On a desktop system? Cost to manufacture. Simpler cores are more space-efficient per IPS (instructions per second) and thus you can squeeze more IPS on a given area of die and die area is money.
In areas where you care about power and heat budget (mobile, datacenter-scale servers) you also get advantages in those terms. What you lose is the sheer single-thread speed of the beefy CPU cores, but then not everything needs to be that fast. Small cores also keep random small loads off the beefy cores (say: move the mouse pointer) meaning that those don’t have to context-switch that often meaning the get to run more instead of waiting for data.
It definitely makes sense to have a couple of them around though they’re not going to make or break a CPU, at least not on the desktop. ARM processors have been using that scheme for ages (called big.LITTLE), hardly surprising seeing as practically everything mobile runs ARM. Also Linux had scheduling support for those kinds of architectures for ages, MS definitely didn’t have to roll out a whole new OS version for that.
Fun side note: AMD’s mini Zen 4 cores are in a sense the exact same cores as their usual Zen 4 cores: They have the same gate layout. What they do is pack them differently (and giving them half the L3 cache), achieving only ~3GHz instead of the full 5.5GHz for the full cores, but fitting two mini cores into the same area as one big core.
On a desktop system? Cost to manufacture. Simpler cores are more space-efficient per IPS (instructions per second) and thus you can squeeze more IPS on a given area of die and die area is money.
In areas where you care about power and heat budget (mobile, datacenter-scale servers) you also get advantages in those terms. What you lose is the sheer single-thread speed of the beefy CPU cores, but then not everything needs to be that fast. Small cores also keep random small loads off the beefy cores (say: move the mouse pointer) meaning that those don’t have to context-switch that often meaning the get to run more instead of waiting for data.
It definitely makes sense to have a couple of them around though they’re not going to make or break a CPU, at least not on the desktop. ARM processors have been using that scheme for ages (called big.LITTLE), hardly surprising seeing as practically everything mobile runs ARM. Also Linux had scheduling support for those kinds of architectures for ages, MS definitely didn’t have to roll out a whole new OS version for that.
Fun side note: AMD’s mini Zen 4 cores are in a sense the exact same cores as their usual Zen 4 cores: They have the same gate layout. What they do is pack them differently (and giving them half the L3 cache), achieving only ~3GHz instead of the full 5.5GHz for the full cores, but fitting two mini cores into the same area as one big core.