That's pretty meaningless. It all depends on how bloated your code is, not on the stack. I have production PAMP apps that process requests in 1-2ms. Most of the request roundtrip time tends to be network latency. It's pretty similar peformance wise to equivalent node code. Except that PAMP is naturally multi-core capable, while node is not. It's much easier to mess up node's performance (latency) by doing too much compute, compared to the PHP app, where the multi-process model will save you, to a point.
How is it meaningless when it is literally the metric by which to determine if a stack is bloated? And I'm not going to argue with the dude that is responsible for Walmart's online shopping platform. Maybe you should talk to him.
It's meaningless to compare platforms based on just one implementation pre-rewrite and post-rewrite. If they re-wrote it into the same stack with performance as a goal, they'd get significantly less compute resource usage, too.
