Launched in August, the Intel Arc family marks the Santa Clara giant’s debut in the gaming video card market. The line promises to shake up the duopoly held by AMD and Nvidia, encompassing competitive performance, Ray Tracing with dedicated hardware and even a proprietary AI upscaling solution, Intel XeSS, which seems to combine the best features of AMD rivals FidelityFX Super Resolution and Nvidia DLSS.
- New Intel LGA Socket2021 for Alder Lake chips leaks in real image
- Intel presents mascot for its boards video ARC
The company has now brought a series of new details about the plates, in an interview with the Japanese portal ASCII.JP
. In addition to confirming some of the features and possibilities that the Arc line will offer, the company also revealed the reasons for choosing TSMC as the responsible for manufacturing the components, and clarified whether or not this partnership could remain in the future.
Partnership with TSMC results from production capacity
Confirming multiple rumors, Intel confirmed during his presentation at Architecture Day 2021 a partnership with TSMC, which would manufacture the first Alchemist generation of Arc plates. The 6 nm N6 process will be used, a refined version of the popular 7 nm N7 present in products such as Ryzen CPUs and Radeon RX GPUs 2021 from AMD, which would put Intel in a certain position of advantage.
Want to catch up on the best tech news of the day?
Go and subscribe to our new channel on youtube, Canaltech News.
Every day one summary of the main news from the tech world for you!
This partnership is quite curious, especially considering the ambitious plans from serving other manufacturers with Intel Foundry Services. That said, billionaire investments are being made around the world to open new foundries, which indicates that it may be a matter of time for Intel to reach a comfortable level of production capacity to serve itself and other giants.
Focus on scalability and launching custom models
So far, they are Two known configurations of the DG2 Alchemist GPU — a simpler one, with 8 Xe Cores, and a more powerful one, with 32 Xe Colors. It is not known whether Intel will announce more models, but the company is keen to highlight the scalability of the Xe-HPG microarchitecture, that is, the ability to increase or reduce the DG2 in an almost linear manner.
Intel highlights the scalability of DG2, whose performance should increase or decrease linearly accordingly with GPU size (Image: Disclosure/Intel)
During the interview, the company did not rule out the possibility of releasing more models in the future, while confirming that the way in which the GPU will be scaled is directly related to the computer it will be destined for, whether a notebook or a desktop.
Another interesting point being considered is the launch of custom models by partners, such as MSI or Gigabyte. This is common practice for AMD and Nvidia GPUs, but apparently it has not yet been defined for the Intel Arc family.
We’ve painted pixels in the sky with 1, Intel drones. Now, that’s a visual experience. What will you create next? https://t.co/FYeygLy6Oh #IntelArc #inteldrones #inteldronelightshows #dronelightshows pic.twitter.com/c0Q4ycNYVS
— Intel Graphics (@IntelGraphics) August 17, 2021
As the manufacturer itself had announced in the past, the reference model is already in the hands of the partners and had its look briefly shown in a presentation of the giant’s drones from Santa Clara. Still, by all indications, Intel is inclined to encourage the production of custom models.
XeSS backwards compatibility and the absence of Xe-Link
Intel’s proprietary solution for upscaling with Artificial Intelligence, Xe Super Sampling, or XeSS, was one of the most interesting news announced by the company during the unveiling of the Arc line . In addition to the potential of the technology, which is closer to DLSS for using data from multiple frames, the compatibility with competing cards and the giant’s desire to make it an open source solution in the future are highlighted.
Also during the interview with
ASCII.JP, Intel confirmed that XeSS will be backward compatible, that is, it can be used with the DG1, the brand’s first dedicated graphics card, as well as with integrated solutions from 000 Tiger Lake generation of processors, based on the more modest Xe-LP microarchitecture.
The Iris Xe DG1 and the integrated GPUs of Tiger Lake processors will also be compatible with the Intel XeSS (Image: ETA Prime/YouTube)
This not only benefits the technology itself, which may see a growth in development adoption. developers as a result, as well as users, who will be able to extract even more performance from graphics solutions that have already shown a lot of potential.
Intel has also committed to recurring driver updates, one of the points the company’s weaknesses at the time, and recently hired former employees of companies like EA and AMD to power up software development.
Intel guarantees that the Arc family will have enough performance to compete with the Quadro and Radeon PRO lines in professional apps (Image: Playback/WCCFtech)
Beyond Moreover, the giant ensures that the Arc family is ready to deliver high performance in professional applications involved with games and animation, for example, bringing competitive performance in programs like 3DSMax to give work to rivals Nvidia Quadro and AMD Radeon PRO.
By f im, the manufacturer has confirmed that it does not intend to bring Xe-Link to mainstream consumers, keeping the technology restricted to GPUs for Xe-HPC Ponte Vecchio data centers. The feature, similar to SLI/NVLink and CrossFire, performs high-speed communication between multiple graphics chips to speed up processing. Still, with the increased power of graphics cards and the lack of support, the practice has become uncommon among gamers.
Source: WCCFTech, Tom’s Hardware
Did you like this article?
Subscribe your email on Canaltech to receive daily updates with the latest news from the world of technology.