We’ve all been in that situation — desperate to add one more plug–in, but with a computer on the verge of collapse. We explain how to wring that last ounce of power from your Pro Tools rig.
Computer power grows every year, but so to do the demands we place on our systems. The increasing sophistication and popularity of software instruments, samplers and the like means that conserving CPU power is still crucial for many of us. In this month’s workshop we’ll be looking at ways in which Pro Tool users can do just that.
Ultimately, the raw processing power of your computer is the key to how well your Pro Tools system will perform, whether you use a Mac or PC, TDM, LE or M–Powered system. It matters even more on host–based systems such as LE and M–Powered, because TDM systems have DSP cards to take care of most audio processing duties, leaving the host processor to manage data management, routing, driving the screen and so on. Either way, your host computer has a finite amount of processing power and when it’s gone, it’s gone! Hard–drive bandwidth can also be an issue, so although you may be able to play back a Session with 32 tracks just fine, the system may start to cough if you do some heavy editing or add lots of plug–ins and automation.
Squeezing the last ounce of performance out of your system is nearly always going to be a trade–off of some description. Whether that means fewer plug–ins for lower latency, or rendering effects to reduce the number of active RTAS plug–ins, to quote just two examples, you will end up having to make a choice between several evils.
The first place to visit is the Playback Engine window (under the Setup menu) where you can adjust various settings to help you get extra performance in some areas at the expense of a hit in other areas. In this window (the screens below are from my HD2 TDM system running on a dual 1.42GHz G4 Mac and my M Box 2 Pro system running on a 1.67GHz G4 Powerbook) you can adjust the following:
The CPU Usage Limit option sets the maximum proportion of your computer’s processing power that will be allocated to Pro Tools’ audio engine. For example, the default setting permits up to 85 percent of total CPU resources to be devoted to running RTAS plug–ins, with only 15 percent left over for all other duties. The default 85 percent is the maximum you can assign on computers with a single processor, but it isn’t always best to set the usage limit to the maximum. If you do, the remaining 15 percent of CPU power will have to do all the other chores, like displaying meters and graphics, controlling automation and dealing with MIDI; ideally, you should allow around 35 percent to handle those processes. In general, the fewer RTAS plug–ins you are using, the lower you can set the CPU Usage Limit. You will soon find out if it is too low, as you’ll get –9128 errors when there aren’t enough processing resources for the RTAS plug–ins you have in your Session. When the CPU Usage Limit is set to 85 percent or above and the RTAS Error Suppression box is ticked, you can experience a slowdown in screen redraw and background CPU tasks. Try lowering the CPU Usage Limit by 5 to 10 percent if this happens.
This setting determines the number of processing cores in your computer that can be allocated for RTAS plug–in processing. Used in conjunction with the CPU Usage Limit setting, the RTAS Processors setting lets you control the way RTAS processing and other Pro Tools tasks are handled by the system. If you have a multi–processor computer it is possible to set the CPU Usage Limit right up to 99 percent, as long as you leave one processor free to take on the housekeeping duties. For example, with sessions containing a large number of RTAS plug–ins on a multi–core machine, you should allocate two or more processors to RTAS processing and set a high CPU Usage Limit, but leave one processor unallocated to RTAS duties. However, setting the CPU Usage Limit above 95 percent may cause –6086, –6093, or –6097 errors. If a –6097 error occurs while Pro Tools is idle, you should reduce the number of plug–ins active in the current session.
The hardware buffer size is the number of audio samples that the CPU passes to and from the I/O hardware in one go. Larger buffer sizes give the computer more time to process audio, but latency is increased, and the system will be more sluggish in responding to commands, as the buffer has to be filled before it can do anything. Small buffer sizes reduce the amount of latency and make the system more responsive, but place more demands on the system, and mean you can run fewer plug–ins.
The usual advice is to lower the buffer size while recording, to minimise latency that will cause monitoring problems, and raise it again when you come to mix, so that you can run more effects. The maximum buffer size available will vary depending on the Pro Tools software version and which type of Pro Tools hardware you have attached at the time. I always set the buffer size to maximum, but do be aware that larger buffer sizes can affect automation accuracy for plug–in parameters and mute data, as well as timing for MIDI tracks.
The higher the sample rate, the higher you will need to keep the H/W Buffer Size to keep the Session playing, especially if automation is getting dense. Low settings can also cause -9128 errors occurring during long recording passes; try giving the CPU Usage Limit the highest available setting if this happens.
If you have just upgraded to v7.3 on Mac OS X and feel that your system is more sluggish and the latency has jumped up even though the H/W Buffer Size hasn’t changed, you are right. Apparently, prior to version 7.3 the DAE (Digidesign Audio Engine) on Mac–based LE systems with a Digi 002/002 Rack interface actually ran at half the displayed H/W Buffer Size. This was changed in v7.3 so that the buffer accurately corresponds to the displayed value. To maintain similar performance after updating to Pro Tools LE 7.3, you should change the H/W Buffer Size to the next setting down.
The DAE Buffer Size sets the amount of memory that the DAE allocates for disk buffers, so that Pro Tools has some buffer space to take content off drives. The default setting is Level 2. It is possible to go below Level 2, and doing so will reduce the amount of time Pro Tools takes to start playback and recording, but the down side is that as the Session gets bigger and you add more edits, the heads on your drives have to travel further to get all the data together. Eventually, the buffers will run empty and Pro Tools will complain, usually with the –9073 Disk Too Slow error.
You can increase the DAE Buffer Size above Level 2, which will allow Pro Tools to carry on working with Sessions containing higher edit densities or more tracks, or which are stored on slower, more fragmented drives. Pro Tools may prompt you to restart the computer so it can allocate more memory to Pro Tools, but even if there is enough memory — and there usually is enough to go up to Level 4 — you should still close and then re-open for the new buffer setting to take effect.
