https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-and-core-i3-8121u-deep-dive-review/14
To sum it up:
>having problems with the 10nm node yields
>investors and the public are starting to get antsy, the "tick-tock" cycle (one year a die shrink, with the next being a microarchitecture optimization) is looking dead
>as an attempt to get SOMETHING 10nm out the door, the i3-8121U is released
>this is a 15 watt dual-core CPU that's basically the least complex thing they could've made
>yet the yields still weren't good enough, so the fucking integrated GPU is disabled
>basically every laptop with that Core i3 needs a dedicated graphics solution
>Lenovo ends up using the RX540 in one of their education laptops
>mind you, this is a 50 watt TDP dedicated GPU in a school device
>it's also paired with a 35 watt hour battery, meaning the battery life is abysmal (going from 100-24% in 90 minutes)
>by the way, that 10nm process used in the CPU isn't actually very efficient either, offering a 10% improvement over the old 14nm process
>even with the advantage of AVX-512 for floating point math it's just not enough
>Cannon Lake itself eventually just fades away
>it took until late 2021 for Intel to finally have good enough yields to ditch 14nm++++++++ entirely (starting with Alder Lake)
To sum it up:
>having problems with the 10nm node yields
>investors and the public are starting to get antsy, the "tick-tock" cycle (one year a die shrink, with the next being a microarchitecture optimization) is looking dead
>as an attempt to get SOMETHING 10nm out the door, the i3-8121U is released
>this is a 15 watt dual-core CPU that's basically the least complex thing they could've made
>yet the yields still weren't good enough, so the fucking integrated GPU is disabled
>basically every laptop with that Core i3 needs a dedicated graphics solution
>Lenovo ends up using the RX540 in one of their education laptops
>mind you, this is a 50 watt TDP dedicated GPU in a school device
>it's also paired with a 35 watt hour battery, meaning the battery life is abysmal (going from 100-24% in 90 minutes)
>by the way, that 10nm process used in the CPU isn't actually very efficient either, offering a 10% improvement over the old 14nm process
>even with the advantage of AVX-512 for floating point math it's just not enough
>Cannon Lake itself eventually just fades away
>it took until late 2021 for Intel to finally have good enough yields to ditch 14nm++++++++ entirely (starting with Alder Lake)
