A discrete GPU and more DRAM than can fit on the SoC package, neither of which the MBP13 now has access to.
I'm betting the M1 can't actually work with discrete GPUs (though it remains to be seen what happens if you plug an M1 Mac into an eGPU. What do Thunderbolt-supporting ARM devices generally do with an eGPU?)
Instead, discrete GPU support (and off-package [hierarchically-managed?] DRAM support) is waiting for the M1X-or-whatever, the next chip. Along with that chip, they'll refresh all the discrete-GPU product lines: the higher-end MBPs, the iMacs, and the Mac Pro.
(Personally, I'm betting that Apple will take this opportunity to get rid of truly discrete GPUs in their products, and instead partner with a graphics company like AMD to license IP cores to put in their higher-end SoCs. So you could see e.g. AMD's next-gen Vega with their Smart Access Memory tech, wired to Apple's SoC-internal UMA bus.)
I'm betting the M1 can't actually work with discrete GPUs (though it remains to be seen what happens if you plug an M1 Mac into an eGPU. What do Thunderbolt-supporting ARM devices generally do with an eGPU?)
Instead, discrete GPU support (and off-package [hierarchically-managed?] DRAM support) is waiting for the M1X-or-whatever, the next chip. Along with that chip, they'll refresh all the discrete-GPU product lines: the higher-end MBPs, the iMacs, and the Mac Pro.
(Personally, I'm betting that Apple will take this opportunity to get rid of truly discrete GPUs in their products, and instead partner with a graphics company like AMD to license IP cores to put in their higher-end SoCs. So you could see e.g. AMD's next-gen Vega with their Smart Access Memory tech, wired to Apple's SoC-internal UMA bus.)