Could Intel's transition to chiplets simplify its manufacturing process?
Could Intel's transition to chiplets simplify its manufacturing process?
Received a new idea. AMD struggled with latency in their chiplet designs for a while, but Big-Little architecture could help Intel avoid similar problems by grouping P cores together. It makes sense to use P cores for workloads that don’t require heavy parallelism, letting chiplets handle the rest. Ideally, if all cores act like a single chip in terms of behavior, latency shouldn’t be an issue for most tasks. Also, detecting and relying on P cores should work with Windows Thread Detector.
Intel is still figuring out what will make their processors fastest, possibly exploring chiplets right now. The approach to P and E cores seems reversed—maximize E cores since using a P core for non-critical tasks wastes energy. Switch to P cores only when E cores fall short. Spotify performs equally well with small portions of P or E cores, but E cores consume less power for the same tasks.
They operate with identical energy consumption for general productivity work. The CPU consumes only the necessary power, such as 60 watts or 180 watts depending on the load, using the same amount when handling tasks that require just 5 watts—because it matches their minimum requirement. The P core handles single-core or clock-sensitive activities (desktop use, gaming, web browsing, Photoshop), while the E core manages highly parallel tasks like video rendering. This setup isn't about energy conservation; it's about delivering extra performance.
Various designs consume different energy levels for identical operations. M1 Macbooks handle more work with reduced power consumption. In an extreme scenario, a 1970s supercomputer would demand significantly more energy for the same tasks as a modern smartphone. There’s no fixed power requirement for every task. P cores focus on boosting speed, whereas E cores prioritize efficiency by handling multiple tasks simultaneously without needing a P core.
It's not really that different. M1 behaves identically to Alder Lake when operating at similar power levels. Intel Alder Lake offers the ability to draw significantly more power for better performance. Any edge they had in energy efficiency comes from software optimizations and tuning, not the underlying chip design.
I wasn't looking at M1 versus Alder Lake, but rather M1 and Intel Macbooks. Intel Macbooks were built with 9th gen Coffee Lake refresh chips, which consumed more power and delivered poorer performance compared to M1 chips. The chart you shared supports my argument that different CPU designs have varying efficiency levels.
Intel 9th generation chips operate on 14nm technology, while the M1 uses a 5nm process. The design teams handle these separately, yet both belong to the same Alder Lake architecture. E-cores don't always outperform P-cores in efficiency, but they share the same core design. They don't deliver much more power compared to P-cores, just limited performance under heavy loads. You won't find my claims alone—E-cores match P-cores in power usage for similar workloads. For single-threaded tasks, E-cores consume around 75w versus 78w for P-cores. Idle power usage is slightly lower, about 56w versus 57w. The biggest efficiency gain comes during demanding operations like stress tests or multi-threaded workloads, where E-cores can cut power by up to a third while maintaining performance. They're identical in architecture but scaled down, sacrificing speed for lower energy consumption.