Skip to main content

How expensive is exception throwing in .NET 9?

· 2 min read
Aaron Cooper
Aaron Cooper
Founder @ CloudCat.dev, Software Engineer @ ASOS.com (Ex. Redgate Software)

Throughout the history of .NET development, a major concern has been the performance overhead associated with throwing exceptions. Best practices have emerged, suggesting that exceptions should be reserved for truly exceptional situations to ensure optimal application efficiency.

But is this practice still relevant today?

In this blog post, we'll explore the performance overhead of using try-catch blocks and throwing exceptions in .NET 9.

Dimage

Benchmark​

Processor Used: Qualcomm Snapdragon X Elite - X1E-78-100

Benchmarking Tool: BenchmarkDotNet v0.14.0

Operating System: Windows 11 (10.0.26100.2605)

.NET SDK: 9.0.101

https://github.com/Snazzie/DotNet-TryCatchThrow-Benchmark

| Method                   | Mean       | Error    | StdDev   | Gen0   | Allocated |
|------------------------- |-----------:|---------:|---------:|-------:|----------:|
| TryCatch_NoError         |   448.1 ns |  1.50 ns |  1.33 ns | 0.0095 |      40 B |
| NoTryCatch_NoError       |   449.9 ns |  1.24 ns |  1.03 ns | 0.0095 |      40 B |
| TryCatch_WithErrorOption |   442.8 ns |  1.67 ns |  1.56 ns | 0.0095 |      40 B |
| TryCatch_WithThrow       | 2,256.7 ns | 15.10 ns | 13.38 ns | 0.0839 |     360 B |

Insights​

Minimal Overhead for Try-Catch​

The mean execution times for Try Catch without exceptions throwing are nearly identical, demonstrating that the presence of a try-catch block itself doesn't add noticeable overhead.

Exception Throwing Overhead​

The TryCatch_WithThrow method shows a relative increase in execution time and memory allocation compared to the other methods. An overhead of ~0.002 milliseconds and additional 320 bytes of memory allocated for throwing and catching an exception, demonstrating that there is still in fact a cost to throwing and catching exceptions.

Conclusion​

While throwing and catching exceptions in .NET 9 still have a cost, benchmark result suggests this overhead is minimal and will go un-noticed in most cases. Strictly reserving exceptions for truly exceptional circumstances is not as important as it once was, but is still beneficial for optimizing application efficiency. Depending on the project needs, it is worth weighing the cost of developer time to produce and maintain graceful error handling vs bubbling up exceptions.

It is also important to keep in mind that this is a very simple synthetic benchmark, the cost could be much greater in a large code base where exceptions bubble up from great depths or where hardware resources are limited.

Feel free to dive into the benchmark code and share your thoughts on these findings or any improvements!