I would recommend R-ing the TFA: >"One solution would be to use standard chip-making techniques to build all 40 GPUs on the same silicon wafer and add intercon

I would recommend R-ing the TFA: >"One solution would be to use standard chip-making techniques to build all 40 GPUs on the same silicon wafer and add interconnects between them. But that’s the philosophy that killed Amdahl’s attempt in the 1980s. There is always the chance of a defect when you’re making a chip, and the likelihood of there being a defect increases with the size of the chip. If your chip is the size of a dinner plate, you’re almost guaranteed to have a system-killing flaw somewhere on it. So it makes more sense to start with normal-sized GPU chips that have already passed quality tests and find a technology to better connect them. The team believes they have that in a technology called silicon interconnect fabric (SiIF). SiIF replaces the circuit board with silicon, so there is no mechanical mismatch between the chip and the board and therefore no need for a chip package."