Intel has plans for future GPUs that are not monolithic, but are built from individual chiplets, or at least there is some thinking along those lines at Team Blue.
We collected that because TechSpot noted an X resident, Underfox, who noted an Intel-filed patent for a “disaggregated GPU architecture” that will likely be the chip giant’s “first commercial GPU architecture with logic chiplets.”
Earlier this month, Intel was finally granted a patent for its disaggregated GPU architecture, which will likely be the first commercial GPU architecture with logic chiplets, which also allows for the power gating of chiplets not used to process workloads. pic.twitter.com/XsNjjdVIOuOctober 26, 2024
What exactly does this mean? All existing consumer GPUs to date have been monolithic, meaning they have a single graphics chip with everything inside. A disaggregated architecture refers to splitting up that one chip so that you have multiple chiplets instead.
This won’t happen with Battlemage, the next generation of Arc graphics cards expected to hit the market in early 2025. If something was in the works for Battlemage, we would have definitely heard about it on the rumor mill by now.
So this could be the plan for Celestial, Druid or one of the future generations of Arc GPUs – assuming Intel gets that far with its discrete graphics card range.
As always with patents, we must remember that they are often filed in a speculative manner, and many of them do not see the light of day in final products on the shelves.
Analysis: The benefits – and pitfalls – of disaggregation
Why choose a disaggregated GPU design like this? Chiplets have certain advantages in terms of chip design flexibility (modularity) and better power efficiency, with the latter being especially important for high-end graphics cards when we get into the realm of some truly wattage-sucking monsters these days.
The tricky part, however, is effectively splitting a monolithic chip into multiple chiplets, which creates the problem of ensuring those chiplets have enough fast connections to ensure performance doesn’t degrade in this way.
AMD was rumored to be looking at a chiplet design for the RDNA 4 flagship before seemingly canning it (and as we know, Team Red has reportedly retreated to just mid-range graphics cards for its next-generation RX 8000 GPUs). It was rumored that Nvidia would also look at a multi-chip design for the Blackwell GeForce flagship, the RTX 5090, but all talk surrounding this idea has died down to nothing.
One way or another, we’ll likely see chiplet designs for consumer GPUs in the future, perhaps from AMD, Nvidia, and even Intel, as evidenced by the patent here.
There are broader concerns about how far Intel will push its discrete Arc GPUs, mind you, and Battlemage graphics cards will likely only be low-end. While work on Celestial has apparently begun, it’s worth noting that Team Blue isn’t really talking much about its Arc line of GPUs these days (outside of integrated graphics, at least).