Skip to main content

Intel just stole a page from Nvidia’s DLSS playbook

hp omen transcend 32 review 13
Jacob Roach / Digital Trends

Intel is giving its XeSS upscaling tech a huge makeover. The aptly-named XeSS 2 steals — or borrows, if we’re being generous — a page from Nvidia’s DLSS 3, which has been a staple feature of some of the best graphics cards you can buy. XeSS 2 comes packed with super resolution like the original version, but also frame generation and a latency-reducing feature called XeLL. And it’s launching alongside the new B580 graphics card.

Point-for-point, XeSS 2 is basically identical to DLSS 3. The super resolution portion functions much in the same way as the original XeSS, providing you with various different quality settings to render your game at a lower resolution in order to improve performance. On the upscaling side, the major change is native support for DirectX 12 and Vulkan, which should open up XeSS to more games.

Recommended Videos

The big update overall is obviously frame generation. This works similarly to other frame generation tools by rendering two frames and comparing them to generate a frame in between. This process, called frame interpolation, is available through features like AMD’s FSR 3 and even third-party tools like Lossless Scaling. However, with XeSS, Intel is running two AI models on the graphics card in real time to support the frame generation, similar to DLSS 3.

The render pipeline for Intel's XeSS 2.
Intel

Because of that, frame generation is only available to GPUs with Intel’s dedicated XMX AI accelerators. Unlike DLSS 3, which is restricted to Nvidia’s latest RTX 40-series GPUs, Intel says XeSS 2 will work on all GPUs with XMX cores. That includes discrete graphics like the Arc A770 and A750, as well as integrated GPUs with the additional cores.

Get your weekly teardown of the tech behind PC gaming
Check your inbox!

The method is interesting here. Basically all frame generation tools use frame interpolation, but Intel has previously researched methods that use frame extrapolation. Instead of comparing two frames, frame extrapolation simply uses previous frames to predict new ones, and given Intel’s previous research on the topic, it could be a big feature in future versions of XeSS Frame Generation.

To go along with XeSS 2, Intel is introducing XeLL, or Xe Low Latency. This feature is similar to Nvidia Reflex in that it sends a frame directly to the GPU after being processed by the CPU, bypassing the typical render queue. Intel says that, by default, XeLL is required for XeSS Frame Generation to work due to the fact that Frame Generation increases latency.

Performance for Intel's XeSS 2 in F1 24.
Intel

With the new Arc B580, Intel says XeSS 2 can deliver up to 3.9x performance over your native frame rate in a game like F1 24. 

Intel XeSS 2 games.
Intel

Intel hasn’t said what game XeSS 2 will show up in first, though the company says it’ll be included in upcoming titles like Assassin’s Creed Shadows, Marvel Rivals, and Like a Dragon: Pirate Yakuza in Hawaii. Intel says the feature will launch alongside the B580, so it should appear in at least one game when that GPU releases on December 13.

Jacob Roach
Lead Reporter, PC Hardware
Jacob Roach is the lead reporter for PC hardware at Digital Trends. In addition to covering the latest PC components, from…
The first Intel Battlemage GPU benchmarks have leaked
Intel Arc A770 GPU installed in a test bench.

A new Intel discrete GPU has surfaced on Geekbench, giving us an early look at the potential specifications. This leaked listing also marks the first Battlemage GPU to surface on a benchmarking platform. First discovered by Benchleaks, the listing details a test system using an Intel Core i5-13600K processor paired to an "Intel Xe Graphics RI" GPU.

The GPU scored 97,943 points in the OpenCL test, placing it just below the Nvidia GeForce RTX 4060 and Intel Arc A770. On AMD's side, it sits between the Radeon RX 7700 XT and RX 7600 XT in performance. Using the GPU's PCIe ID (E20B), further investigation into the Linux kernel driver database confirms that it's based on Intel's next-generation Battlemage architecture. The database lists five Battlemage SKUs, and this particular GPU seems to be the second-most powerful in the lineup.

Read more
Intel just gave up on a plan it laid out years ago
An Intel executive holding a Lunar Lake CPU.

Intel is giving up on its 20A node, or at the very least, it won't show up in any desktop processors. Intel announced that it would shifting resources away from developing 20A toward its smaller 18A node. Intel 20A was the foundation of Intel's upcoming Arrow Lake CPUs. The company says it's now using "external partners" to create Arrow Lake chips, which will likely be chipmaker TSMC.

We first heard about the 20A node in 2021, where the then-new CEO Pat Gelsinger laid out a road map detailing how Intel would move off its storied 14nm node onto smaller manufacturing processes. You could read this shift to 18A as Intel jumping forward to more exciting future technology, but it doesn't bode well given Intel's long-standing road map. The release of 20A was supposed to start the "Angstrom era," as Intel called it, where we would move beyond measuring transistor size in nanometers.

Read more
Intel’s new chips just set an exciting benchmark for graphics
Intel Xe2 architecture.

Despite all the talk around the efficiency and AI performance of Intel's new Core Ultra Series 2 processors, code-named Lunar Lake, the real star of the show just might just be its integrated GPU.

I got all the details at a special event in Berlin just ahead of IFA 2024, and walked away very excited for what the new Xe2 graphics architecture could offer, as it could position Intel as the leader in graphics solutions for low-powered, ultrathin laptops.

Read more