What simply occurred? At this 12 months’s Scorching Chips convention in Palo Alto, California, Intel revealed a ton of details about its 2024 Xeon CPUs for information facilities. The corporate additionally detailed how it’s utilizing AI expertise to enhance the facility effectivity of its next-gen desktop processor lineup.
Beginning off with the Xeon CPUs, Intel defined how workloads in trendy information facilities have gotten moderately particular with each passing 12 months, requiring various kinds of chips for various kinds of jobs. For instance, some workloads like AI processing require excessive efficiency cores, whereas others could require trade-offs between core efficiency and core density.
To deal with the various necessities within the sector, Intel’s 2024 Xeon CPUs will are available two separate lineups with Efficiency Cores (P-cores) and Environment friendly Cores (E-Cores), respectively. Whereas the previous is codenamed Granite Rapids and is being optimized for compute-intensive and AI functions, the latter shall be a part of the Sierra Forest household for high-density and scale-out workloads. Each lineups, nevertheless, will share the identical platform and software program stack, which means they are going to be suitable with each other.
Intel additionally introduced that the 2024 Xeon platform will make the most of modular SoCs for elevated scalability and adaptability. The corporate believes that the brand new design will meet the processing and energy effectivity wants for AI, cloud and enterprise installations. It’s going to additionally supply help for as much as 12-channel DDR/MCR (1-2DPC) reminiscence and as much as 136 PCIe 5.0 lanes with CXL 2.0 and as much as six UPI hyperlinks.
Intel additional claimed that the Granite Rapids lineup will convey greater reminiscence bandwidth, core rely, and cache for compute intensive workloads, thereby providing 2-3 occasions higher efficiency for combined AI functions in comparison with the Sapphire Rapids household. As for Sierra Forest, Intel claims it is going to supply 2.5 occasions higher rack density and a pair of.4 occasions greater efficiency per watt in comparison with Sapphire Rapids. Each will embody a number of SKUs with variable core counts and TDPs.
Lastly, Intel additionally introduced that the Meteor Lake household will use AI to handle energy to make them extra environment friendly than their predecessors. As reported by PC World, the chips can have a VPU (visible processing unit) that won’t solely optimize AI workloads but in addition assist enhance the battery lifetime of laptops. In accordance with the corporate, the AI within the new chips can have the flexibility to determine when to shift between excessive energy (efficiency) and low energy (idling) states as a part of a brand new function that the corporate is looking ‘Dynamic Voltage and Frequency Scaling,’ or DVPS.