Efficiency projections leaked to Igor’s Lab present what Intel’s inner expectations are for its subsequent two generations of processors. The numbers spotlight an as much as 21% efficiency uptick for 2024’s Arrow Lack CPU, over the present high thirteenth Gen CPU and a large increase to its gaming graphics efficiency, too.
Although the upcoming Raptor Lake refresh, or 14th Gen of Intel’s desktop processors, do not look wherever close to so comparatively spectacular. We’re speaking a 4% enchancment, at finest.
Examined towards the Core i9 13900K as a baseline, subsequent yr’s Arrow Lake chips not solely look to be up 21% quicker total—in a choice of integer and floating level benchmarks—but additionally smash their predecessor with round double the iGPU efficiency in 3DMark Time Spy.
That is fairly a efficiency increase for Arrow Lake’s tiled GPU, however simply as was the case for the Core i9 13900K that got here earlier than it, the chip is not precisely centered on gaming. Double the UHD 770’s Time Spy determine would put it round 2,000 marks, which is round two-thirds the efficiency of AMD’s spectacular Radeon 780M GPU within the Ryzen 7 7840U utilized in fashionable gaming handhelds.
We’re nonetheless a 240% enhance over Raptor Lake’s iGPU efficiency for Arrow Lake, although, which is fairly spicy.
Alongside these efficiency numbers, it has been discovered that upcoming Arrow Lake-S and Lunar Lake chips will help a couple of necessary directions, together with one significantly fascinating extension meant for bashing out the sorts of calculations wanted to run some fairly highly effective AI functions. We’ll be seeing an enormous deal with neural networking, then, together with some spectacular enhancements in effectivity, safety, and total efficiency from the upcoming chips, together with the Raptor Lake refresh.
The anticipated 24-core (8 P-core, 16 E-core) chips won’t solely each make use of the safety and encryption enhancing SHA512, SM3, and SM4 directions, however importantly the AVX-VNNI-INT16 perform meant to enhance the velocity and effectivity of AI workloads, i.e. the ever sophisticated duties being run by neural networks. That is in response to an structure data sheet from Intel (PDF warning) noticed on Videocards.
The latter AVX-VNNI instruction does so by bettering the chips’ capabilities round inference duties, resembling these present in AI functions, due to its means to run specialised 8-bit and 16-bit integer duties.
Whereas we knew Arrow and Lunar Lake would offer completely different high quality of life options, with Intel heralding Lunar because the performance-per-watt chief and giving us the impression that it may effectively be a cellular first chip, we are able to at the least say now each chips will probably be AI focussed.
Arrow Lake will probably be primarily based on the Intel 20A manufacturing course of—that is the primary to modify from the traditional FinFET transistor design to a RibbonFET, or gate throughout (GAA) design. It is also bringing bottom energy supply, or PowerVia, to its processors, which is designed to simplify the wiring and probably ship efficiency enhancements of its personal.
Lunar Lake ought to then observe in both late 2024 or later in 2025, on the newer, even smaller Intel 18A course of. In response to Intel, the roadmap is six months forward of schedule, placing the possible launch date someplace in late 2024. So put together yourselves for some intense AI-centric chips by the tip of play subsequent yr, with the Raptor Lake refresh anticipated to hit cabinets later this yr.
