Optimized / Faster recording speeds?

[Dechno1337]

Is it possible to optimize the way the images (and/or videos) are being written on the drive in order to make use of the full capabilities of the drive?

A drive is the fastest when reading or writing in a sequential order + making use of many "channels" (queue depths) on the drive controller, and using more CPU threads, as far as I understand it. I don't know what the limitations are in the Source engine, but would it be possible to change (increase) compared to what is being used now?

I think the most common applications can use up to ~8 queue depths, but most modern SSDs can handle up to 32 or possibly more. Is this something that can be exploited at all, or is there a hard engine cap?

[dtugend]

Currently HLAE only writes out the images on the drawing thread, meaning 1 image at a time.

I wouldn't expect writing multiple images being notably faster on Windows 10, since it uses operating system file functions to write them, but I don't know to be honest.

We would have to gather data / clues on this.

[Dechno1337]

Not sure how relevant this is, or may shed some light on how data is sent through the system: https://devblogs.microsoft.com/directx/directstorage-is-coming-to-pc/

[Dechno1337]

Since HLAE has support for FFMPEG, I feel like this post is somewhat obsolete, unless there is something that can be improved with both FFMPEG and the classical image sequence recording method. ¯\_(ツ)_/¯ (#doubt)

[dtugend]

Merging this into #295, thus closing.