6000-series GPUs had four GPU designs spread out across a full dozen different graphics card models — and that’s not counting integrated graphics solutions. AMD RDNA 3 trims things down somewhat, with three primary GPUs and seven different graphics card models (plus some integrated versions).

The three GPU designs are called Navi 31, Navi 32, and Navi 33, going from highest performance to lowest performance. The largest die has three different models, while the other two only have two models. But there are some significant differences in how AMD gets there.

AMD RDNA 3 uses GPU chiplets for the first time, at least on the top two configurations. Navi 31 and 32 each have a large GCD (Graphics Compute Die), with anywhere from three to six MCD (Memory Cache Die) chiplets. The smallest die, Navi 33, retains the traditional monolithic die.

Each MCD offers a 64-bit memory interface and 16MB of L3 cache. The MCDs link to the main GCD via AMD’s Infinity Fabric, using what AMD calls the Infinity Fanout Bridge. It uses less power per bit than other external interfaces, though there’s certainly a power penalty relative to monolithic designs. In effect, AMD saved money on the costs of manufacturing the MCDs on a prior generation process node, which in turn shrinks the size of the GCD.

While future designs could benefit from breaking up the GPU die into multiple chiplets, for this first generation of GPU chiplets it’s basically a cost-saving measure. The Navi 33 die meanwhile still uses N6, a refinement of the N7 node used with RDNA 2, because it’s small enough it wasn’t worth the effort of breaking it into a GCD plus two MCDs.

Clock speeds haven’t changed much from RDNA 2, though AMD has a two different clock domains this time — one for the shaders and one for the front-end. AMD’s stated clock speeds are somewhat conservative this round, meaning you’ll generally see higher clocks in most games. But there are games where you may not hit the boost clock, which is what we’ve listed — AMD also has a «Game Clock» that’s even more conservative, but in practice it’s not as meaningful.

The CUs and shader cores have received a big upgrade this generation, but curiously AMD doesn’t directly call each of the potential ALU processors a shader. Instead, it lists a base value that’s half of the effective rate. So as an example, the 6,144 GPU sahders in the RX 7900 XTX can at times behave more like 12,288 shaders. That’s where the much higher compute comes from, more than doubling what was available with RDNA 2.

The top model has 61.4 teraflops of compute, and double that for FP16 workloads. Even the mid-tier 7800 XT has 37.3 teraflops of compute, which almost doubles what the RX 6800 XT offered. However, actual gaming performance doesn’t scale directly in proportion to the theoretical compute. So, as an example, the RX 7800 XT ends up performing quite similarly to the RX 6900 XT.

Memory bandwidth on the 7900 XTX increased by over 50% relative to the RX 6950 XT. That’s thanks to the move to a 50% wider interface, plus running the GDDR6 memory at 20 Gbps (versus 18 Gbps on the 6950). Effective memory bandwidth also improves thanks to the second generation Infinity Cache, which for most models decreased in size compared to RDNA 2. But that’s mostly offset by higher throughput and other tweaks to the caching hierarchy.

AMD Radeon 7000 Performance

(Image credit: AMD)

As we saw with the past two generations of AMD hardware, AMD targeted at least a 50% improvement in performance per watt with RDNA 3. While that might seem reasonably specific, it leaves plenty of wiggle room. For example, RDNA 2 delivered 50% better performance per watt than RDNA. Let’s give some concrete examples of how that played out.

According to our GPU benchmarks hierarchy, the RX 6900 XT consumes 308W of power while delivering 130 fps at 1080p ultra, 106 fps at 1440p ultra, and 63 fps at 4K ultra. A lower-tier RX 6700 XT uses 215W and averages 96 fps, 71 fps, and 39 fps, while the RX 5700 XT consumes 214W while delivering 74 fps, 53 fps, and 29 fps at those same resolutions.

Do the math, and the 6900 XT provides a 22% to 50% improvement in performance per watt, while the 6700 XT only provides a 29% to 34% improvement. If we add in all the other RDNA and RDNA 2 GPUs as reference points, the potential spread for performance-per-watt improvements becomes even wider.

AMD has also discussed some architectural improvements. The ray tracing hardware improved, but not as much as we would have liked. The biggest change architecturally was the move to dual CUs with double the execution resources. AMD also added AI Accelerators that increase the throughput of FP16 and INT8 calculations — these share resources with the shader execution cores but optimize the data flow.

Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)

Now that all the RDNA 3 GPUs have launched, we can of course provide independent assessments of how they stack up — both in performance as well as performance per watt. The above charts show the overall results from our 19-game test suite. Below, we have tables of the individual test results. About half of the 6000-series GPUs are ‘missing’ because we don’t have updated test results (yet), but we’ve provided enough data to put the 7000-series into context.

