MOTUnation.com

I run two 896's Aggregate in Tiger. I run at a 256 buffer setting. When monitoring through DP and effects, I have no problems with a bunch of plugins and such. However, once I record, when I play back tracks OR have track data in the audio tracks, I started getting spikes. It only occurs when I have audio data in the tracks.

Given I have two firewire interfaces hanging off the computer, there seems to be overhead associated with that. Also, given my machine has just two gig of ram (1 gig for each processor) RAM is also an issue.

I tried running my audio from a firewire 400 and a firewire 800 interface. In both cases, I still got drop-outs. If I was using only the internal firewire ports it was worse as both the 896s and the drives were fighting for bus time.

I bought a FireWire card and moved the 896s over to that.

As you notice, I haven't mentioned the internal drives. They had been mirror RAIDed software wise (from disk utility) but last night I went through the pain of unRAIDing them. (I have two SATA internals).

Now, when I run my projects, ALL the spikes are gone and I hover around 50% CPU for this project as I should (regardless if I am streaming from disk or the 896 Interfaces).

For others using Firewire audio interfaces, this may help you.

As a reference, these are my Xbench scores for all the drives:

70 = Internal SATA Mirrored Drives RAID
108 = Internal Drive 1 DATA (NOT RAIDED - BOOT DRIVE WITH OSX)
114 = Internal Drive 2 SATA (NOT RAIDED)
80 = External Firewire 400 IDE 7200
91 = External Firewire 800 IDE 7200 Mirrored RAID Box

Now, what you do NOT see is the drive performance when the audio interface is being used. This I think is the bottle neck.

Has anyone compared using say 2 x 896 interfaces as compared to like two 2408mkIIIs using the PCI(e) cards? Is there less overhead if you are not on FireWire?

Anyways, hope some of this helps. Like I've told some of you, I really do push my system hard in a real-time sense. When I am doing my final mixing at 1024, I can go A LOT more than what I am doing now, and I am amazing at what I can do now!

Thanks for this info, Rick.

I've always had my doubts about FW performance, and it sounds like the FW card was indeed the trick. I don't have time to go into tech detail, but streaming along FW has been the culprit in my case. I moved a ton of sample libraries recently from my FW400 drives to a new internal SATA. Performance improved by about 30-40%, I'd say, and there has only been one crash since due to an unrelated issue (undo history and iSight).

I thought about going to FW800, but I may just keep my FW400 drives and use them for general back up and not for streaming. Some sort of external SATA configuration seems to be the best move-- or getting a FW card might be more cost effective with clear benefits as in your case.

Hard drives and their busses remain the biggest weakness on Macs, imho. Once these configurations are user-refined, the CPU really comes to life.

I want to at least give back what I can for what everyone else has given. That's the least I could do.

As for FW cards, it did improve the issue, but ultimately, running on the Internal SATA in this case has been far superior. I tested both a FW400 and FW800 enclosure. Now, in each case, faster drives and a higher quality external case could have been used which could be an improvement. I just don't have the excess funds sitting around to fully test everything.

In addition, if you do some reading, there is some FW800 performance issues related to the G5. It's fairly well discussed.

I too think an external SATA setup would be ideal. For now, since I am putting my extra cash into other things, I will backup to FW devices. So far, with RSYNC, things are certainly going okay.

What I really want to test is an internal card versus the firewire devices. I just don't want to go through the hassle of installing a PCI card and all the configuration now that my system is tuned. I am thinking ahead to when the Intel towers are stable, then I might make a switch then, and keep one of the 896s for live use with the powerbook.

Good advice, Rick. It really pays to plan ahead while moving forward today. Thanks again for your generosity of info.

Rick, when you say 896's, are you talking about 896HD, or the older 896 such as the one I use? Also, when you say "aggregate," do you mean that you turned the two boxes into a single "aggregate device" using Apple's Audio/MIDI Setup?

If yes on the latter question, there may be more at work here than just hard drive performance and bus speeds.


Shooshie

Hey Shooshie...

I have an 896 and an 896HD. As for aggregate, yes, I mean running them both as one device from Audio/MIDI. I tried running them both selected in DP but I still get audio clicks, and I am running them off of a Lucid for timing now that I know is setup right.

For the test of running on the internal drive and external drive, I don't see how that couldn't be a bus performance issue. The only difference in the two tests is where the audio data is stored. One was using external firewire with nothing else on the bus (because I have the 896s hooked up to a firewire card in the G5). Although I will agree, there could be other aspects at work.

I tried hooking the MOTU interfaces and audio drives in the reverse setup as well (running the MOTU off the internal interface for FW and the hard drives on the card), this ran worse.

What I have not tried yet is shutting off the drives (FW) to see if their is any polling or such anymore.

What I have done is remove the spikes. I think it is also a combination of having too little RAM as well. Having 2 Gig on a dual processor is hurting me too. I made this determination when I shut down FireFox that had been running for awhile. Even though nothing was happening in it (no self reloading pages or animations). DP started running better.

I also notice sometimes DP is spiking for no reason when playing, and when I set the buffers up then down, it goes away. Again, this could be related to how much memory was available when I loaded DP. Could also be swap related. What separates me from a lot of people is that I run all my stuff live till the end, and the record down. The reason, I change a LOT of stuff. As I get more proficient, I am sure that will change, but for now, I am a tweaker!

Wow, ok, I've got my fingers crossed that the Aggregate Device issue is solved, and I'm going to be happy if it is. Would you keep us posted on any strange behavior that may occur? If this works, there is nothing I'd rather do than create an aggregate device for my 896 and my two 424PCI boxes. You may or may not remember when hell's fury broke loose after I did the same thing a while back. Solved the spiking issue and the clicks, but suddenly DP lost track of its chunks, tracks, and windows, mixing chunks together, and operating on tracks other than the active one displayed. Butchered a few files, but I didn't really lose any information. (had backups, and this all happened before I was able to do any significant work)

So, if you are successful with this, I want to know! What version of OSX are you using?

