One of the commenters on the original article (https://medium.com/@tarthedark/did-a-custom-bench-between-https-github-com-kataras-iris-blob-master-benchmarks-ir

One of the commenters on the original article (https://medium.com/@tarthedark/did-a-custom-bench-between-https-github-com-kataras-iris-blob-master-benchmarks-iris-main-go-fb2c413a1ae2 https://medium.com/@tarthedark/did-a-custom-bench-between-ht...) changed a few lines of the C# to use some different ASP.NET facilities which more than doubled the reqs/sec and dropped the std dev of latency by 2/3. That provides some evidence that it wasn't really an apples-to-apples comparison. Beyond that, a synthetic benchmark such as this one in which no real work is being done to service an HTTP request is unlikely to translate into real-world performance since the "real work" factors such as disk IO or database access are likely to dominate the HTTP request/response code paths. IMO, one of the places where C# shines is its thorough support for asynchronous I/O (of all kinds -- not just network) so that these kinds of services that do "real work" scale up efficiently on multi-core systems. I am also impressed with how fast Golang is, but you really have to compare the systems with real workloads over time periods sufficient to fully stress GC on their production platforms to see which will be faster.