Low latency and Intra Refresh

Gaming and Visualization Technologies Video Processing & Optical Flow
1

I’m using Video SDK 11.0.10, drivers 461.40, CUDA 10.1.243 on a RTX 5000 on Windows 10.
The docs state that for low latency encoding we should be using Intra Refresh, but the none of the example code, including the Low latency example, use it. I understand it doesn’t technically make sense for that example since there’s no chance of data loss when writing out to a file, but is the lack of example telling us we shouldn’t use it, and the docs are out of date? Shouldn’t the tuning preset for low latency have it on by default?

Beyond that, I have an issue where for whatever value I set my intraRefreshCnt, when that frame # is sent in, the whole driver gets into a bad state.
nvEncLockBitstream() returns NV_ENC_ERR_INVALID_PARAM in one thread, another thread that is doing encoding has a call to nvEncEncodePicture that returns NV_ENC_ERR_GENERIC, and on my main thread a call to cudaGraphicsMapResources starts returning cudaErrorIllegalAddress on a buffer it’s succeeded calling that on N -1 times before for all the previous frames.
Any ideas what’s going on?

2

To answer my own question about the errors I’m getting. I had sliceMode = 0 and sliceModeData = 1, and this was causing these errors. I changed over to sliceMode = 0, sliceModeData = 0 and that’s fine, also sliceMode = 3 and sliceModeData = 4 also works fine.

Still curious about suggested low latency workflows vs. tuning presets.