11.1 and 11.2... still getting CPU spikes?!

[Keith Moore]

Originally Posted by neilvg

I have a very similar setup to you. Read below as I am experiencing these issues too and my testing as revealed (posted from another post but relevant. Will discuss below this repost):
===
Any latest with this, or over at the duc? I've been trying to speak with Avid Tech for a while and no real results and they hardly seem to acknowledge the issue. However, after much research I do NOT think that it is pro tools or HD|Native per se. I can get this to happen in Logic and with Internal Speakers/Audio set.

Also, my issue might be a little different, but I would definitely characterize it as CPU spiking.

Here is my system:
PT11HD
HD|Native
Mac pro 5,1 12 Core 3.06 (2012) 16gb RAM
11.2.1
Lynx Aurora (using HD Interface)
Mountain Lion 10.8.5

I can open slate VTM and just click SHOW/HIDE to display the tape machine graphic and get my CPU to hit 100% (or higher in Activity Mon) every time. If I'm playing audio, this will stop playback with an AAE error. VERY Odd. Switching instruments in VI's can cause this, and so does toggling record ARM repeatedly with several tracks with plugs on them grouped together at once.

I come recently from PTHD TDM where life was a bit limited but good. Now life is limited (in a different way) and not good at all.

I did DISABLE Hyperthreading (not use XCode but using a Sudo Command:
OFF: sudo nvram SMT=0
ON: sudo nvram -d SMT

This keeps it off even on reboot. It's obviously working because sys†em usage only shows 12 cores for me now. ***This makes quite a difference for the better! Another reason why I think it's OS based - or at least an interaction between multicore processing and a given software.

Like I said, I can get spikes to occur without using the HD|Native card and using a different software - one made by Apple.

Anybody else experiencing this Show/Hide CPU spike issue with VTM? It's an easy way to see it happening. When I called Slate, they weren't able to get this to happen but I've tried this on other machines and other vanilla installations - no difference - at least with AAX and AU.

But in general, even with my fast computer, I still get all sorts of CPU 'movement' on the monitors (activity mon and sys usage) without any real reason.
====

I did some more test and it's almost like a conflict between the PCI-E bus and the harddrive. I've tried SSD's, Hackintosh's, non pro tools software (Logic), etc.. If you JUST input monitor a bunch of tracks and loudly enough so you can hear the room / preamps, go and try to instantiate an installer, or just use the compute a little. I hear little audio 'gapping' blips that sound like a quick faint DTMF glitch every now and then. When this happens the Pro Tools system usage spikes. This also can happen when you open up activity mon and take a 'sample' of the pro tools thread (hit sample button in activity mon under protools). I feel like this spiking is happening all the time on the OS interaction level and its tripping up the software at random intervals.

Any sort of disk access causes this to happen. Slate VTM just clearly shows the issue and when I rapidly toggle show/hide I get clicking and popping at even very high buffers (> 256)

Did you ever find a solution to the Slate VTM CPU issue?

[Lucid Audio]

Hello good people.
I am new to the DUC forum, so please forgive any transgression of responder protocols, such as being too verbose.
I have found a pretty reliable workaround to this deeply frustrating flaw, in an otherwise exquisite program.
My observations were that plugins acting on audio regions with little or no audio, try to amplify that audio, and the resolution required to do that places a heavy load on the processor causing the spikes.

The first work around I found was to insert the Dither plugin at the top of any channel that involved amplification such as Sans Amp PSA, or Eleven. I thought this worked because noise produced by the Dither plugin, although inaudible stopped the plugin trying to process silence and thus prevented the massive processor allocation. But then I noticed that even compressors and limiters had the same effect, causing processor spikes.

So work around two was to insert a dither plugin at the top of every channel. This was even more reliable than the previous workaround and pretty much allowed un interrupted work. As I was working with this fix, I wondered if the accumulative effect of all these plugins would result in some kind of audible artefact, so I tried globally bypassing the entire Dither plugin set, and yes there is an audible difference, (actually not a bad difference but a difference none the less).

And that led to fix number three. Run with all the Dither plugins in bypass. Don't ask me why it works, it blows my theory out of the water, but on all my systems it does.
The Dither plugin uses very little system resources so doesn't really limit the availability of other plgingins.
I'll be curious if others find this works for them.

I hope this helps
Ben

[The Weed]

Your "Dither First" idea is for Denormalization. Be interesting if that's the issue - or one of them.