| Author |
Comment | ElmerJFudd
Guru
Posts: 3653
Loc: USA
(07/31/07 02:52 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Velvet, send me your most often used GP template and let me see what my CPU looks like. But more importantly than the CPU monitor. Let's see what kind of real world performance I am getting with it. Mind you, my G5 is pre PCI-X, and I don't have a Duende.
You have my email. Just make sure you .zip before sending. AOL mangles anything it doesn't recognize.
White MacBook '09 (4gb RAM), OSX.5.8, v.1.2 mLAN tools, Logic 9, 01X, S90ES, i88x & i88x
Post Extras:
  | velvet
Guru
Posts: 648
Loc: Space, wheresoever is enough ice to skate on
(07/31/07 02:53 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Acelera, thanks for your PMs.
Of-course I tried that, but the thing is that even after I removed the PCI-firewire card (even after clean OSX installation), the CPU Load is higher than with previous version of mLAN tools.
It is exactly as I explained in my report.
and about daisy-chaining Duende: I tried that, but I did hear clicks / pops occasionally, more than when Duende was on it's own firewire bus.
01x, i88x, PowerMac QUAD G5 2.5,
Post Extras:
| velvet
Guru
Posts: 648
Loc: Space, wheresoever is enough ice to skate on
(07/31/07 03:00 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Elmer, the file has been sent. Thank you.
01x, i88x, PowerMac QUAD G5 2.5,
Post Extras:
| AndyLuvsMacs
Enthusiast
Posts: 300
Loc: London
(07/31/07 03:30 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Getting the same CPU increase here, unfortunately, though not Velvet's "Matt's streaming driver" message (like Elmer, I'm not PCI-X, and no additional PCI cards - possible factors ?).
Most of my projects end up right on the CPU edge (multiple VIs), and every single one I've tried so far with the new drivers goes over. :-(
I was hoping that Geosync's 'removal' advice might sort the problem, but if Velvet's already tried a fresh Tiger install, that would seem to rule out that possibility; except ... isn't AppleMLANAudio installed by Tiger as part of Apple's FW streaming package ? Might it still be being activated when it shouldn't under some circumstances ?
Offsets rock solid, but variance between devices on the network about the same as I was getting.
G5 dual 2.7, 2.5GB
Internal SATA only for recording, sample lib's on external FW.
OSX 10.4.8
Sample Offsets @ 44.1/24bit:
01X round trip +33
01X round trip including D/A/D +68
i88x D/A/D patched to 01X +120
i88x D/A insert tapped to 01X +61
All offsets completely consistent across reloads of GP and restarts, haven't differed once.
The new GP patching seems fine, but I've noticed GP stays busy/unresponsive (wont hide) for 10 or 15 seconds after applying, which it didn't before; not really a problem.
Hope those of us experiencing higher CPU usage can get it sorted ...
Incidentally, was it just me, or were 01Xray/mlancental and motifator all down for a few hours from 7-ish GMT this morning (monday) ?
01x, G5 Dual 2.7, OSX 10.4.11 & 10.5.6, DP5.13, SX3, NI Kmp5, Melodyne.
Post Extras:
| velvet
Guru
Posts: 648
Loc: Space, wheresoever is enough ice to skate on
(07/31/07 03:43 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Andy, I recieve the "Matt's streaming driver ...etc.." message only when I have PCI-firewire card installed inside my Mac, and mLAN rig is connected to the port on this PCI card...
But it seems that I am really not the only one who found out higher CPU load. And don't forget Chris McCormick - he did have almost the same results as me!
Andy, do you have any PCI firewire card, or can you borrow it somewhere just to clarify that the same message appears on your computer?
Well - only in case you are not afraid of doing that - it seems that the guy Matt is really jackanapes - he is comming uninvited 
01x, i88x, PowerMac QUAD G5 2.5,
Post Extras:
| Yamaha_US
Guru
Posts: 2684
Loc: Buena Park, CA
(07/31/07 05:04 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
"Incidentally, was it just me, or were 01Xray/mlancental and motifator all down for a few hours from 7-ish GMT this morning (monday) ? "
Yes, we logged in first things at about 6 am and every Keyfax site was down. They went back up here in LA around 9 am , but it seems it was another issue because other people in different areas had access to the site when we didn't.
Hopefully, issues with " Matt" will get sorted out soon and this will prove useful for everyone when the driver is widely available for download.
We would still recommend trying the removal process Geosync recommended.
Post Extras:
| chrismcc1
Guru
Posts: 1936
(07/31/07 05:14 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Just got in to Seattle and I have caught up on the posts...
I did about 6 hours of testing Monday evening. I did not try the offset test yet as I am confident that this has been resolved for most intents and purposes (I understand this is not as well resolved on multiple node Mlan setups... I use a dual node Mlan system made up of an 01X and I88x, actually three nodes if you count the Mac of course). Others will testify.
Please be sure to indicate your configurations when posting your results.
The Negatives:
1. Higher CPU consumption:
Velvet and I exchanged experiences, and I can confirm that I did notice increased CPU usage on my Dual 2.5 G5. In fact I find very much the same thing that Andy said. Most of my projects are very close to the limit of CPU performance in Cubase 4 now, and they were not before.
I use a fairly "full" GP setup with all connections from the 01X to the Mac. And the I88X feeds to the 01X so I can control the I88x inputs with the 01X's faders. Midi is connected from the O1X and I88x to the Mac although I don't use it as I have an 8x8 USB audio interface. But the fact that they are connected in the GP, means they are contributing to CPU usage.
