Ncu to produce similar output as nvprof

nvprof is not working with compute capabilities 8.0+, but it was producing extremely useful information and was barely slowing down the running process.

ncu is extremely slow and tries to output too many things, which can be useful in some situation, but not always.

Is there a way to get ncu to output the same kind of data nvprof did?

It’s not clear from your post, but I am presuming you are comparing nvprof’s default tracing mode to ncu’s kernel profiling, which don’t attempt similar things. For low-overhead tracing of applications using newer architectures, you should be using Nsight Systems, not Nsight Compute. The latter replaces the event/metric profiling functionality of nvprof, with lower overhead than the old tool.

You can find more related information in the nvprof and Visual Profiler transition guides.