In the most recent episode of PCWorld’s “The Full Nerd” podcast, Thomas Petersen, who is commonly known as TAP in the industry, spoke as a representative of Intel Arc. He talked about the journey of their first generation GPU architecture, Alchemist, and what we can anticipate in the future. TAP began by discussing the sales of Intel’s Alchemist or 1st Generation of Arc, stating that they have met the company’s expectations, and that they could potentially improve even further, especially given the value proposition that Intel is offering, including a recent price reduction on the Arc A750 down to $249 US.
Petersen also discussed the progress made by the software and development team in terms of drivers over the last few months. They were able to address some major performance issues with DX9 API and delivered a wide variety of driver releases for the latest AAA games on day-0 and day-1. In fact, the number of those releases has surpassed what AMD offers and is now comparable to NVIDIA, which is quite impressive.
All of this effort will undoubtedly prove beneficial in the development of the second generation of discrete GPUs and graphics card lineup, which is known as Arc Battlemage. While Tom couldn’t reveal specifics at this time, he did mention that the majority of their design team, including those working on the architecture and software, are currently focused on this project. The architecture itself will include some exciting new technologies, and most importantly, it will address some of the issues discovered at the architecture-level in the Alchemist GPUs. These issues would not have existed if Intel had already released a first-generation GPU. However, now that they have, they can address these concerns and ensure that Battlemage works even better right from the start.
The focus was on improving DX12 scaling, enhancing ray tracing capabilities, and even exploring AI features. However, it seems that Intel is not yet prepared to adopt NVIDIA’s AI-driven Frame Generation technology. They claim that it is not desirable for AI to handle that type of work. This is not to suggest that Frame Generation is not an improvement, but the approach taken by NVIDIA may not be necessary. AMD expressed similar sentiments in a recent interview, where David Wang, the head of RTG, commented that while NVIDIA’s approach is undoubtedly impressive, they believe that AI can be beneficial in other areas.
It’s important to note that both Intel and NVIDIA utilize AI hardware on their GPUs for upscaling techniques, such as XeSS and DLSS. AMD’s FSR 3 is scheduled to be released later this year, but the company has yet to provide much information beyond the claim of a 2x performance increase over FSR 2. AI-driven Frame Generation is an entirely separate topic, and if Intel and AMD were to adopt it on their GPUs, it would be intriguing to see what approach they would take.
Regarding the product lineup, recent leaks from the roadmap indicate that there will be a soft refresh of the Arc Alchemist series, codenamed Alchemist+, later this year. However, Battlemage is not expected until 2024, and we can anticipate more details later this year. We are aware that Intel is developing two Battlemage GPU versions, known as Xe2-HPG for Discrete and Xe2-LPG for integrated, and once again, Raja Koduri is dedicated to the Arc GPU roadmap. Intel has also stated that it will make every effort not to disappoint the gaming community in the future, and gamers can anticipate the same performance per dollar value as with the first generation of Arc.