I sent my setup to Elmer for testing on his system. We got similar results with the All4You demo, which one could consider to be good at first glance. But that demo does not use any VSTi and very few VSTs. So I don't consider the test to be conclusive with this demo. We need to get a pretty intensive project on an FTP website somewhere that uses C4 VSTi (like halion etc...) and plenty of VSTs, so we can all test the same thing.
2. Last State not remembered by the GP:
I also noticed that the GP does not remember the last state when you quit. Since the AC is no longer usable that means that every time we startup the computer, the GP can be set to autolaunch at startup, but it won't "apply" the last remembered state to the Mlan network. Really annoying if you use the same GP most of the time for all your work... I do, which is why I use a fairly complex GP setup. This really needs to be corrected quickly. As anyone else noticed this? I was surprised nobody commented on this...
I don't want to sound too negative. Yamaha have taken a great step forward. But I was hoping for a little more given the time it has taken... So here are,
The positives:
1. The new patching works great, and really is a very big improvement. Although I was disapointed with the CPU usage, I find this new features really is a great compensation. By connecting only the needed ins and outs instead of all of them (as was required before), it is possible to reduce CPU consumption and improve the buffer in the DAW quite drastically. The more connections made the higher the CPU usage. With just a few connections made, the CPU usage drops quite considerably. This means that we need to get used to using multiple GP setups for tracking and mixing. Its unfortunate that Mlan cannot support full interconnexion among several Mlan nodes without becoming a CPU hog, but now that we can connect properly, we can really make full use of the GP as a versatile Patchbay using specific setups that use less CPU for each specific task.
I didn't expect this feature to be in this release. But now that it is, I realise this should have been n°1 on our priority list. Well done Yamaha for realising this and getting it to work.
So at this time, it sounds like the new driver brings:
1. CPU improvement uncertain, and varies from system to system and greatly depends on the complexity of the GP setup and number of nodes.
2. Offset: Gone for most purposes although some doubt remains on multiple node systems.
3. GP connectivity: Works to the extent that it can compensate for the disapointment in 1. (Velvet reports redraw errors after applying the new setup...)
This release is a good step forward. I would rank it a 7-8 out of 10.
Questions: The AC is no longer usable for multiple node systems. We were always told the GP uses more CPU then the AC. Will the AC come back and be able to open GP configs in later releases?
I am away from my studio for the next three weeks and will follow and react to this thread as I connect off and on to the internet.
Good testing everyone!
PS: Hats off to Velvet for a very complete PDF report. Nice job. I did not get any of the "Matt's..." messages you got. But I did a complete uninstall of Mlan at one point to be sure nothing was polluting my system. I also used the Reset tool on my 01X and I88X as per Geosync's recommendation. Finally, I did not notice the error in the GP where it doesn't redraw the connections correctly. Did anyone find this?
Christian
G52x2.5,OS10.5.6,SX4.1,O1x,I88x
http://www.christianmccormick.com/Page_7.htm
Post Extras:
| Yamaha_US
Guru
Posts: 2684
Loc: Buena Park, CA
(07/31/07 05:50 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Chris, Thanks for the detailed info.
Some things are starting to make more sense to us here about the reports.
We know that there is a second round of CPU improvement planned along with fix to the clock slave issue.
Because the improvements were divided into two areas (Driver side and OS side) , it may depend on the situation how much CPU usage is seen.
This would account for some people showing CPU usage improvements and others not. As Chris mentioned testing the same file in different setups would be optimimal.
We will try to get some technical details which should help people with real world applications.
Post Extras:
| AndyLuvsMacs
Enthusiast
Posts: 300
Loc: London
(07/31/07 07:34 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
Well, I tried Geo's removal instructions to the letter and reinstalled, and it's made no difference to the CPU usage I'm afraid; still noticeably higher than before the new drivers.
And just for good measure, and following the time honoured tradition of kicking the television when it doesn't work, I unceremoniously dragged AppleFWAudio and AppleMLANAudio out of extensions and made them stand in the corner for a couple of boots. Yeah, well, it didn't help at all, but at least it made me feel a bit better ...
Ho hum.
Acelera seems very happy with his CPU usage, and he's Intel I think ? And Velvet, Chris and me are all Dual G5. So is this a factor ? But then Elmer's Dual G5 as well ?
?
???
01x, G5 Dual 2.7, OSX 10.4.11 & 10.5.6, DP5.13, SX3, NI Kmp5, Melodyne.
Post Extras:
| AndyLuvsMacs
Enthusiast
Posts: 300
Loc: London
(07/31/07 08:00 PM)
Edit - Reply |
Re: OSX mLAN Release Candidate Released! Post Ple
In reply to:
Andy, do you have any PCI firewire card, or can you borrow it somewhere just to clarify that the same message appears on your computer?
No, sorry Velvet, don't know anyone with a PCI FW I could borrow. I've got an Audigy 2 ZS PCI/breakout which includes FW (which I have successfully used for mLan) on my P4, but I don't think it can be made to work on a Mac ?
Chris,
In reply to:
2. Offset: Gone for most purposes although some doubt remains on multiple node systems
There's no doubt, the offset difference between 01X and i88x is almost exactly the same as it was before the new drivers, for me anyway:
OLD DRIVER loopback offsets:
01X -56
i88x (back via 01X) +29
85 sample difference.
After a restart:
01X -50
i88x (back via 01X) +35
85 sample difference.
New driver:
01X +33
i88x (back via 01X) +120
87 sample difference.
Offsets now consistent.
Direct i88x loopback including D/A/D gives +89
01X loopback including D/A/D gives +68
01x, G5 Dual 2.7, OSX 10.4.11 & 10.5.6, DP5.13, SX3, NI Kmp5, Melodyne.
Post Extras:
| | | Pages: 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | (show all) | |
