2018-04-05, 10:14 PM
Further tests showed mp_pipeline is faster (usually) than running a script without it; the problem is, it takes the CPU at 100% with just one instance; if I run, for example, four parallel avisynth script at once, CPU is more or less at the same workload, but summing single performance, speed is about doubled in comparison to single mp_pipeline; I should admit that sometimes (often?) one (or more) of these parallel encoding crashes, but if I have to re-run one or few of them, it's a small price to pay, in exchange of taking "only" two days instead four, for example...
Now, I'm curious to know if a CPU with more than 4 cores takes advantage of the mp_pipeline, or, as I suspect, running eight (instead four) concurrent instances would always take half time...
EDIT: forgot that, when I did the last parallel encoding, I used a lighter script... with the new one, with added filters, parallel encoding crashed almost immediately - even if still a bit faster, and CPU around 80%... so, mp_pipeline is still faster than not using it, AND it does not crash. I guess I should wait...
Now, I'm curious to know if a CPU with more than 4 cores takes advantage of the mp_pipeline, or, as I suspect, running eight (instead four) concurrent instances would always take half time...
EDIT: forgot that, when I did the last parallel encoding, I used a lighter script... with the new one, with added filters, parallel encoding crashed almost immediately - even if still a bit faster, and CPU around 80%... so, mp_pipeline is still faster than not using it, AND it does not crash. I guess I should wait...





