nMP application optimization

[MediaBrain]

I've been working on my nMP since late January now. It is a fast machine, but I'm a little disappointed in the application usage of all that horsepower or lack thereof.

I most often work in Pro Tools, Logic, Motion, DaVinci Resolve, Compressor, and Final Cut. I don't notice this so much with the audio apps, but Final Cut performance has been a little underwhelming. I do not believe the machine is the limitation, because I monitor memory and CPU usage with menumeters app. I frequently only show 35% processor usage when I get dropped frames on HD multi cam projects with lots of effects or heavy filtering.

My system configuration is nMP, 6-core, 16GB RAM, D700. My editing drive is a 10TB thunderbolt RAID0 that tests >700MB/s read/write speed. I don't believe the disk is the issue.

I suppose it is frustrating to see that the machine has lots of gas left to give and yet FCPX is dropping frames. I'd be ok with it if the CPU was peaking, but the fact that there are cores sitting idle when this happens bugs me. In fact about the only time I ever see that CPU meter peg above 80% is during compressor jobs.

Anyhow, perhaps this is just normal performance and since I'm so hardware obsessed with the new machine I'm just now noticing it. Hopefully continued application updates will make better use of the hardware. I'm curious what experience others are having though.

----------

Also I did run geek bench just to verify the hardware. My scores are pretty comparable to the norm. I believe my hardware is perfectly fine. It just seems the software isn't fully taking advantage of it.

 

[deconstruct60]

... I frequently only show 35% processor usage when I get dropped frames on HD multi cam projects with lots of effects or heavy filtering.

My system configuration is nMP, 6-core, 16GB RAM, D700. My editing drive is a 10TB thunderbolt RAID0 that tests >700MB/s read/write speed. I don't believe the disk is the issue.

If pulling several streams from one drive then that is the more likely source of the problem. Latency is measured in seconds; not MB/s. Once the drive is locked onto pulling data from a single file then 700MB/s is probably normal. shifting between pulling in 5 or 6 sources at different rates is going to more represent random I/O access than sequential file access when all those requests are dropped onto 1-2 drives. It would not be surprising to see seek access hiccups in that context.

10TB RAID 0 is indicative of a wide stripe width but depending upon how many streams and how much the "stripe" is scaffolding higher sequential access that may not be enough.

I'd run a test were start some disk access to some other set of files on the RAID set running in background and see if the number of dropped frames goes up. If so that is more likely linked to the root cause here ( running out of latency coverage resources).

 

[JQuick]

deleted.

 

[MediaBrain]

Awesome. I'll try that. Thanks so much.

 

[VirtualRain]

How big is your internal SSD? How much free space?

I ask, because you should move as much of your workflow to your SSD as possible. Obviously don't fill it to the brim... (so Deconstruct60 doesn't need to chime in and say that it's a gross over simplification and flawed idea )

Can you move your current project files and media to your SSD?