Last week Raja Kodori went to Twitter To explain Intel’s intention to stop focusing on: Intel Xe-HP By applying this technology to the latest Intel Xe-HPG GPUs, a newer and more focused ecosystem with a five-year growth plan for advancements in AI, gaming, cloud-based visual technology, optimization and high performance. This was surprising to a large audience due to the amount of time they spend discussing Intel Xe-HP publicly.
Intel targets Xe-HPG in gaming and visual cloud/AI exploratory GPU division after Xe-HP cancels
Originally announced in mid-2020, Intel’s Xe-HP GPU is a “multi-square graphics series designed for data centers, with the primary goal of media supercomputing accelerators.” Intel has gone so far as to create three separate monitors with between “1 and 4” tiles. It will now help Intel use Intel Xe-HPs for its on-premises cloud servers.
We used Xe HP to build a developer ecosystem for HPC, AI, and Visual cloud. It is used locally in Argonne and other clients.
We will have Xe HPG-based data center products for AI visual/inference cloud products and Xe HPC-based products for HPC/AI training Embedded Tweet pic.twitter.com/iHZa2dMLqd
– Raja Kodori (@aRajaontheedge) October 30, 2021
Koduri further explained that the Xe-HP GPU will help develop the “Ponte Vecchio Architecture Ecosystem”. As for the production lines, the Xe-HPG lines will replace the Xe-HP and add the “AI and Optical Cloud Interference” technology that was Xe-HP’s original goal.
As Intel Xe-HPG focuses on analytics, processing, immersion, and cloud technology media, such as cloud gaming and cloud graphics, Intel is expected to start competing with GeForce NOW, Google Stadia, or Amazon Luna.
But what is interesting is the use of media conversion servers / video content providers. Intel was one of the biggest hardware suppliers at the recent Olympics in Tokyo, Japan. Intel’s Xeon servers were the primary use for streaming 8K 60fps HDR video sent and applied directly in the cloud to TVs everywhere. While a large group of consumers have not been able to reach the full accuracy potential of the signal, the fact that this technology is now available is evidence that we are back to what is possible.
Finally, Koduri also discussed Intel’s Xe-HPG scheme to support high-end applications such as 3DS Max, and the extension of Xe-HPG to high-end workstation sections. It can be suggested that the Xe-HPG microarchitecture will be available for up to three different markets, just like its competitors NVIDIA and AMD with their own Ampere and RDNA2 architectures.
|GPU family||Intel Xe-LP (1e Genatie)||Intel Xe-HPG (1st generation)||Intel Xe-HP (1e Genatie)||Intel Xe-HP (2e Genatie)||Intel Xe-HPC (1st generation)|
|GPU Part:||Entry level (integrated + separate)||General/Advanced Games (Separate)||Data center and workstation||Data center and workstation||high performance computing|
|GPU Gen||Tech 12||Tech 12||Tech 12||Tech 13||Tech 12|
|start button:||Intel 7||TSMC 6 nm||Intel 7||It is announced||Intel 7 (base tile)
TSMC 5nm (Teagelbreaking)
7 nm TSMC (Xe Link-tegel)
|GPU Products||tiger lake
DG1 / SG1 cards
|ARC Alchemist GPU’s||arctic sound||Buyer’s Voice||old bridge|
|Specifications/Design||96 EU / 1 tegel / 1 GPU||512 EU / 1 tegel / 1 GPU||2048 EUs / 4 pieces per GPU||It is announced||8192 EUs / 16 pieces per GPU|
|memory subsystem||LPDDR4 / GDDR6||GDDR6||HBM2e||It is announced||HBM2e|
“Lifelong entrepreneur. Total writer. Internet ninja. Analyst. Friendly music enthusiast.”