Before this thirty day period, AMD held an occasion to reveal the approaching Ryzen 5000 sequence CPUs, primarily based on the Zen three architecture. To near the celebration was a teaser of an approaching Radeon RX 6000 series GPU and nowadays, with the “In which Gaming Commences” celebration, the collection has finally been unveiled.
To start off the celebration, a quantity of impending video games have been shown off, and later on it was confirmed these game titles, which includes Grime 5, Considerably Cry six, and Globe of Warcraft: Shadowlands will feature AMD technologies and optimizations, which include ray tracing, variable fee shading, and FidelityFX CAS. Just before essentially having to any of the certain GPUs, AMD shared some fascinating technical specs about the new RDNA2-primarily based GPUs, like that the style and design is 50% extra successful than RDNA1, which alone was 50% additional economical than the GCN-centered GPUs just before it. This accomplishment is coming with out a node transform, as the RX 6000 series will be made on the very same seven nm procedure at TSMC as the latest RX 5000 sequence. Later on it was said that 50% was the performance purpose, but what was really attained was a 54% advancement, with that coming from frequency increase, holistic style and design optimization, and the functionality per clock enhancement from the new Infinity Cache.
Design specifics will evidently be shared closer to the launch, but the Infinity Cache was likened to the L3 cache on the Ryzen CPUs, and in one of the graphs it was shown to a lot more than double the bandwidth, even when paired with memory on a narrower bus. Precisely, the graph displays what GDDR6 memory on a 384-little bit bus supplies, and then the Infinity Cache paired with GDDR6 on a 256-little bit bus being two.17 moments increased, although also applying about 90% the ability. It will be attention-grabbing to see the specifics on how Infinity Cache performs when that info is produced, but with the comparison to a CPU’s L3 cache, it would appear it will serve to as a regional buffer so the VRAM does not will need to be identified as up as much. Staying on the die this final results in electrical power personal savings and a lot increased bandwidth, as now stated
In addition to the innovation of the Infinity Cache, AMD also mentioned the compute models have been enhanced to be 30% a lot more electricity effective. This was accomplished with pervasive great-grain clock gating, intense pipeline rebalancing, and redesigned facts paths for efficient motion.
For element assist, we can hope comprehensive DirectX 12 Ultimate guidance with DirectX Ray Tracing, variable amount shading, Mesh shaders, and sampler feed-back, but also aid for the DirectStorage API. This is some thing that will be deployed on the Xbox Sequence X as a means for the property in storage to be streamed immediately to the GPU, with it decompresses the belongings rather than waiting around on it all to go through the CPU to start with.
There is some far more of the know-how to include, but it is possibly the graphics cards by themselves we are most intrigued in. Starting off the components reveals was the RX 6800XT, which was the GPU employed for the teaser previously this thirty day period. It will characteristic 72 compute models, a Enhance Clock of 2250 MHz and a Video game Clock of 2015 MHz. As a brief reminder, the Recreation Clock is the regular clock one can hope even though gaming, but dependent on the title it may perhaps be lower or increased. It will be paired with 16 GB of GDDR6 memory and 128 MB of Infinity Cache. The complete board ability for all of this is 300 W. The RX 6800XT is currently being positioned against NVIDIA’s RTX 3080, which it trades blows with in the graphs of 4K, optimum options, ideal API functionality. As AMD also notes in the graph, the RTX 3080 consumes 20 W far more electrical power for this similar overall performance. The RX 6800XT is to start on November 18 with a recommended value of $649.
Launching on the very same day is the RX 6800 (non-XT), which is a lower-down version of the GPU with only 60 CUs and decrease clock speeds of 2105 MHz for the Boost Clock and 1815 MHz for the Video game Clock. There is even now 16 GB of GDDR6 VRAM and 128 MB of Infinity Cache. With the less CUs and reduce clock speeds the TBP is just 250 W. For evaluating it versus an NVIDIA GPU, AMD selected the RTX 2080 Ti, as it does not have accessibility to an RTX 3070, but as was revealed in our modern assessment, the two GPUs supply similar efficiency. This would make it rather fascinating then that in the 10 online games AMD as opposed the RX 6800 to the GTX 2080 Ti at 4K, the RX 6800 gained noticeably in 8 of them, and it looks eked a victory in the other two. The RX 6800 did have the advantage of Intelligent Entry Memory, but I will arrive back again to that just after covering the third GPU reveal. The RX 6800 will be priced at $579 appear November 18.
If the effectiveness of the RX 6800XT is not enough, and you have the revenue to invest, then there is the RX 6900XT sitting at the top of this new stack. Significantly of the technical specs amongst the two GPUs are the similar, like clock frequencies, but the RX 6900XT has a entirely unlocked die with 80 CUs lively. Inspite of the more compute units and identical frequencies, it however has a TBP of 300 W, producing it a 65% performance per watt raise more than RDNA1. With these specs, its opposition is the RTX 3090, and in the 10 game titles examined at 4K, the two are buying and selling blows with the RX 6900XT winning five, tying in a person, and dropping in the other 4, but not by much. The RX 6900XT experienced the gain of Sensible Access Memory and Rage Method to assistance it nevertheless, but to start with the launch particulars. The RX 6900XT is launching on December 8 at a price of $999, so just two-thirds the value of the RTX 3090.
Now to make clear what Good Obtain Memory is, as perfectly as Rage Method. Clever Entry Memory is a system-dependent technologies that will allow an RX 6000 sequence GPU paired with a Ryzen 5000 CPU, enabling complete accessibility to the GPU’s VRAM for the CPU. This allows enhanced effectiveness, but it also involves utilizing a 500 collection chipset motherboard and the attribute to be enabled in the BIOS. The functionality advancements shown were without having builders optimizing to consider benefit of the technologies.
Rage Method is the name presented to the 1-click on overclocking preset that will be designed into the Radeon Software program drivers. On the lookout at the picture proven, this is a preset like Silent or Well balanced, which are demonstrated along with it. Hunting at the computer software for my RX Vega 64, it has Turbo in the position of the Rage preset, and this is various from the automatic Undervolt GPU, Overclock GPU, and Overclock VRAM readily available below a distinctive tab.
It will be fascinating to see how the RX 6800 and RX 6900XT perform with no these two technologies At minimum with Rage Method, it seems it will be accessible to all buyers, no matter of their platform.
Some other technologies that have been touched on are coming to the FidelityFX suite, in which you now obtain FidelityFX CAS. These include a software for variable price shading based on luminance and movement, an optimized denoiser for DXR, and a “Super Resolution” aspect that is nevertheless being worked on with activity developers. Potentially this very last attribute is AMD’s response to NVIDIA’s DLSS, but by remaining underneath FidelityFX that signifies it really should be open up resource underneath the MIT license. In concept that will support with adoption, and potentially also innovation.
It is undoubtedly hunting like the GPU house is heading to be additional interesting than it has been in several years, with AMD back again in the significant finish.
Source: AMD Occasion, Anandtech  (Live web site), and