Intel Arc Alchemist Desktop Graphics Cards In Q2 2022, Workstation In Q3 2022, Celestial Architecture For Ultra-Enthusiasts In The Works
It looks like Intel themselves have confirmed that while Arc Alchemist GPUs will be coming to the notebook segment in the first quarter of 2022, desktops will get the discrete graphics card update by the second quarter of 2022 while workstations will get the first Arc Alchemist powered products by the third quarter of 2022. In addition to Arc Alchemist, it looks like Intel is also well underway on the development of its next-generation Arc graphics architecture. The company aims to offer enthusiast-grade products with its Battlemage architecture next year but the design work is already underway on its Celestial graphics architecture which comes after Battlemage and will be addressing the Ultra-Enthusiast segment. It is expected to be launched beyond 2024. It is also interesting to see Intel use the ‘Ultra-Enthusiast’ wording for the Celestial lineup. This could mean that while Battlemage is aiming for the enthusiast segment, it still won’t have GPU options to compete against NVIDIA and AMD’s highest-end offerings so those who want to see those insane discrete graphics cards that sit at the top of the hill should wait for 2024 for next-generation Celestial products.
Visual Compute Roadmap and Strategy Intel Arc Graphics Timing and Roadmap Update – AXG expects to ship more than 4 million discrete GPUs in 2022. OEMs are introducing notebooks with Intel Arc graphics, code-named Alchemist, for sale in the first quarter of 2022. Intel will ship add-in cards for desktops in the second quarter and workstations by the third quarter. Architecture work has begun on Celestial, a product that will address the ultra-enthusiast segment. Project Endgame – Project Endgame will enable users to access Intel Arc GPUs through a service for an always-accessible, low-latency computing experience. Project Endgame will be available later this year.
Intel also announced that it plans to offer a new ecosystem known as Project Endgame which will allow consumers to access Intel Arc GPUs through a service for an always-accessible and low-latency computing experience. Think of it as Intel’s answer to GeForce now and it does sound pretty cool. As for adoption, Intel plans to ship more than 4 million discrete GPUs in 2022 and aims to deliver $1 billion in revenue in 2022 but AXG plans to scale that revenue up to $10 billion by 2026.
Super Compute Roadmap and Strategy – More than 85 percent of the world’s supercomputers run on Intel Xeon processors. Building on this foundation, AXG is extending to higher compute and memory bandwidth and will deliver a leadership CPU and GPU roadmap to power high-performance computing (HPC) and AI workloads. To date, Intel expects more than 35 HPC-AI design wins from top OEMs and CSPs. Additionally, AXG has set a course that paves the way to zetta-scale by 2027. Sapphire Rapids with High-Bandwidth Memory (HBM) – HBM integrated into the package with Sapphire Rapids offers applications up to 4x more memory bandwidth, providing a generational improvement of 2.8x over Intel 3rd Gen Xeon processors. On the same computational fluid dynamics application, Sapphire Rapids with HBM outperforms competing solutions by up to 2.8x. Ponte Vecchio – AXG is on track to deliver Ponte Vecchio GPUs for the Aurora supercomputer program later this year. Ponte Vecchio achieved leadership performance results with up to 2.6x more performance compared with the leading market solution on a complex financial-services workload. Arctic Sound-M – Arctic Sound-M brings the industry’s first hardware-based AV1 encoder into a GPU to provide 30% bandwidth improvement and includes the industry’s only open-sourced media solution. The media and analytics supercomputer enables leadership to transcode quality, streaming density, and cloud gaming. Arctic-Sound M is sampling to customers and will ship by mid-2022. Falcon Shores – Falcon Shores is a new architecture that will bring x86 and Xe GPU together into a single socket. This architecture is targeted for 2024 and is projected to deliver benefits of more than 5x performance-per-watt, 5x compute density, 5x memory capacity, and bandwidth improvements. Custom Compute Group – AXG’s Custom Compute Group will build tailored products for emerging workloads such as blockchain, supercomputing at the edge, premium infotainment for cars, immersive displays, and more.
Intel also talked about its server and data center parts and has announced the AXG is on track to deliver the Ponte Vecchio Xe-HPC GPUs to the Aurora Supercomputer. Also, it looks like Arctic Sound is back in the form of Arctic Sound-M which will bring the industry’s first hardware-based AV1 encoder into a GPU, providing a 30% bandwidth uplift and is expected to sample by mid of 2022. The Data Center GPU will feature a single-slot form factor and will come with a single 8-pin connector to boot in a passive design. It was reported that Intel had canceled Arctic Sound to focus more on the Xe-HPC and Xe-HPG designs but in reality, it was just the Xe-HP GPUs that were canned. Arctic Sound-M have been positioned to serve the cloud segment and will be powered by the Xe-HPG DG2 GPU, allowing for up to 8 simultaneous 4K streams, 30+ 1080p streams, 60+ virtualized functions, 30+ game streams and AI performance of up to 150 TOPs. In addition to that, Intel is working on a new MCM design known as Falcon Shores which is going to fuse x86 CPU and Xe GPU cores together into a single socket. We have been waiting to see AMD’s Exascale APU solution for a while now but it looks like Intel may be the first to offer the said technology when it comes out with it by 2024, featuring 5x the performance per watt, 5 times the compute density and 5 times the memory capacity and bandwidth compared to traditional socket solutions. The Falcon Shores platform will be leveraging from Intel’s Xe-HPC architecture and will be based on either the 20A or 18A process node while featuring the next-generation of advanced packaging tech from Intel.
Intel ARC Gaming GPU Lineup
