Intel Xe HP MCM GPU Teased With Massive Die Size Indicating Powerful Exascale Computing Capabilities For Commercial Operations?

Intel’s Xe GPU has been in the news for quite some time. However, Intel has only offered some hints about the existence of the same. But a recent remark by Jim Keller (Intel’s senior VP in Technology, Systems Architecture and Client Group and general manager of Silicon Engineering Group) on social media seems to have offered some compelling evidence about the size and scale of the Xe GPU.

It appears Intel’s Xe DG1 GPU might not be for entry-level gamers or even enthusiast PC builders. If the response of Jim Keller to a photo posted by Raja Koduri (senior video president, chief architect, general manager for Architecture, Graphics and Software at Intel) is to be believed, Intel is readying Xe GPU to lead the GPU segment. In other words, the Xe GPU could outperform nearly even GPU from other companies like AMD and NVIDIA.

Intel Xe HP ‘Father of All’ MCM GPU Can Easily Accommodate More Than 3000mm² Die Size?

Late last year, Raja Koduri sent out a congratulatory Tweet which included the photos of the Intel Xe Team that was working in Bengaluru, a state in India that is informally referred to as the Silicon Valley of India. The team’s photos aren’t relevant, as Koduri had included one more photo of what appeared to be a massive processor. The photo also included an AA Battery for scale. Back then he had christened this GPU, “the baap of all”, which means “Father of all”.

Koduri recently tweeted about the images and noted “The “baap of all” is back, battle-fielding and b-floating.” The image of the GPU indicates it sports a standardized LGA array. Simply put, Intel appears to be readying the mystery Xe GPU as a swappable GPU module which can be slotted inside a receptacle that’s similar to the CPU slot on a motherboard.

It appears either Intel is merely testing the Xe GPU or even exploring the actual delivery format for enterprise uses. Koduri also confirmed the support of 16-bit float with this tweet for the Intel Xe HP GPU.  This aspect is essential for the majority of AI and Deep Learning applications.

[Image Credit: VideoCardz]
According to rough calculations, the alleged Xe GPU sits on a die size that measures about 48.9mm by 75.6mm in height and width respectively. This means the die size is approximately that nears 3,700mm². The usable area will be obviously lesser than the IHS (Integrated Heat Sink). In other words, the actual Xe GPU might take up about 2,400mm². Keeping a safe margin, it would be safe to assume the actual silicon chip size sitting beneath the IHS could be close to 2000mm². Needless to conclude, such a massive GPU hasn’t yet been conceived, especially for commercial production, irrespective of the intended end-use.

Intel Xe GPU and Ponte Vecchio GPU Deploy Similar MCM Design?

MCM or Multi-Chip Module is the new methodology being adopted by the silicon chip makers. Instead of a single silicon chip, MCM chips contain multiple chips on a single die. Such a design wasn’t commercially viable earlier. However, modern-day manufacturing techniques, which Taiwan’s TSMC has perfected, have allowed accommodating multiple processors on a single die.

Although Intel’s Xe HPC GPU and Ponte Vecchio GPU might adopt similar fabrication technologies, going by the images, it is highly unlikely that the two are similar. This is simply because the Graphics processor would rely on Xe-HPC architecture. In other words, the Xe-HP is designed for media transcoding, workstation, and possibly even gaming. Experts indicate the massive GPU that Intel’s senior employees are hinting on social media, could be a worthy opponent to NVIDIA Quadro or AMD Radeon Pro.

Alap Naik Desai
A B.Tech Plastics (UDCT) and a Windows enthusiast. Optimizing the OS, exploring software, searching and deploying solutions to strange and weird issues is Alap's main interest.