Intel is making the right steps with its Alchemist gaming GPU


In fact, rendering games in 4K seems like fools. In a new interview, Intel revealed more details about some of the features of its first Alchemist graphics cards, including a version of AI upscaling that works similarly to NVIDIA’s excellent Deep Learning Super Sampling (DLSS).

First, let’s talk about their upscaling offering. It will be called XeSS, and Intel is hoping to use it to boost frame rates that much in games as a factor of two. A live version of the AI-based technology was shown during Intel Architecture Day in August with Unreal Engine 5, but only now do we have more details on how XeSS works compared to Nvidia or AMD’s competing technologies.

DLSS has been one of the most impressive technologies in gaming for years, which is why fans have been pushing AMD so hard to find an alternative solution. Intel has also read the tea leaves and is therefore more closely following in Nvidia’s footsteps and at the same time committed to making the technology open source.

But first a quick primer. For those who haven’t followed it, AMD and Nvidia approached the idea of ​​upscaling in essentially two different ways. The technology pioneer, Nvidia, uses neural networks and machine learning powered by specialized cores in its RTX line of GPUs. However, the first version of it, called DLSS (Deep Learning Super Sampling), had to be trained from game to game. That was slow and inefficient for mass adoption, so Nvidia came up with a new implementation that does not rely on game-specific content. Today, DLSS provides a sharper, cleaner image – especially when moving – and some of the technology’s earlier limitations, such as:

But DLSS still has one problem: you need an Nvidia GPU. AMD wanted a more open solution and built FidelityFX Super Resolution (FSR), an algorithm-based upscaling technology that works with any GPU. FSR doesn’t require any special AI hardware or neural network training, which can be especially helpful for entry-level computers or gaming laptops that are limited to lower resolutions. And since no special hardware is required, there is hope that one day FSR will appear in the PlayStation 5 and Xbox Series X, as both consoles are built with the same RDNA 2 architecture that powers AMD’s desktop GPUs.

Of course, there is a little catch. FSR’s ability to upscale images while not using an advanced neural network is pretty good. FSR can get a larger increase in frame rate even at the most aggressive settings, but image quality is degraded much more under the same circumstances than with DLSS. So which one is better really depends on your individual needs and what you can live with.

The big problem with FSR right now, of course, is adoption. DLSS starts introducing new games relatively quickly, thanks to its more general algorithm and the release of engine-level plugins in Unity and Unreal that make things massively easier for developers. Intel’s answer, according to XeSS chief engineer Karthik Vaidyanathan, is to make their technology open source while leveraging their own specialized AI hardware to potentially get the best out of the world of AMD and Nvidia.

The most modern game [techniques] Use lots of heuristics, lots of handmade approaches, to try to use as many pixels as possible, but try to do so in a way that you don’t end up incorporating wrong or invalid information, but they do. It doesn’t work all the time and so you will often see artifacts like ghosting, blurring and these are issues often associated with techniques like TAA, checkerboard rendering. And this is where neural networks come in because this is almost an ideal problem for neural networks because they are very good at recognizing complex features and this is where we can incorporate just the right amount of information and if that information is not there, try this recognize and reconstruct complex features. So that sums up the technology.

To ensure that XeSS is compatible with other vendors, Intel built its technology around an instruction set called dot product acceleration. This is currently supported by AMD (since RDNA 2) and Nvidia (since the release of their Turing 10 series cards). And similar to the current implementation of DLSS and FSR, users can choose from a range of resolution and quality settings. There will also be no game-by-game training required, which should make it easier not only for developers but also for engine makers like Unity and Unreal.

“XeSS from day one, our goal must be a generalized technique,” said Vaidyanathan in one Interview with WCCFTech. “It’s similar to the other system [DLSS 2.0], but the underlying technology is likely to be very different. Because if you have two independent groups trying to solve a problem in their own way, they are likely to come up with very creative solutions to the problem. “

An SDK for developers using Intel’s Xe Matrix Extension cores – these are the AI-powered parts of Intel’s upcoming Alchemist GPUs – will be available this month. If developers want to play around with a version of XeSS based on dot product acceleration that plays with hardware that consumers actually own today, it is said to be out later this year.

Intel is currently working with “multiple partners” to implement XeSS, which Vaidyanathan claimed to be a “game developer”. It’s likely a similar set of partners like AMD and Nvidia that AMD and Nvidia started out with – motor manufacturers like Unreal, companies like EA with their own massive internal motors like Frostbite, and so on. Intel has also worked a lot with Creative Assembly in the past few years – Total war was used as a highlight for Intel’s mobile gaming chops, so I wouldn’t be surprised if we see it Three kingdoms or Total Warhammer 3 Pop up in a future benchmark or tech demo.

All of this, of course, will answer the question people are wondering whether Intel’s Alchemist gaming GPUs are actually worth getting. There is increasing speculation that, as with AMD’s first attempts at its RDNA architecture, Intel will aim to do this the lower to middle market when Alchemist hits the market next year. In the real world, that means a competitor to the RTX 3070, RTX 3060 Ti, and the Radeon RX 6700 XT, cards that are all currently priced at or above $ 1000 in Australia. You have to imagine that Intel has to calculate the price very aggressive to convince all buyers to jump on board. But if the performance and the range of functions come close enough and the developers are satisfied with the state of XeSS, 2022 could be a very interesting year for the gaming landscape.


Leave A Reply