Is Bun much faster than Node.js?

medium.com

6 pointsredbell4 comments

This seems incomplete: why was the only real package tested this QR package? Is this specific package a performance anomaly or a real reflection of real-world performance difference?

Interesting findings. I was always a bit wary about just benchmarking the raw http implementation instead of a more real world scenario.

Problem is that "real world" adds way more variables that you cannot control.

I can make a hello world api in go that will handle way more requests than an express app but as soon as I put a database call we'll the go/node are pretty much neck and neck because the dB doesn't really respond any fast.

Now I know the go server will literally be able to handle more concurrent clients but that is not "speed" benchmark anymore.

I've seen benchmarks where Node outperforms Go when utilizing Node 's worker threads [0]. But you're right, making real world benchmarks is hard. Even though the mentioned article is using a pure javascript qr code module with no external dependency (eg a database) in order to test a bit more than just the http implementation which is probably just the C implementation in Node.js and the ZIG version in Bun.

[0] https://medium.com/@Cazineer/go-goroutines-vs-node-cluster-w...

Reply on news.ycombinator.com