During the Computex congress we get information about the new Intel Lunar Lake SoCs that bring interesting features.
It seems that the new generation of notebooks, mini PCs, tablets, AIOs and consolidated PCs is going to get very interesting. On the one hand, the new AMD AI 300 Series APUs with RDNA 3.5 were presented and now a revolution from Intel. These new SoCs manufactured in part by TSMC offer us in the same package, CPU cores, RAM, controller and the new generation Xe2 GPU, undoubtedly a change that will give us something to talk about. A solution that we have seen in Apple M processors with ARM and that now makes the leap to x64 PCs.
With this integration we will theoretically have a considerable efficiency improvement thanks to the high-speed links between the disaggregated components that form this SoC.
Intel Lunar Lake
It’s new Soc Intel Lunar Lake as we discussed, has a disaggregated microarchitecture design. A much more economical and flexible concept than the monolithic system. Intel uses the word tilesthat are connected to each other by high performance links. Each part allows the use of different integration technologies, it is divided into two parts, the first is the NOC(Network-On-Chip), and the second IO(Input-Output).
Within the NOC block there are two tiles, we have the Compute tile (CPUs) iGPU and also the NPU. In another IO block, we have the Wi-Fi 7 and Bluetooth controllers, PCI Express 5 management, USB Thunderbolt 4 controllers, sound and video controllers with support for the new VVC codec (H.266).
Computing power and GPU Xe2
As for the CPU, we have a configuration of 4 high-performance P-Core Lion Cove cores and another 4 efficient E-Core Skymont cores. As we can see, we do not have Hyper-Threading so the execution threads will be the same as the number of physical cores. The power efficiency of this new package is supposed to improve by 40% by integrating all the hardware in one SoC.
One of the most interesting points we can see in this SoC is the new integrated Intel Xe2 GPU, a hardware about which we still have little information. According to the data provided we have an improved performance 1.5x times better than the Xe1, it is not a giant leap but it is interesting. We will see how it competes with the new AMD RDNA 3.5 iGPUs and in what state their drivers are in.
Integrated RAM and AI power
In terms of RAM we have two LPDDR5X-8500 memory stacks that allow configuration of 16 or 32 GB. Thanks to this design the latency is reduced helping the iGPU, in addition we also have a lower power consumption. Of course, we have the disadvantage that we will not be able to expand the main memory, something that already happens in many devices with LPDDR soldered memory.
As it can not be otherwise, these new Intel Lunar Lake focus part of their marketing on AI power. In this package we have a dedicated NPU that can reach up to 48 TOPS. But it also allows us to add the capacity of the integrated GPU, its performance, which gives us a total figure of 120 TOPS.
For now we do not have an exact release date for the first Intel Lunar Lake laptops, but the idea is very interesting and it seems that we will finally have some competition in the low power PC sector.
