It's interesting that the results on Techempower (on TE's hardware) remain relatively stale.
On the more "focused" application and micro benchmarks the difference between 7, 8 and 9 is very significant. The annual compiler improvements match those that take about 3 years for OpenJDK, and more for Golang. This has reached the point where you could reasonably expect e.g. LINQ to breathe down the Rust's iterators neck in un-vectorizable code (which is 98% of it), I didn't expect this would actually happen.
I have been told that each Linux kernel update tends to hit Techempower scores particularly hard, and practically everything in 75th percentile tends to be very sensitive to specific hardware configuration and DB driver implementation quality.
GC improvements while somewhat hurting Techempower too should make everyone deploying on small containers very happy as it pushes GC design in a Go-like direction with very minor hit to throughput (which is still massively better) while yielding significant sustained RSS reduction.
Personally, I feel like .NET has an identity crisis of sorts - it is being marketed as a regular high-level language for regular applications and web development, where it has plenty of competition.
At the same time it never seems to be marketed for low-level, systems and systems-adjacent scenarios, where .NET's competition is far behind. Any new DB or a message broker or a web server that is written in Java or Go would have benefited massively from being built with .NET instead, as proven by projects like Ryujinx and Garnet/FASTER. Nor it is being marketed at FP public - F# has access to the whole .NET ecosystem and easily slots into existing .NET projects better than Kotlin, Clojure or Scala do into Java ones.
> Personally, I feel like .NET has an identity crisis of sorts - it is being marketed as a regular high-level language for regular applications and web development, where it has plenty of competition.
>
> At the same time it never seems to be marketed for low-level, systems and systems-adjacent scenarios
I think it would be really good if they spun off a "T#" which moved closer to TypeScript and effectively chopped off some of the older syntax and even some of the features in favor of a more streamlined language.
C# and TypeScript are already fairly similar. In my mind, T# would be like a trimmed down C# with a bit more influence from F#.
On the more "focused" application and micro benchmarks the difference between 7, 8 and 9 is very significant. The annual compiler improvements match those that take about 3 years for OpenJDK, and more for Golang. This has reached the point where you could reasonably expect e.g. LINQ to breathe down the Rust's iterators neck in un-vectorizable code (which is 98% of it), I didn't expect this would actually happen.
I have been told that each Linux kernel update tends to hit Techempower scores particularly hard, and practically everything in 75th percentile tends to be very sensitive to specific hardware configuration and DB driver implementation quality.
GC improvements while somewhat hurting Techempower too should make everyone deploying on small containers very happy as it pushes GC design in a Go-like direction with very minor hit to throughput (which is still massively better) while yielding significant sustained RSS reduction.
Personally, I feel like .NET has an identity crisis of sorts - it is being marketed as a regular high-level language for regular applications and web development, where it has plenty of competition.
At the same time it never seems to be marketed for low-level, systems and systems-adjacent scenarios, where .NET's competition is far behind. Any new DB or a message broker or a web server that is written in Java or Go would have benefited massively from being built with .NET instead, as proven by projects like Ryujinx and Garnet/FASTER. Nor it is being marketed at FP public - F# has access to the whole .NET ecosystem and easily slots into existing .NET projects better than Kotlin, Clojure or Scala do into Java ones.