Intel has no fear of NVIDIA on ray tracing – already competitive or even better Arc GPUs

“Yes, we are definitely competitive or superior to NVIDIA with ray tracing hardware“. According to Inteland in particular the executive Tom Petersen, the ray tracing unit on board the Alchemist GPUs based on the X and HPG architecture already competitive with NVIDIA’s current offeringsnamely the GeForce RTX 3000 generation.

This is no small statement, considering that NVIDIA is already in the second generation of its RT core and AMD, with the first generation of the Ray Accelerators, has shown that it has to catch up. Of course, NVIDIA and AMD are now close to introducing new GPUs that will surely take a step forward in all aspects, including ray tracing, but without a doubt Intel is not starting badly from this point of view (at least according to what it says, we have not tried the GPUs yet …).

During an interview with PC GamerPetersen explained that behind the competitiveness in ray tracing there are some key technologies which will be detailed in a video that the company will publish in the coming days.

The first advance technology to PC Gamer a BVH cache inside the GPU. This solution takes care of operations “BVH traversal“which are the way real-time ray tracing is implemented in current titles. The latter, it must be remembered, are based on rasterization and for this reason we speak of” hybrid rendering “. Another important element is the thread sorting unitwhich plays an important role in determining how to calculate ray tracing operations on the GPU.

In addition to the hardware part, Intel also made sure that these technologies require little or no commitment from the developers to work. Something everyone should do in general, but even more so if you enter a new industry and have to chase. “We tried to make our solution generic because we know we are not the GPU supplier of choice“explained Petersen.” So all the technology has to work with a low level of relationship and technological involvement of the developers. Therefore, things like the cache structure and hierarchy, as well as the thread ordering unit, work without any relationship or technological commitment on the part of the developers. “

And while Intel seems on the one hand sure of how Arc GPUs handle ray tracing, at the same time I notice the company it needs to take a leap forward in rasterization performancewhere Arc A750 and A770, the top-of-the-range solutions of this generation, compete primarily with the GeForce RTX 3060 (NVIDIA’s mid-range) and only in modern API-based titles such as DX12 and Vulkan, waiting for the drivers to mature.

Intel is aware of this and knows it is facing a long way. The first generation of Arc without a doubt arrived late, but technically complete in all respects because XeSS upscaling technology (here the latest), AV1 and much more is added to ray tracing. In short, Intel has opted to build a solid foundation and then grow in the future.

“We’re not going anywhere with our dedicated GPU business,” he said explained Petersen removing the rumors of a possible disengagement of the company from the sector. “Our GPU business is dedicated to basic technology development that goes into both datacenters and integrated GPUs.”

“Most of our ASIC team working on Battlemage. A small part about our future, Celestial. And then there is a small part on Alchemist today [] but most of the design team on Battlemage. Our plan to start from here and gradually add on top. And it’s not difficult to understandbecause we are starting with the higher volume market segment and then, over time, grow into higher performing segments“, he concluded.