AMD to Unveil New Radeon RX 6000 Graphics Card Series later this year. According to the most recent leaks, it is likely that in September we will know all the keys to the RDNA 2 architecture, which was partially presented not long ago, and that soon after, the first graphics cards based on this architecture begin to hit the market.
A recent leak suggests that there will be a total of ten versions of the Navi 21 graphics core, a chip that, as we can see in the attached table, will bring to life the mid-range and high-end graphics solutions that will form the Radeon RX 6000 family. all those variants we find some that will be exclusive to Apple, information that fits perfectly with what we have seen in recent years.
Regarding its specifications, the most reliable sources indicate that the Navi 21 XT chip will have a total of 80 active computing units, double the ones found in the Navi 10 XT core. If AMD does not introduce changes at the level of shaders for each computing unit activates the top of the range of the Radeon RX 6000 series will have a total of 5,120 shaders, twice as much as the Radeon RX 5700 XT.
Radeon RX 6000: at or above RTX-series 20 level?
It is important to note that the rival to beat for the new Radeon RX 6000 is not the GeForce RTX 20, but the future RTX 30 series, whose presentation is also expected for the month of September.
We cannot make a direct comparison, but if we start from the basis that gives us the performance of the RX 5700 XT, which has 2,560 shadersWe can carry out a well-founded and quite sensible estimation. Compared to an RTX 2080 Ti, the RX 5700 XT offers, on average, 50% less performance, although in 4K resolutions the difference becomes greater in specific cases.
The top of the range graphics card within the Radeon RX 6000 series will have twice as many shaders, but This does not mean that we will find twice the performanceIn fact, the RTX 2080 Ti practically doubles the shaders of the RTX 2070 and performs around 60% more than that.
In terms of gross power, the performance improvement derived from the increase in shaders is likely to be between 60% and 80%, at best. As we see it would be enough to overcome the RTX 2080 Ti even in the worst case, but we must bear in mind that the Radeon RX 6000 series will use a new graphic architecture, RDNA 2, and that this will bring a significant improvement at the IPC level and advanced features, thanks to the integration of dedicated hardware to accelerate ray tracing.
If the increase in the CPI is around 25% (quite reasonable) and AMD achieves increase the working frequencies between 200 MHz and 300 MHz We may find an additional performance increase of between 10% and 15%, more than enough to put earth between the top of the range of the Radeon RX 6000 series and the RTX 2080 Ti.
So far so good, but what about the RTX 3080 Ti? I know, it’s the question you all have in your head. In several leaked slides we saw that the performance of said graphics card will be, on average, 55% higher than that offered by the RTX 2080 Ti. If this forecast is met, the most powerful model of the Radeon RX 6000 family would be a little below the new flagship of NVIDIA, specifically it would be 15% to 10% slower.
Efficiency and advanced technologies: beyond gross power
Performance matters, of course, but support for advanced features like ray tracing is going to play a bigger role in the new generation of graphics cards. Such technology it is no longer a promise for the future, it is a reality that it is being taken advantage of in games, and whose short-term future has been guaranteed by its inclusion in the new generation consoles.
AMD will have to scale up in gross performance and efficiency, but it will also have to demonstrate to the user that its dedicated hardware to accelerate ray tracing can deliver an acceptable experience and is, at a minimum, above what we have seen in Turing. . I’m unclear what model the Sunnyvale company has adopted to speed up that rendering technique, so I don’t have a base to serve as a starting point, and therefore I cannot leave you an estimate analysis as in the previous case.
Ray tracing will start to become truly “affordable”, in quotes because it remains a very demanding technology, starting with the next generation of graphics cards, and Ampere will multiply Turing’s performance by four with that workload. . If this is true AMD will not have it easy, but neither will the company had it difficult after the FX Bulldozer and gave the chime with Ryzen, so let’s give it a vote of confidence.