You sometimes can get a DAE error –6047, which is ‘Not enough system memory for audio buffers’. This error requires a restart, and is more likely with larger DAE buffer sizes. At the other extreme, though, I don’t recommend you drop to Level 0; for low memory usage, use Level 1. If the disks are too slow and you are suffering –9073 errors, go to Level 4. I don’t recommend going to Level 8, as there is a documented problem on LE systems which can give rise to an ‘Assertion Error’ whilst bouncing to disk.
Typical 7200rpm hard drives are usually able to provide enough seek performance for use with Pro Tools, but you’ll have more ‘headroom’ with 10,000rpm drives, and you’ll see fewer –9073 Disk Too Slow errors when using lots of tracks and high edit density. Many laptops still use 5400rpm drives, which usually have seek times above 10ms and are not recommended with Pro Tools. You can use them, and I do for smaller jobs on the road, but you won’t be able to get as many audio tracks per drive, and you may need to increase the DAE Buffer Size to avoid –9073 errors.
The ‘wrong’ type of Firewire drives have also been known to cause -9073 errors. All Firewire drives should have a rotation speed of 7200rpm and must have one of the Oxford Bridge chip sets. For more details, see the Compatibility section of the Digidesign web site. This is very important: I have lost count of the number of folk complaining of poor track counts where it turns out that they’re using some dodgy drive in some unknown drive case with the wrong chip set.
Audio tracks with lots of edits, as are often created by using Beat Detective, can also cause –9073 errors. If this is a problem, try bouncing to disk or using Consolidate Selection to create a single file and Region from the track. (Make a copy of the track first and disable it if you want to keep the edits to go back to.)
Once you have tweaked the CPU Usage and H/W Buffer Size, and have installed the maximum amount of RAM and the fastest drives available, what if it still isn’t enough? There are a number of things you can do, some of which we have alluded to already:
A bypassed plug–in doesn’t process audio but still consumes resources. You can make a plug–in inactive by Command+Control–clicking (Mac), Control+Start–clicking (Windows) or right–clicking on the plug–in. Names of inactivated plug–ins are displayed in italics.
There are two ways to make a track inactive. You can turn off its voice in the mixer by clicking where it says ‘dyn’ and choosing ‘off’ instead; this will stop audio from playing back and deactivate any inputs to and outputs from that track, but any plug–ins will remain active. More useful for saving CPU resources is choosing Make Inactive from the Track menu. Doing this makes sure that all inputs, outputs, sends, inserts and all plug–ins are deactivated and all DSP resources for that track are freed up. An inactive track is greyed out. When you have finished and want to reactivate that track, turn the voice back on to ‘dyn’ or choose Make Active from the Track menu.
All tracks use processing power whether they have content on them or not. Hiding a track doesn’t reduce the resources it consumes by very much, but deleting unused or unwanted tracks can free up useful amounts of resources. You may want to Save a Copy of the Session before you start deleting tracks, so you can go back to them if you change your mind later.
To save another chunk of processor resources, you can turn off the sends view metering that normally displays when you show an individual send view, by unticking the option in the Meters section of the Preferences Display tab.
Automation uses computer processor power, and if you don’t have the CPU Usage Limit correct, Pro Tools won’t be able to track it accurately either. You can help your Pro Tools system by thinning down your automation, especially where you have recorded fader moves from a hardware controller. Pro Tools stores all those moves in great detail, and this can often be simplified without changing the sound of the final mix. You can set the amount Pro Tools thins the automation data in the Mix tab of the Preferences window. The more you thin, the simpler but less accurate the automation will be; one option is to leave this automatic setting to thin only a little, and manually thin the data for particularly dense sections. This is done by highlighting a section of automation and choosing the appropriate option from the Edit menu; LE and M–Powered systems have a Thin Automation option, while TDM systems have an entire submenu. Take it in stages to make sure you don’t over–thin, and remember that you can always use Undo.
As discussed earlier, you can reduce the load on your hard drive by consolidating heavily edited sections into one Region. I always either duplicate the track or Playlist so I can get back to the edited version should I need to, then consolidate using the Consolidate Selection option from the Edit menu. This will produce one continuous Region for the highlighted selection.
I often try out an effect using a real–time TDM or RTAS plug–in and then use the Copy & Paste Settings option to apply the same effect off–line via the Audiosuite version of the plug–in. This also consolidates the selection you apply it to if you choose Continuous Region in the Audiosuite plug–in menu options. Again, it’s worth saving an edited and unprocessed version of the selection, perhaps by duplicating a track and turning the voice off, before applying the process.
If a software synth or sampler track is putting too much demand on your system, bounce the output of that instrument track onto an audio track and then turn off the instrument track. The deactivated track will retain all the settings, should you change your mind, but you’ll only be playing back a standard audio track.
In general, to get the best out of your Pro Tools rig it’s helpful to have as much RAM installed as your computer will take, especially if you plan to use software synths and samplers. Also, make sure you buy your memory from a reputable source, as there are a lot of dodgy RAM modules out there.
The Pro Tools CPU Usage Meter can help narrow down what is causing CPU overloads. When you see it peak, take a close look at what is going on in the Session at that point, and see if you can simplify it by thinning the automation or using Consolidate to reduce the number of edits.
Remember, though, that the CPU Usage Meter in Pro Tools LE indicates the percentage of overall CPU resources being used by the Pro Tools audio processing engine. This includes RTAS plug–ins, but does not indicate the CPU usage of Pro Tools LE as a whole. Changing the CPU Usage Limit does not change the value displayed in the meter.