Radeon RX 6000: Are RDNA 2's Ray Accelerators slower than Nvidia's RT cores?
Source: HW Upgrade added 29th Oct 2020
AMD has presented in the past few hours the Radeon RX series cards 6000, showing very convincing performance: the company has, however, glossed over the performance with ray tracing and a first synthetic test suggests that Nvidia has an advantage in this regard.
by Manolo De Agostini published 29 October 2020 , at 07: 43 in the Video Cards channel
AMD Radeon Ships RDNA
In the past few hours AMD has first cards announced Radeon RX 6000 and showed very convincing performance on paper , at the level of competition from Nvidia with “traditional” games. However, the presentation did not touch on the subject of ray tracing and the performance of what the company calls RA, Ray Accelerator : within the new RDNA 2 architecture we find a Ray Accelerator for Compute Unit that exclusively takes care of managing all the calculations related to ray tracing (the intersections of the rays and more ).
Waiting for the company to express itself directly and / or distribute the cards to reviewers for testing, on the AMD website, by going to read the notes at the bottom on this page, you can find the following paragraph: “Measured by AMD engineering labs 8 / 17 / 2020 on an AMD RDNA 2 based graphics card, using the Procedural Geometry sample application from Microsoft ?? s DXR SDK, the AMD RDNA 2 based graphics card gets up to 13. 8x speedup (471 FPS) using HW based raytracing vs using the Software DXR fallback layer (34 FPS) at the same clocks . Performance may vary “.
AMD then tested the performance of an RDNA 2 GPU (it doesn’t say which one, but the Radeon is assumed RX 6800 XT) with the DXR Procedural Geometry Sample application both with ray tracing accelerated by RA units and managed by software, recording respectively 471 FPS and 34 FPS. There are therefore those who have decided to check the performance of Nvidia video cards with the same application , recording an interesting picture, which at the moment we do not take as definitive but it is always better than groping in the dark.
The capabilities of the tested RDNA 2 GPU appear in line with those of the GA GPU 104 Ampere on board the RTX 3070 , with an advantage of approximately 20% sui 390 FPS totaled by RTX 2080 Ti, Nvidia’s previous flagship based on Turing architecture. La RTX 3080 and the RTX 3090 , with more second generation RT cores than RTX 3070, total respectively 630 FPS and 750 FPS, respectively 33% and the 50% more than indicated by AMD .
We repeat, it is only a scenario and moreover synthetic, “too little” to draw definitive conclusions, but conspiracy lovers could easily combine the choice not to speak ray tracing during the event at performance that promises to be lower than Nvidia. However, it should be added that the test was carried out on August 8 and in the meantime AMD may have worked on optimizations to improve the results: Nvidia also managed to profoundly improve ray tracing performance only after some time from the launch of the Turing architecture, working about drivers and with game developers.