Since the power consumption of these cores are low, we have lesser cooling requirements. Would this enable us to stack them closer to each other and have a larger performance density (datacenter space-wise)?
On a sidenote, for a typical desktop computer with discrete graphics, the GPU consumes 1.5X to 2X the power of the CPU. But on a per-instruction basis, GPU is more efficient.
Since the power consumption of these cores are low, we have lesser cooling requirements. Would this enable us to stack them closer to each other and have a larger performance density (datacenter space-wise)?
On a sidenote, for a typical desktop computer with discrete graphics, the GPU consumes 1.5X to 2X the power of the CPU. But on a per-instruction basis, GPU is more efficient.