Advanced Caching

So I just installed a Kingston Predator PCIe based SSD (capable of 1.4GBytes/sec) and did a clean install of 10.8.5

I’m really hoping to utilize it instead of RAM caching, but there doesn’t seem to be any way to specify it.

I can still watch my RAM get depleted down to nothing. How does Amount of Texture RAM to Preserve affect caching in general, or is that GPU RAM?

When I watch the tmp folder, playing the clip through doesn’t necessarily invoke writing cache files, while “Cache Original Clip” does.


Also, on the old system, after manually caching a clip (switch to 1/4 and play it through) I could shut down mocha and the cache would still be valid, even when switching between a half dozen different shots. Now, no matter which option I choose when I shut down Mocha (retain or don’t retain rendered caches) the clip seems to recache when I open it again.


I guess what I’m saying generally is, with the advance of fast SSD’s and the industry’s ridiculous trend to move to 6k 32bit (almost 200MB/frame) plates, it’d be nice to see a major rework of the caching methodology.



also, I’m thinking about upgrading the old 8800 GFX card to something better. Will this affect playback in Mocha? Would there be a noticeable difference between a Quadro or Geforce card?
Does Mocha take advantage of the CUDA system?


Texture RAM is based on the GPU. This helps things like the GPU rendered matte files etc. We don’t presently use CUDA to assist this however.

We cache clips at time of playback, so it will appear slow on the first playback but then will speed up after that. Checking “cache original clip” does this beforehand, but it usually not necessary unless you’re caching from a slow server.

We are improving caching for V5, but can you tell me which version you are on? The cache not persisting seems like a new bug.

4.1.2 build 9658

Just did a clean install of 10.8.5

when it writes to the disk seems almost random unless I click the “Cache original clip” option

It’d be nice to something more like Syntheyes’ caching, where you can manually limit the hit on RAM to prefetch and have the bulk of the cache come off of a superfast SSD.



This is something we can look at. I’ll add a feature request here.

But the non-persisting cache definitely seems like a defect. We’ll take a look.