The top two GPUs, the 7900 XTX and 7900 XT, surpass the performance of anything offered by AMD’s previous generation. That’s chiefly thanks to both having more CUs, more VRAM, and more memory bandwidth than anything the prior gen offered. The 7900 XTX has 96 CUs and 24GB of memory with 960 GB/s of bandwidth; 7900 XT has 84 CUs and 20GB of memory with 800 GB/s of bandwidth. Compare that to the 6950 XT with 80 CUs, 16GB of memory, and 576 GB/s of bandwidth and it’s easy to see why performance improved at the top.

Below that mark, things become less impressive. The 7900 GRE trades blows with the 6950 XT while the 7800 XT and 6800 XT land very close together. This shows the big concern with RDNA 3: Breaking off the memory controllers and cache to create GPU chiplets wasn’t really about improving performance.

Similar CU counts end up delivering relatively similar performance, gen-on-gen. RX 7800 XT has 60 CUs and the 6800 XT has 72 CUs, so the older GPU had 20% more CUs. However, the new GPU also has higher clocks — 8% higher by specifications, but more like 5% in our test results. Overall, the 7800 XT ends up being about 5% faster, meaning the architectural improvements appear to only provide about a 10% improvement in overall performance.

Image 1 of 11
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)

Here you can see the performance per watt metrics from our full testing suite. AMD claims up to a 50% improvement in perf/W, and if you pick the right GPUs to compre, you can get there. but it requires some serious cherry-picking of results.

The 7800 XT versus 6800 XT for example shows a gen-on-gen performance per watt improvement of just 19–25 percent overall. The 7900 XT versus 6900 XT improvement ranges from 7% (at 1080p medium) to 25% (at 4K ultra). If you want more than a 50% improvement, you need to do something like compare the RX 6600 with the RX 7700 XT at 4K ultra. Except the RX 6600 was never intended to be a 4K gaming card, so that’s a pretty questionable comparison.

It’s not that the RDNA 3 architecture represents a step back from RDNA 2; it’s that it doesn’t provide the big step forward that many hoped to see. RDNA 2 had better performance per watt than Nvidia’s competing RTX 30-series, at least in rasterization games, but it also had a process node advantage. Nvidia leapfrogged AMD on process nodes and efficiency this round with the RTX 40-series, generally providing around 30~60 percent higher performance per watt.

AMD RDNA 3 Architecture: GPU Chiplets

AMD’s EPYC CPUs feature up to eight CPU chiplets surrounding the central I/O chiplet. (Image credit: AMD)

One of the biggest architectural updates for RDNA 3 is something we’ve already mentioned — the move to GPU chiplets. There are good reasons for moving to multiple chiplets, though the overall net benefit largely depends on implementation. For this first generation of consumer GPU chiplets, AMD mostly seems to be focused on reducing costs.

Previous GPU designs were monolithic, meaning everything needed for the GPU to function came from a single piece of silicon. Because of that, GPUs build in a lot of redundancy, and chips get sold in «harvested» configurations. Navi 21 was used in the 6950 XT, 6900 XT, 6800 XT, and 6800, with up to 80 compute units (CUs) or as few as 60 CUs.

Interestingly, the Navi 21 die also has eight 32-bit memory interfaces, but AMD didn’t release any Navi 21 variants with fewer memory channels. The same mostly applies to Navi 22, Navi 23, and Navi 24 — memory and cache sizes seem to have been an all-or-nothing thing. Perhaps yields on the memory controller were just really good, as the only GPU that didn’t use all the memory channels possible was Navi 22 in the RX 6700 (non-XT), with a 160-bit interface instead of the full 192-bit interface.

With Navi 31 and 32, AMD will pull all of the memory interface and Infinity Cache blocks out of the main die and move them into separate dies. Each MCD contains a 64-bit GDDR6 interface and a 16MB chunk of Infinity Cache. There were rumors that the MCDs had the option to use 3D chip stacking with another 16MB die on top of the main MCD, but if that rumor was true, AMD never implemented such a configuration. The MCDs link with the GCD via AMD’s Infinity Fabric, which will see some updates to the previous variants used in Ryzen CPUs.

Breaking out the Infinity Cache and GDDR6 memory interface has some clear advantages. Transistor densities scale best with logic circuits, less so with cache, and analog interfaces (i.e., for memory controllers) scale the worst. The on-package Infinity Fabric links on both the GCD and MCDs still require some die space, but any faulty memory controllers or cache blocks will no longer be a problem — they just get tossed.

