Intel announced the first details of its next generation of processors Core Ultra (Meteor Lake), which will debut before the end of 2023. The chips are characterized by using the Intel 4 manufacturing process – which promises greater performance without sacrificing power consumption – as well as integrating a neural processing unit (NPU) for artificial intelligence tasks.
During the start of the Intel Innovation Day, the manufacturer announced that the Core Ultra processors will be presented on December 14. The new chips move away from the conventional arrangement and adopt a design with chiplets for each component, similar to what we see in SoCs from manufacturers such as Qualcomm or AMD. According to Intel, this methodology is one of the biggest architectural changes in 40 years and will be the spearhead for the processors of the future.
Under this design, the Meteor Lake architecture is divided into four tiles, each focused on a specific area (CPU, GPU, I/O SoC). The CPU is made up of P (performance) and E (efficiency) cores in a combination similar to that of current processors. The novelty in this generation are the electronic cores (e) that are responsible for performing tasks with lower energy consumption and they can work on their own.
For its part, the GPU tile has features that would allow games to run at a decent frame count. In terms of performance, the new Xe-LPG architecture doubles the efficiency of the previous generation Iris Xe. Although we will not obtain the advantages that a dedicated card offers, The Intel Core Ultra GPU will receive support for XeSSthe intelligent scaling solution that offers a considerable gain in frames per second.
Artificial intelligence and low consumption: the keys to the new Core Ultra
One of the most important features of the Core Ultra is the presence of a neural processing unit (NPU), the first on an Intel chip. He coprocessor is designed to run artificial intelligence workloads that need higher quality or efficiency, and that would normally run in the cloud. The NPU can connect with AI applications to execute tasks locally, speeding up the process.
Although Intel did not offer specific examples, it mentioned that the new Core Ultra disaggregated architecture offers balance in performance and power in artificial intelligence processes.
- The GPU has performance parallelism, ideal for media-infused AI, 3D applications and rendering processes.
- The NPU is a dedicated low-power AI engine for sustained AI and AI offload.
- The CPU has a fast response ideal for lightweight, single-inference, low-latency AI tasks.
The arrival of Core Ultra could represent a turning point for Intel. The manufacturer has not only included an AI coprocessor in its processors for the first time, but will rely on one of its competitors. Three of the four tiles that make up the chip will use TSMC siliconone of them manufactured under a low-consumption process that serves as a central communication bridge.
Unfortunately, Intel did not reveal information about specifications, performance tests, or configurations of its upcoming chips. So far, the only model that has been leaked is a Core Ultra 5 1003Hwhich will arrive with up to 18 threads and a GPU with 128 computing units.
The new processors Core Ultra will be officially presented on December 14 and will mark the end of the “Core i” nomenclature that the manufacturer used for more than 15 years.