Windows, motu2408mkII, express XT usb, logic, and the art of latency

[pvolcko]

I've been having a hell of a time getting my setup to work. A brief description of the system:

P4 1.7ghz
768MB DDR333 RAM
(2) 40Gig IBM 7200rpm HD's (IDE ATA100)
(1) Yamaha CD-RW (IDE)
ATI Radeon VE AGP
3Com 905C-TX
Motu 2408mkII
Motu Express XT (USB)

I have one HD and the CD-RW on the primary channel, this is the os and data "backup" drive. I have the other HD on the second channel by itself, I use it for audio recording/playback storage.

I'm currently running Win98se and I'm using Logic Audio Platinum 4.8.1.

I have a few interrelated problems with some minor "successes".

I'm using the 2408 drivers that are on the motu site and listed for Win98 only (R9 drivers I think). These have more options in the buffer size selection in the PCI-328 console app, which seems to be a necessary thing.

My "umbrella" problem was that I was fully expecting this system to be capable of recording/playing back 24 simultaneous audio tracks (raw tracks, no effects) and it is not working. Indeed I'm having trouble getting this thing to record more than one track at a time.

Mini problem 1 is that I can turn on DMA for either HD. When I do this I start getting errors from logic complaining about trouble syncing audio and midi and some wacky sampling rate being returned. Without DMA on I can only possibly record up to 6-8 tracks simultaneously. I've tried looking for updated IDE drivers for the mobo/chipset (the mobo is a MSI 645 ultra, based on the SiS645 chipset), but there doesn't appear to be anything newer than the drivers included with win98se. When I try the latest MSI/SiS supplied drivers for the SiS5513 IDE controller in this chipset I end up not being able to boot into windows and having to do asafe boot and reverting the drivers.

Mini problem 2 I have managed to be successful in getting 6 simultaneous tracks of recording, but I had to set my 2408 buffers to 4096, thus giving me bad latency, even when using cuemix. Given that this machine is really frikin fast and doing nothing but audio and midi I think that latency should be at most 10-15ms (or about 5ms or less through cuemix). If I try to drop the buffer size I start getting dropouts and ASIO overload errors. This large buffer size is necessary due to excessively long interrupt latency from what I can tell, it's 1-6ms in idle conditions (as reported by cuemix) but with HD activity it jumps to 20ms or more, usually more. I'm assuming this is because of not having DMA turned on for the drives.

My questions to the crew here...

Would moving to Windows 2000 or (god forbid) XP reduce that interrupt latency and/or give me better luck with the DMA use on the HD's? If yes, what kind of reductions? I've read whitepapers claiming that win2k would provide a solid worst case interrupt latency of around 5ms. Is this people's experience here on the forum?

Would a move to a SCSI subsystem make a difference in terms of getting that interrupt latency down during Disk I/O? I am loath to spend the $500-$1000 more on a SCSI controller and drives. I have used SCSI on my other systems for years now, but wanted to get out of it because of the added expense of the hardware involved.

Is the USB midi interface causing any of my headaches? I've tried running with it and without it and it doesn't seem to make a bit of difference. Even in regard to the audio/midi sync error described above it still reported the error when no midi data was being read in or send out of logic and the interface box was powered down.

Any and all help here is appreciated. I've searched the forums here and at other sites and done a lot of other web searching on these issues and have found nothing to give clear answers to these issues, even though they have been brought up before. Well, the SCSI switch is pretty much universally accepted to be a good move, but it isn't clear that it is a necessary move given the high speed ATA100 devices available now.
Thank you.

 

[pvolcko]

Some additional info and a bump.

I tried the setup using DMA turned on and standard virtual memory settings with the Sonar demo instead of LAP, worked with no problems at all. Recorded 24 tracks at the same time with no jumps or studders. Recorded 12 while playing back 12 and also had no problems. CPU meter didn't break 5% and DISK maxed at 15% or so. I'm rather pissed off with this development, but Sonar was my second choice for software so I guess I can go grab a competitive upgrade price deal somewhere for it instead. I hate to piss away the $450 I paid for LAP though. If anyone has any pearls of wisdom on how I can get LAP to work or if an upgrade to Windows 2000 would get LAP working well I'd appreciate it.

I note that Sonar uses WDM drivers and LAP uses the ASIO drivers for the motu 2408. Are the ASIO drivers for this interface problematic? Does LAP have problems using ASIO drivers?

 

[pvolcko]

In my searching for a solution to this problem I came across many posts here and at other places that mentioned similar symptoms, but almost never gave any kind of resolution. Many were two message threads consisting of the problem description in one message and a "Solved it..." in the next message. As this does no good for people who know how to use search functions, I thought I'd post a problem resolution.

Problem was that LAP would not record even a single track if I had DMA turned on in Windows 98se. It would error out with Midi/Audio sync problems or simply stop recording and but continue playing. The system is a P4 1.7ghz machine: 768MB RAM, 2 x 40GB HD's (7200rpm/ATA100), MSI 645Ultra mobo, Motu 2408mkII, Motu express xt (USB), 3c905CTX.

Resolution was to turn off 32bit disk access for each HD in BIOS. They needed to be set to "standard" 16bit mode for DMA to work correctly. Upon further searching, after having stumbled across this solution myself, this is a somewhat rare, but documented problem with older versions of NT 4.0 (pre-service pack 2) and even now is known to pop up here and there.

This fix allowed the machine to record/playback 24+ tracks of audio with no ASIO overload errors or the like.

Also of interest, I could not get DMA to be used in Win2K (despite having the IDE channels set to use it if available). After making this bios change DMA worked fine in win2k.

The HD's are IBM Deskstar 40Gb 60gxp models. The chipset on the mobo is SIS 645/691. I'm guessing this is a problem specific to this chipset or chipset/drive combination. Performance is stil very good even at 16bit transfers so I guess I'll score it a win and let my blood pressure lower a bit before moving on to solving problems of latency, random video wigging out, and other such nonsense.