Shooshie

Important to note that I am NOT using any PCI interfaces. In fact, issues could be caused by that as well. I wish I could afford to purchase two 2408s to test all this further, that just isn't happening any time soon!

When I added the apple internal audio to the Aggregate a few months back, my CPU spikes were significant (I was trying to use the optical in for two more ports). When I removed the internal audio, I could process a significantly larger amount of tracks. And the important thing to note here is EVEN IF THE INTERNAL AUDIO WAS NOT USED IN ANY TRACKS, NOT IN THE AUDIO BUNDLES, ONLY INCLUDED IN THE AGGREGATE DEVICE it still affected the performance of DP altogether.

In your case, if you are using a firewire device and then a PCIe device, timing issues between them each being used could cause CPU spikes. Essentially, just timing issues (bus wise). Not saying this will happen to you, but certainly keep an eye on your performance. I may end up eventually buying a pair of 2408s from a local dealer and if they don't work, just take them both back.

I remember the fury, definitely back up before you do all this. One thing I noticed was that once you create the aggregate device, older projects do seem to act a little strange. I went through this process and they seemed okay.

1. Create Aggregate Device
2. Rebooted (Just to make sure CoreAudio wasn't leaving anything strange unsetup properly)
3. Create a new project and setup the Audio Bundles.
4. Opened up an old project, went into setup and selected the new Aggregate device.
5. Loaded the new Audio Bundles.
6. Played the sequence (after making any changes necessary).
7. Went into setup and selected next higher buffer setting.
8. Played the sequence.
9. Went into setup and selected the buffer setting I wanted.
10. Played the sequence.
11. Saved the project

Every older project I put through these settings worked with no problem. No issues of any kind. I'm actually amazed with the performance after switching to the internal drive. Something is strange with the buffers and it seems switching between them recreates them and really helps some time.

I am using 10.4.5 at the moment, I have some work I have to get done by tommorow so I have not ventured to 10.4.6 yet. Not to mention, I probably wont update till I have some down time in what I am working on. (I will be going to DP5 immediately though, I need input monitoring without record NOW).

And I will keep everyone updated as to my progress.

My next task is to get Reaktor running in DP4.6 which I hope works now with Reaktor5 (I knew there were troubles in Reaktor4).

So which pci card firewire would you recommend for a motu 828 mkII?.
And i have a SATA RAID 0, do you think that i will get better performance deleting this RAID?.

I am using the Adaptec firewire card in a Dual 1.8 (with the older PCI-X slots).

And if you are running RAID0, which is striped, you should get better performance than me, stripping is supposed to do that. I am not sure if it matters if you are hardware or software raid, so that could make a difference.

Do this, download Xbench and just do the drive test on that drive and post your results.

Now, if you are running both OSX and your audio off of the same RAID0 drive, I think this will hurt you a bit. You want to stream your audio off of an independent disk if you can. But I've done both, and it is still very usuable, you just can't push it to the absolute extreme.

Well the problem with a stripped RAID in a powermac are the Hardrive slots limitation. So i can only build a RAID 0 that has to work with DP and Audio files. Cause i think that if i use this raid only for osx and dp i'll be wasting a lot of storage space.
So i have decided to rebuild my system with to independent SATA drives. One for OSX and DP and other for Audio files.
One more question, if i connect MOTU 828 mkII and an external LaCie FW HD (kontakt dedicated) to the same firewire controller, should i find problems?

Shooshie,

I‘m running OSX 10.4.5 on dual 2.75 G5 with 4G ram with PCI424(2408mkIII and1296) with Mackie DXB200 connectd via light pipe(2408mkIII) and firewire cable directly to G5 (24 in and out). I have created an Aggregate Device and have no problems with the setup (except Mackie clicking when enabling too many eq‘s on FW channels but anyway).

2 weeks ago I connected our Lacie external FW drive and started getting CPU spikes. Sometimes I got them even with just one MIDI channel and no audio. Then again, I managed to have 100+ tracks from Lacie with many plugins and no spikes..

Point being: using Aggregate devices with PCI424 and Mackie DXB200 connected with firewire, no problem. Added 2nd FW device, being Lacie external firewire drive, some spiking appeared.

New setup working. This are the steps that i followed:

- Raid 0 erased. Build two independent SATA Drives. One for OSX 10.4.6, DP4.61, and VI. Other for Project‘s Audio files.
- External LaCie USB 2.0 for Library samples.
- Motu 828 mkII conected to Powermac‘s G5 back firewire plug.
- Buffer Size: 64, Host Buffer Multiplier: 4, Work Priority: Low.
- Disk read/write size: 500, Buffer Size per voice: 333. (I checked lower settings, but i got two much spikes and clicks).

Now my DAW seems stable, i‘m finishing a project with 32 tracks, effects plugins, 3 absynth v3, 1 Battery 2, Reaktor 5, reaching 50% on CPU meter. No clicks, no spikes.
Seems that my problem was the SATA RAID 0 working with all project‘s audio files.

Awesome, yeah, I think the software RAID in either 0 or 1 with the internal slots just seems to degrade the performance of the machine. EVERY part of my G5 runs faster now I am not running a mirror. I believe it's the latency of the reads and writes that has gone away. I believe you aren't supposed to see a read hit on the mirror, just a write hit, but I know running like I am now, everything is so much smoother.

Given you are running at 64, that's pretty low. I normally run at 256, but there are issues with that for sure when I am doing live stuff.

What's the difference here running OS X's (core audio) Aggregate Device vs running DP's own Aggregate Device setup?