Given its current manufacturing troubles, a resurgent AMD, an incursion from Qualcomm, and Apple’s shift from buyer to competitor, it’s been a tough few years for Intel’s processors. Laptop consumers have extra viable choices than they’ve in a few years, and in some ways the corporate’s Meteor Lake structure was extra attention-grabbing as a technical achievement than it was as an improve for previous-generation Raptor Lake processors.
However even given all of that, Intel nonetheless gives the overwhelming majority of PC CPUs—almost four-fifths of all pc CPUs offered are Intel’s, in accordance with current analyst estimates from Canalys. The corporate nonetheless casts an extended shadow, and what it does nonetheless helps set the tempo for the remainder of the trade.
Enter its next-generation CPU structure, codenamed Lunar Lake. We’ve identified about Lunar Lake for some time—Intel reminded everybody it was coming when Qualcomm upstaged it throughout Microsoft’s Copilot+ PC reveal—however this month at Computex the corporate goes into extra element forward of availability someday in Q3 of 2024.
Lunar Lake can be Intel’s first processor with a neural processing unit (NPU) that meets Microsoft’s Copilot+ PC necessities. However trying past the countless movement of AI information, it additionally consists of upgraded architectures for its P-cores and E-cores, a next-generation GPU structure, and a few packaging modifications that concurrently construct on and revert lots of the dramatic modifications Intel made for Meteor Lake.
Intel didn’t have extra data to share on Arrow Lake, the structure that can carry Meteor Lake’s large modifications to socketed desktop motherboards for the primary time. However Intel says that Arrow Lake continues to be on observe for launch in This fall of 2024, and it could possibly be introduced at Intel’s annual Innovation occasion in late September.
Constructing on Meteor Lake
Lunar Lake shares just a few issues in frequent with Meteor Lake, together with a chiplet-based design that mixes a number of silicon dies into one large one with Intel’s Foveros packaging know-how. However in some methods Lunar Lake is easier and fewer bizarre than Meteor Lake, with fewer chiplets and a extra typical design.
Meteor Lake’s elements have been unfold throughout 4 tiles: a compute tile that was primarily for the CPU cores, a TSMC-manufactured graphics tile for the GPU rendering {hardware}, an IO tile to deal with issues like PCI Categorical and Thunderbolt connectivity, and a grab-bag “SoC” tile with a few extra CPU cores, the media encoding and decoding engine, show connectivity, and the NPU.
Lunar Lake solely has two practical tiles, plus a small “filler tile” that appears to exist solely in order that the Lunar Lake silicon die generally is a good rectangle as soon as it’s all packaged collectively. The compute tile combines the entire processor’s P-cores and E-cores, the GPU, the NPU, the show outputs, and the media encoding and decoding engine. And the platform controller tile handles wired and wi-fi connectivity, together with PCIe and USB, Thunderbolt 4, and Wi-Fi 7 and Bluetooth 5.4.
That is basically the identical break up that Intel has used for laptop computer chips for years and years: one chipset die and one die for the CPU, GPU, and every little thing else. It’s simply that now, these two chips are a part of the identical silicon die, relatively than separate dies on the identical processor bundle. Looking back it looks like a few of Meteor Lake’s most noticeable design departures—the division of GPU-related capabilities amongst completely different tiles, the presence of extra CPU cores inside the SoC tile—have been issues Intel needed to do to work round the truth that one other firm was truly manufacturing many of the GPU. Given the chance, Intel has returned to a extra recognizable assemblage of elements.
One other large packaging change is that Intel is integrating RAM into the CPU bundle for Lunar Lake, relatively than having it put in individually on the motherboard. Intel says this makes use of 40 p.c much less energy, because it shortens the gap knowledge must journey. It additionally saves motherboard house, which may both be used for different elements, to make techniques smaller, or to make extra room for battery. Apple additionally makes use of on-package reminiscence for its M-series chips.
Intel says that Lunar Lake chips can embody as much as 32GB of LPDDR5x reminiscence. The draw back is that this on-package reminiscence precludes the utilization of separate Compression-Hooked up Reminiscence Modules, which mix lots of the advantages of conventional upgradable DIMM modules and soldered-down laptop computer reminiscence.