Blackmagic Forum

What Should be best set it the Preferences under the Memory 4 frames at once or 8 or 1 or 10 and how to know what's best or it's not based on the cores at all?

I'd set it to 10, with simultaneous branching on. Seems to be faster. I could be wrong but I believe simultaneous branching increases the CPU utilisation.

I would note increasing to 10 frames at once increases memory usage.

Worth a look:

http://www.jasonmyres.com/2014/11/fusio ... ce-tuning/

Frames at once and core counts aren't related. Fusion doesn't assign one core to one frame or anything like that. It's useful in cases where you have non-CPU bottlenecks, like I/O or GPU rendering.

Ahh okay I thought this might be the case. So what about simultaneous branching, is this also better used for non cpu bottlenecks?

Generally speaking, if you want to optimize performance, you'll need to keep an eye on where the bottlenecks are. It happens that RAM is stuuuuupid cheap right now, so it's probably the best bet to default to using more RAM to make things faster. But that may not always be the case, and for users on OSX machines or laptops in general, might not be an option. But if the work you do is I/O bound or GPU bound or CPU integer or float bound, you need to keep an eye on those and adapt accordingly, assuming performance is important at the time (it often isn't).

Thanks that helped, I also found a lot of explanations in the manual (which I forgot to check in the first place) on what the options are doing for tweaking in the preferences...

Is anyone getting the same behavior, cause I do on 2 PCs: when I choose to cache loaders to disk Fusion starts playing loaders extreeeemely slow?.. Should be the other way around, shouldn't it? Is it a bug?

No, it doesn't precache, so it needs to do more I/O on demand. This causes a huge performance hit initially, but you can improve results for subsequent requests for the data.

Also, it makes the most sense when the difference in performance between the original storage and the cache storage is largest.

Thanks, understood..
And what is the subsequent request for the data?

Ideally it would just read from the cache, but I think it still checks to see if the data is valid by modified time.