Meanwhile, the MCDs will be made on TSMC’s existing N6 process, which costs less than the newer N5 node, and die size won’t even be a serious concern. The MCDs are only 38mm^2, meaning a 300mm N6 wafer should have enough room for about 1,600 MCDs, give or take. There’s no real difficulty in simply tossing every faulty chip rather than worrying about building in redundancies.

The GCDs for Navi 31 and Navi 32 in turn shrink quite a bit compared to the prior generation, and on the costlier N5 node, that could be very helpful. AMD can still get around 180 of the larger Navi 31 GCD per wafer or nearly 300 of the smaller Navi 32. And it can still use harvested chips with fewer compute units or even MCDs to create lower-tier products — which is why we have three cards based on the Navi 31 GCD, and two using the Navi 32 GCD.

Had AMD stuck with monolithic designs, it likely would have ended up with Navi 31 using closer to 550mm square and getting maybe 100 chips per wafer. Since N5 wafers likely cost more than twice as much as N6, this should be an easy net win from a cost perspective. It also gives AMD more flexibility with cache sizes, at least on the higher-tier products. The stacked cache is only on the MCDs, leaving the GCD free to dissipate heat into the cooling setup — that’s something 3D V-cache struggled with on its CPUs.

Comparing AMD Radeon RX 6000 and NVIDIA RTX 3000 performance

Source: AMD (Image credit: Source: AMD)

The main stats most people were awaiting were raw performance in popular PC games compared to NVIDIA’s mighty RTX 30-series GPUs. And for those on Team Red, the numbers look great. The RX 6800 XT is set to compete with the RTX 3080, and at 4K resolution in games like Battlefield V, Borderlands 3, and Call of Duty: Modern Warfare, the RX 6800 XT pulls ahead in frame rates. That’s with maxed in-game graphics settings, so you can be assured that there is no compromise.

Even dropping down to 1440p, the RX 6800 XT beats the RTX 3080 in almost all games tested. Keep in mind the RX 6800 XT is doing this at a lower 300W TDP compared to the RTX 3080’s 320W. And now with the RX 6700 XT on the table, coming in at a lower price than the RTX 3070 and beating it out in raw performance in many scenarios, NVIDIA has some close competition.

Source: AMD (Image credit: Source: AMD)

The RX 6800 compared to the RTX 2080 Ti tells about the same story. With the Smart Access Memory feature (more on that below) enabled, the RX 6800 beats out the RTX 2080 Ti at a 4K frame rate in all games tested. The same holds true when dropping down to 1440p. Considering the RX 6800 is expected to retail for about $649, this is a huge amount of performance.

Source: AMD (Image credit: Source: AMD)

Stepping up to the RX 6900 XT — which is actually no larger physically than the RX 6800 XT — you’re getting performance to rival the beefy RTX 3090. AMD padded the stats a bit with Rage Mode and Smart Access Memory enabled on its side, but at 4K with maxed setting the RX 6900 XT manages to beat the RTX 3090’s frame rates in six out of the 10 games tested. And where it fell short, it wasn’t far behind.

With the RX 6900 XT hitting the 300W TDP mark (compared to 350W) and costing $500 less than the RTX 3090, this seems like a big win for AMD.

Bad title

The requested page title is invalid. It may be empty, contain unsupported characters, or include a non-local or incorrectly linked interwiki prefix. You may be able to locate the desired page by searching for its name (with the interwiki prefix, if any) in the search box.

Possible causes are:

• an attempt to load a URL such as https://en.wikipedia.org/wiki/| (the | character is unsupported);
• an attempt to load a URL pointing to a «non-local» interwiki page (usually those not run by the Wikimedia Foundation). For example, the URL https://en.wikipedia.org/wiki/meatball:WikiPedia will give this error, because the «meatball:» interwiki prefix is not marked as local in the interwiki table. Certain interwiki prefixes are marked as local in the table. For example, the URL https://en.wikipedia.org/wiki/meta:Main_Page can be used to load meta:Main_Page. All interlanguage prefixes are marked as local, and thus URLs such as https://en.wikipedia.org/wiki/fr:Accueil will work as expected. However, non-local interwiki pages can still be accessed by interwiki linking or by entering them in the search box. For example [[meatball:WikiPedia]] can be used on a page, like this: meatball:WikiPedia.

