Intel has introduced new Lunar Lake chips. The company has developed a new SoC design that promises 14% higher processor performance at the same clock speed, 50% higher graphics performance, and 60% longer battery life. This was reported by The Verge.

Lunar Lake will appear in laptops designed to work with artificial intelligence this fall. Intel’s technical marketer Rob Hallock assures that the company has achieved unprecedented x86 power and will surpass Qualcomm.

Lunar Lake devices will no longer have separate memory strips or chips. The company “bakes” 16 or 32 GB of LPDDR5X memory into the chassis itself without the possibility of connecting additional RAM. According to Intel, this change reduces power consumption by about 40%.

The previous generation, Meteor Lake, had a hybrid 3D performance architecture with Performance (P), Efficiency (E) cores and a pair of new Low Power (LP-E) cores on a separate tile called the “Low Power Island”.

This island was built as a smartphone with its own Wi-Fi, Bluetooth, display controllers, memory, and low power processor cores (LP-E). The idea was that it would theoretically be possible to increase battery life without heating up other tiles and larger cores unless heavy tasks were being performed.

But it didn’t work. Programs like Microsoft Teams heated up the entire chip. So Intel abandoned LP-E cores in favor of a new 4-by-4 system.

Lunar Lake will have up to four new Lion Cove performance cores and four new Skymont efficiency cores. These E-cores now run as fast as LP-E cores, but consume one-third less power or increase performance by 2 or 4 times (single-threaded and multi-threaded).

Microsoft Teams now uses 35% less power on Lunar Lake thanks to these changes, Intel claims.

As for the powerful E-core, Intel’s Skymont is more powerful and efficient than previous P-core. At typical laptop clock speeds, they have 20% more single-threaded performance.

In the GPU space, Intel is even more optimistic: the company claims that its Xe2 GPU offers 1.5 times the graphics performance of Meteor Lake (in 3DMark Time Spy). It still has the same number of Xe cores and other functional units, but with a variety of performance and efficiency improvements.

Intel says it did not have to split its GPU into two different versions with lower and higher power: Xe2 can now scale independently across the full range of light and midrange notebooks. Intel also claims that the GPU offers 67 TOPS of AI performance in addition to the NPU.

As for the NPU, this generation of Lunar Lake triples the number of NPU hardware on the chip, doubles the memory bandwidth, and increases the clock speed from 1.4 GHz to 1.95 GHz, offering up to 48 TOPS. Microsoft requires 40 TOPS of performance for its latest Copilot+ PCs.

The tripled hardware consumes a little more power, but Intel claims it’s significantly faster: just 5.8 seconds for 20 iterations of Stable Diffusion for Lunar Lake versus 20.9 seconds for Meteor Lake, while consuming 11.2 watts instead of 9 watts previously.