Intel’s Meteor Lake integrated graphic performance has been greatly improved

This year, AMD introduced a chip codenamed “Phoenix,” featuring Zen 4 architecture CPUs and RDNA3 architecture GPUs. The top-of-the-line model boasts an 8-core, 16-thread CPU and a Radeon 780M integrated GPU with 12 CU units. From the information available thus far, the graphical performance appears quite impressive, rivaling mobile platforms such as the RTX 2050 and GTX 1650 Ti to some extent.

According to Moore’s Law is Dead, Intel’s next-generation Meteor Lake will also exhibit significant graphical performance improvements. With a 45W power output, the integrated GPU featuring 128 EUs is on par with the Radeon 780M, even comparable to AMD’s next-generation RDNA 3+ architecture integrated GPU. It surpasses the mobile platform’s 50W GTX 1650 and approaches the 35W RTX 3050.

Meteor Lake’s integrated GPU utilizes the Xe-LPG architecture and is manufactured using TSMC’s 5nm process. It comes in two configurations: 64 EUs (MTL GT2) and 128 EUs (MTL GT3), with frequencies reaching up to 2.1 GHz. The computational performance exceeds 4 TFLOPs, supporting DirectX 12 Ultimate and XeSS technology. Previous reports suggest that Meteor Lake will also feature a “package cache” function. The so-called Adamantine cache (L4 cache) will become a fundamental component of the multi-chip design, accessible by any module, likely further enhancing GPU performance.

With its transition to a multi-chip design, Meteor Lake encompasses five modules: CPU, GPU, SoC, I/O, and a base module. The Adamantine cache will provide faster access times than any typical L3 cache, which is usually part of the CPU module. Its primary function is to improve boot optimization and enhance security around the host CPU. The L4 cache will retain the cache upon reset, thereby shortening load times that would otherwise be experienced throughout all boot/reset cycles. Rumor has it that the L4 cache size currently being tested for Meteor Lake ranges from 128MB to 512MB.