Retrieved from «https://en.wikipedia.org/wiki/Special:Badtitle»

• Privacy policy
• About Wikipedia
• Disclaimers
• Contact Wikipedia
• Code of Conduct
• Developers
• Statistics
• Cookie statement
• Mobile view

Bad title

The requested page title is invalid. It may be empty, contain unsupported characters, or include a non-local or incorrectly linked interwiki prefix. You may be able to locate the desired page by searching for its name (with the interwiki prefix, if any) in the search box.

Possible causes are:

• an attempt to load a URL such as https://en.wikipedia.org/wiki/| (the | character is unsupported);
• an attempt to load a URL pointing to a «non-local» interwiki page (usually those not run by the Wikimedia Foundation). For example, the URL https://en.wikipedia.org/wiki/meatball:WikiPedia will give this error, because the «meatball:» interwiki prefix is not marked as local in the interwiki table. Certain interwiki prefixes are marked as local in the table. For example, the URL https://en.wikipedia.org/wiki/meta:Main_Page can be used to load meta:Main_Page. All interlanguage prefixes are marked as local, and thus URLs such as https://en.wikipedia.org/wiki/fr:Accueil will work as expected. However, non-local interwiki pages can still be accessed by interwiki linking or by entering them in the search box. For example [[meatball:WikiPedia]] can be used on a page, like this: meatball:WikiPedia.

Retrieved from «https://en.wikipedia.org/wiki/Special:Badtitle»

• Privacy policy
• About Wikipedia
• Disclaimers
• Contact Wikipedia
• Code of Conduct
• Developers
• Statistics
• Cookie statement
• Mobile view

Bad title

The requested page title is invalid. It may be empty, contain unsupported characters, or include a non-local or incorrectly linked interwiki prefix. You may be able to locate the desired page by searching for its name (with the interwiki prefix, if any) in the search box.

Possible causes are:

• an attempt to load a URL such as https://en.wikipedia.org/wiki/| (the | character is unsupported);
• an attempt to load a URL pointing to a «non-local» interwiki page (usually those not run by the Wikimedia Foundation). For example, the URL https://en.wikipedia.org/wiki/meatball:WikiPedia will give this error, because the «meatball:» interwiki prefix is not marked as local in the interwiki table. Certain interwiki prefixes are marked as local in the table. For example, the URL https://en.wikipedia.org/wiki/meta:Main_Page can be used to load meta:Main_Page. All interlanguage prefixes are marked as local, and thus URLs such as https://en.wikipedia.org/wiki/fr:Accueil will work as expected. However, non-local interwiki pages can still be accessed by interwiki linking or by entering them in the search box. For example [[meatball:WikiPedia]] can be used on a page, like this: meatball:WikiPedia.

Retrieved from «https://en.wikipedia.org/wiki/Special:Badtitle»

• Privacy policy
• About Wikipedia
• Disclaimers
• Contact Wikipedia
• Code of Conduct
• Developers
• Statistics
• Cookie statement
• Mobile view

Источники:

https://www.tomshardware.com/reviews/amd-radeon-rx-7800-xt-review&rut=8f286a8b079442021b5576aecfed0cba5c8bcaea7eaa65c91fb9215586abb505
https://www.techspot.com/downloads/drivers/essentials/amd-radeon/&rut=3d13c288a4657e7e4266c76a99760bdffb06ac56d2791b715e640def59ecfb34
https://www.techradar.com/news/amd-graphics-cards&rut=5c9d6d97ca7ff867e9c204ceecbe76940e385ba396f2c746839990ad5be06b7c
https://www.tomshardware.com/reviews/amd-radeon-rx-7900-xtx-and-xt-review-shooting-for-the-top&rut=39ed82632a8263a62eef485d496645bab55e60659b9b5224ad7f9c6c5c0cf686
https://www.tomshardware.com/features/amd-radeon-rx-7000-rdna-3-price-performance-benchmarks-release-date&rut=60983dc7ec38640200ef24b3340b9e4eb778e83bec2ca387fdeac0fadd9d9201
https://www.windowscentral.com/amd-radeon-rx-6000-need-know&rut=33e39704f2a3a2951feb07b0d057c84d836c34e66eafa0b5e362199d33170e6b
https://en.wikipedia.org/wiki/Radeon&rut=e9ae74260327ffa1e4c2efa2a46d282365b14e4baba1e9c8f37ab1ef56fa380a
https://en.wikipedia.org/wiki/AMD_Software&rut=3dbfd12bb4ff6d24709bbdf8dc17eece86c0a307a345f955e0f9d246aeaad7c4
https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units&rut=df6be3c9e69f0d87504dda98d89c4ebe9c0f5086e4b0dc501cbc60565f2d6e5b