Baffling Loss of Performance!

[Sean McCoy]

Help! I finished a mondo 5.1 film mix for DVD in January and am now doing the touchup mixes for a theatrical release. This project, which uses nearly every track and return plus six Buss Extensions in XY panning, always pushed my system to its limits, but now it's virtually unworkable. Absolutely nothing has changed in the EDL with regards to track count or CPU-hungry FX, yet I'm seeing high MT Loads in sparse areas of the mix and quick spikes to buffer underruns in places that defy logic.

I've done everything I can think of to reduce the load, though because of the dynamic 5.1 panning I can't print as many effected tracks as I'd like. Of course I've altered the buffers ad nauseum, tried single CPU on and off, tried switching to the ASIO driver, and even reinstalled SS 4.0c in case it had something to do with 4.1. No change. Certain plugins, particularly Waves, seem to be particularly hungry, yet I still see bizarre CPU loads even after completely removing many of them. Much of the demanding reverb is being handled by the UAD-1, which is never running at a high load.

What could possibly have changed in my system to cause this drop in performance? I'm out of ideas, so I'm open to any suggestions***8212;but my system already has all of Bob's tweaks applied and, as I said, I was able to get through this project much better a few months back.

[Bob L]

If the data is on an external drive... perhaps move it to an internal drive and see if that helps performance.

Also... perhaps there are some sector issues cropping up on the current drive and moving the data may eliminate them, allowing disk bottlenecks to be bypassed.

Also... if you are using MME drivers... perhaps a switch to ASIO may improve the performance with a maxed out heavy loaded session... MME requires lots of extra background threads and adds a lot of overhead to the system in the background... perhaps ASIO may give back enough of the cpu to get your session playing smoothly.

Are you using Region Memory caching? If so... try without... if not... try with... see if there is a difference.

Of course you should be able to build the final mix without regard for cpu load... and if you can get at least a min or two of playback without stopping, then you should be able to make session tweaks in sections... then build the entire mix.... even if it will not playback from start to finish.

Bob L

[AcousticGlue]

Use perfmon and find out what is happening. Check processes, check disk queue lengths. Make sure nothing has changed on system. Checking disks for fragmentation and queue lengths might shed some light. Are you running a pagefile?

[Sean McCoy]

Use perfmon and find out what is happening. Check processes, check disk queue lengths. Make sure nothing has changed on system. Checking disks for fragmentation and queue lengths might shed some light. Are you running a pagefile?

I've run the Performance Monitor, and see nothing drawing heavy resources except SS and the always baffling "System Idle." The only changes in the system (that I remember) are that I replaced the old 120 Gb data drive with a 500 Gb drive, and added a third UAD-1 card. I may pull that card to see if anything changes. The data drive has lots of free space and isn't fragmented. I do have a pagefile. Should I get rid of it?

[DominicPerry]

There's a good chance that the 500GB drive doesn't have the seek time performance of the 120GB drive. If you have a lot of memory (2GB or more) it's worth trying without a page file. It's also worth turning the page file off, defrag'ing the system drive and then turning it back on - this will give yuo a contiguous page file instead of one which is scattered all over the place.

Your new UAD-1 could be sharing an interrupt with something important - like the sound card or the hard disk controller.
That would definitely cause an issue if you were right up against the performance limit of the machine.

Dominic

[MMP]

Is the third UAD sharing an IRQ with anything other than the rest of the UADs? If so, that could very well be the problem. I found the best way to run Uads is in a large Magma where they can all share one Irq, and to arrange it so nothing else is sharing with them.

Good luck!

Regards,

MM

[Sean McCoy]

Thanks, Dominic and Michael. I'll definitely try taking out the UAD-1 card. It occupies a spot that used to have a FireWire card in it, which I was able to lose because of the Asus' on-board FireWire. Perhaps the IRQ configuration shuffled itself around in a bad way. I'll also experiment with the pagefile. The system drive is somewhat fragmented, so that could be an issue. As I have 3 Gig of RAM, I should be able to operate without it. I doubt the hard drive speed is an issue, as the project rarely has a super heavy simultaneous track count, and spikes are happening even in very sparse areas of activity. (though stranger things have happened!)

Interestingly, if I leave the system inactive for a few minutes after it overloads, it frequently plays through the same section with less load on the first couple of playbacks. It's almost as if the CPU is getting gradually taxed and taking a long time to recover.

[Pedro Itriago]

Hmmmm, sounds like cpu overheating

[Sean McCoy]

Hmmmm, sounds like cpu overheating

Ah, that's an interesting thought, as my A/C just went out today while we're in the middle of what will be record May temperatures. I suppose it's possible the system has been getting less efficient at cooling my machine room recently as it approached its current state of complete non-functionality.

[Ian Alexander]

The other replies seem more likely, but have you done the arithmetic on the power supply? A larger drive may not draw much or any more power than the older drive, but does the additional UAD take power from the computer supply?