Why Intel's Lunar Lake chip looks a lot like the Apple M

Deal Score0
Deal Score0

If the flagship chip of Computex Taipei was the Qualcomm Snapdragon X Elite, which for the first time was able to convert a large number of manufacturers to integrate its processor, Intel's low-consumption mobile chip Lunar Lake has been the subject of numerous analyses. On the one hand because Intel took advantage of the show to offer a technological dive into its chip. On the other hand because the physical structure of its processor closely resembles another component that we know well: Apple's M chips.

Advertising, your content continues below

It is important here to clarify what we are talking about: basic M chips, and not the Pro, Max and Ultra versions. SoCs which share the same microachitectures of CPU, GPU and NPU cores, but whose internal organization is quite different from simple M. Which are low-consumption, high-efficiency chips that are integrated into both the Macbook Pro and Macbook Air as well as the iPad Pro – the M4 was actually first integrated into an iPad.
From a physical point of view, a physical element is directly reminiscent of the M1 at Lunar Lake: the chip integrates its own memory.

Memory control, a major issue for energy consumption

Intel seems confident about the progress it has made in energy consumption. Not only with regard to its CPU microarchitecture, but also in all its SoC optimizations: reorganization of logic blocks, strategic placement of cache memory, tips for reducing consumption of the multimedia engine, etc.
However, the nature (type of modules) and physical form (SODIMM or soldered) of a PC's memory have a great influence not only on performance, but also on energy efficiency.

So, if we can rail against the non-removable nature of integrated/soldered RAM modules, the reality is that SODIMM modules are more bulky, but above all slower and more energy consuming than LPDDR5x modules. Intel therefore decided to protect itself.

Advertising, your content continues below

Intel delivers chips to its customers who design products with them. With these products, they can create ranges: for the same processor, the same brand can offer chassis at very different prices depending on the rest of the equipment. Whether it's the screen, an additional graphics card… Or the memory.

However, the risk is great that a manufacturer uses all the technical and marketing promises that Intel will put in place for its processor with less good memory. Intel has therefore, like Apple, integrated the memory directly on the support (package in the jargon), in order not only to eliminate this risk, but above all to carefully calibrate the interface between these two components.

Apple's invisible advantage

Lunar Lake is therefore a chip that integrates its own memory into the package. And will only be available in two memory versions: 16 GB or 32 GB. If frequency changes are possible, or even disabled CPU/GPU units, on paper, Lunar Lake should not have many variants.

Because we must emphasize here another advantage that Apple has with its chips: the number of variants is very limited. Thus, a renowned analyst told us at Computex, on condition of anonymity, that “the low number of variants of M chips allows Apple to perfectly adjust not only the memory exchanges, but also the frequencies. Whereas AMD and Apple must, on the same technical basis, decline their chips in ranges of different frequencies, different numbers of cores and compatible with a wide variety of memories, both in quantity and nature. Memory control and the limit on the number of SKUs (different models, editor's note) is critical for the performance/watt ratio of the chips“, we were told.

Advertising, your content continues below

The fundamentals of Lunar Lake were laid after the arrival of the Apple M1 at the end of 2020. A chip which benefits, in addition to its fine engraving and its optimized architecture, from another advantage: there is only two variants, only distinguished by the deactivation of a GPU core.

The fundamentals of Lunar Lake were laid after the arrival of the Apple M1 at the end of 2020. A chip which benefits, in addition to its fine engraving and its optimized architecture, from another advantage: there is only two variants, only distinguished by the deactivation of a GPU core.

© Wikipedia

It seems obvious that Intel is not going to completely shake up its practices to continue to meet the demand of its partners, particularly in other ranges such as chips for towers or high-performance laptop PCs. But Lunar Lake is a product whose mission is to restore the image of x86 chips in ultraportables. Intel therefore had to take more control than usual.

As for whether Intel could generalize this design in all its future ultramobile chips, that also depends, in part, on the success of CAMM2 memory. A format that combines some of the advantages of soldered memory and the modularity of SODIMM.

Advertising, your content continues below

More Info

We will be happy to hear your thoughts

Leave a reply

Bonplans French
Logo