Intel has just released its second-generation gaming GPUs, codenamed Battlemage and starting with the Arc B580 and B570 graphics cards. We’re a little disappointed that they’re only targeting Nvidia’s low-end RTX 4060 for now. But when it comes to features and technology, Battlemage seems to have an advantage over AMD. And that includes scaling.
To complement the Battlemage launch, Intel is also introducing a up-to-date version of its scaling technology, known as XeSS 2. The gigantic news is the addition of frame generation, making it highly competitive in terms of features with Nvidia’s DLSS.
There is no denying that Intel has an advantage over AMD. This is because Intel GPUs are equipped with XMX cores, which are AI accelerators analogous to Nvidia’s Tensor cores. This puts Intel on par with Nvidia in terms of AI-accelerated scaling, while AMD’s scalers are essentially hand-coded. For now, AMD GPUs do not contain cores equivalent to Intel XMX or Nvidia Tensor.
On the surface, Intel’s Frame Generation technology looks similar to Nvidia’s DLSS algorithm. It uses data from previous frames, including motion vectors and depth information, to feed it into optical flow and motion vector reprojection algorithms, which are stitched together to generate transition frames that can be inserted between fully rendered frames in the 3D pipeline.
Intel’s claims for overall performance gains are surprising. In top performance mode with F1 24 at 1440p resolution, XESS 2 scaling combined with the up-to-date Frame Generation feature increases performance from 48 fps to a somewhat astonishing 186 fps. That’s a 3.9x augment. Even at the highest image quality, performance increases by 2.8x to 136 fps.
Picture 1 With 3
To complement the XeSS Super Resolution scaling and frame generation features, known as XeSS-SR and XeSS-FG for tiny, Intel is also introducing XeLL. Here “LL” means low latency.
This feature is analogous to Nvidia’s Reflex technology, and Intel claims it can reduce latency when running at just under 50 frames per second from 57ms to 32ms. Intel also claims that XeLL can be used to essentially eliminate the latency added when enabling frame generation technology.
In other words, the latency achieved by the up-to-date B580 GPU is the same when using super resolution, frame generation and low latency mode as when using regular super resolution scaling. The idea is that you get the best of both worlds – the lower latency that scaling provides and the incredibly sleek animation that frame generation allows.
When you add it all up, you get a very complete feature set that, on paper, can rival Nvidia and outperform AMD gaming GPUs, which have essentially no AI support. We’ll have to wait and see how well XeSS 2 performs in practice and how its image quality compares to both Nvidia DLSS and AMD FSR.
