40nm-ish. IO logic doesn't really scale, so going much beyond that doesn't really make sense when the price per mm2 keeps increasing with every node. If your lo

40nm-ish. IO logic doesn't really scale, so going much beyond that doesn't really make sense when the price per mm2 keeps increasing with every node. If your logic is simple enough that it doesn't really benefit from the increased transistor count or power savings, IO is going to be the main factor when it comes to node selection.