Tailor-made for AI and high-performance computing (HPC) jobs, Intel claims that Falcon Shores is a powerhouse bit it needs serious cooling due to its appetite for energy. The brains of the operation in Falcon Shores are a bit like a jigsaw, slotting together x86 cores for everyday tasks and Xe cores that are all about AI and HPC.
Intel's having a bit of a boast, claiming that Falcon Shores will knock out five times the grunt per watt and pack five times the memory oomph and speed compared to what they had on the shelves in 2022. They're also banging on about a programming model that's a doddle to use, hoping to catch the eye of the coding crowd.
That power-hungry spec puts systems design on the back foot, especially when keeping things cool. Intel might have to resort to some clever tricks like liquid cooling or dunking the whole shebang in coolant—tactics they've had a gander at for other kit that gets a bit hot under the collar. These nippy solutions are bang on to keep things ticking over nicely and stop the processor from throwing a wobbly.
This lofty ambition for power from Intel means we might need to rewrite the rulebook on hardware. The current Open Accelerator Module (OAM) specs, like the OAM 2.0, only manage up to 1000W. That means Intel's either cobble together its own gizmos or give the OAM standards a nudge to handle the 1500W draw.
Looking ahead to 2025, the Falcon Shores GPU is shaping up to be a bit of a game-changer for Intel's GPU kit, showing off its next-gen Xe graphics architecture. It's elbowing its way into the AI accelerator market, backed up by Intel’s oneAPI coding toolkit. They're hoping Falcon Shores will be just the ticket for the ever-hungry AI and HPC scene, even if it does have a bit too much power.