Graphics and MIDI problem - PCI latency questions

Tau · Post by **Tau** » Thu Dec 27, 2007 8:47 am

Hi!

I'm having some serious problems configuring my scope stand-alone box. I have been searching here, and I have tried mostly everything, but I still get this problem. I have to ask for your help!

I've built a computer for scope only. It's based on an AsRock P4i65g mobo, a 3Ghz PIV and 1GB DDR400 stick of RAM, with 2x15 + 1x6 DSP cards. I would prefer to have two screens, so I've installed an AGP GeCube radeon 9600XT 256MB. The problem is that, whenever I touch the GUI of any Scope plug-in, the MIDI notes start to drift - sometimes just a little, while other times a lot - it's unplayable! And it gets much worse if I have two screens and drag one interface from one to the other.

This is what I did so far:

My first install was as Standard PC, and no Hyperthreading. I always had a shared IRQ between one scope and USB, VGA or LAN, and I couldn't make the 6DSP card to come last, although I edited the cset.ini file. I could get 17 MVs - but MIDI was terrible.
I tried to use MIDI over LAN, Scope hw MIDI ports, an Emagic MT4 USB MIDI interface and lastly, a parallel port MOTU MTP/AV - no luck.
I have reinstalled XP in ACPI mode, and not only did I get rid of IRQ sharing, but the 6DSP now show as last. I haver also unistalled USB and LAN drivers, and tweaked the PCI latency with DoubleDawg (this makes it a little better, but not much). My last attempt involved swithching the ATI for a MatroxG450, turning busmastering off, disabling comb writing, but it's always the same problem at all resolutions and bit-depths.

I went through many posts here, but I need to ask for some direct help. The only thing I haven't done is to use the built-in VGA adapter - but that's single output, and I realy need 2 screens... If I use a single screen, it's all more or less normal with any Vid card.

I never had this on the other DAW, and it would do scope and all the VST/audio work too with 2 screens... Maybe this MOBO is really bad for graphics? Is anyone else using it?

Any help will be appreciated. I would like to use this holiday week to get some music finished, but I've been spending my nights re-booting the pc

And another thing - PCI latency: I can tweak this in the BIOS, it is set at 32. But I can also do it in DoubleDawg: Scopes usually start at 96, the Matrox at 64 and the ATI at 240 or so. It's still not clear to me what these values mean. For eg., should I increse or diminish the PCI timer in the BIOS? And in Double Dawg, what do higher values mean? Should the VGA have a higher or lower latency than Scope? I got the impression that the higher the value, the better hold of the PCI bus for the device, so Scope's should be set high, while the VGA should be set low - is this correct?

Thanks in advance!

T

astroman · Post by **astroman** » Thu Dec 27, 2007 12:43 pm

yes, it's more a thing of resource allocation (how long is a card allowed to control the bus), than a timing issue (Asio latency).
You're right - higher values mean longer 'time-slices' on the bus.

It's probably enough to just set the ATI to similiar values as the other cards - it would rather be counterproductive if the Scope card hog the bus, because other vital tasks would be delayed.

Sorry for my bad memory, but I have experienced the same drifting of midi notes by activity in the GUI ... I think I solved it, but I just can't remember the clue.
At least since I used a software midi port for Band in a Box - supplying sounds by Gigastudio instead of an external sound module - I didn't notice it anymore.

cheers, Tom

Tau · Post by **Tau** » Thu Dec 27, 2007 1:32 pm

Thanks for clearing that up, Tom. I'll be finetuning those PCIs tonight!

BTW, I don't have a sequencer app on the scope box, I'm using Hubi's Loopback device to send MIDI from the MOTU in port to the Creamware seq source, and I also tried with MIDI-OX and Miditrix... My luck to be able to experiment with a lot of different hard and software - Unfortunately, only one MOBO, and it's an AsRock...

I was very happy with MIDI over LAN (using molcp driver), it seemed to work very well, and it doesn't even need a harware interface. I thought the graphics problem was due to this, so I stopped using it, but I'll give it anoher try when I nail this problem. I was using UltraVNC as well to swap files between DAWs and to access eachother's desktop - my idea was to see if it is possible to run a desktop DAW with scope from a laptop, without having an extra monitor/keys/mouse - so I could take a rackmount live without the additional cargo. It was all working very well, too, until I plugged in the 2nd monitor, and all went down the drain.

Right now, my pc doesn't have LAN or USB, for the sake of IRQ saving. It's crazy trying to get some files in there

I'm back to the drawing board.

Cheers!

T

P.S.: Tom, I just noticed what you said about the soft outs, and coming to think of it, I have always had trouble with Scope when I use hw MIDI ports. It happened with Magma, but I thought it was because I was using the hw ports on Scope cards, and then I never really checked it out again. On the other DAW, MIDI would be handled by Cubase or Ableton Live, therefore only soft MIDI outs were used into scope, and it went well. But I remember that, once, I "hardwired" my Bitstream controller USB MIDI out to the Scope Seq in via Hubi, and I noticed a few glitches on the timing...

So a new picture comes to mind, now. If cubase is "playing" MIDI into Scope, I have no problems, but if Hubi's receiving incoming MIDI in realtime and sending it to Scope, or if Scope hw MIDI in receives the data, I cannot touch the GUi without severe disturbances to the timing...

Hummm...

Is there a difference in MIDI handling between Hubi's and Cubase, for example? Or is the fundametal difference in the "realtime" nature of one signal vs. the latency-compensated MIDI streaming of the other? and how do graphics fit into this equation?

Could it be that, when an ASIO app is running, MIDI is handled with a higher priority and stability? I'm only running Hubi's and SFP, so there's no ASIO, for reduced latency, I thought...

Obviously, I'll have to give it some thought. I'll make some more tests tonight (Nothing else I can do, as my projects won't play like this), and I'll bring back some more news tomorrow. in the meantime, if any of you has any ideas on how to solve this, please share them here.

Thanks again, and excuse my thinking aloud!

T

Post by **hubird** » Thu Dec 27, 2007 2:01 pm

no help, yet good luck wished to you

Tau · Post by **Tau** » Thu Dec 27, 2007 3:21 pm

Thanks Hubird!

I got some of that luck straight away! I seem to have found it, but it is weird.

I added ASIO modules in SFP (ASIO1flt, just for the record), and installed Reaper on the Scope DAW. I then created a single MIDI track routing the parallel MOTU MIDI in to the Creamware MIDI out, set it to record, enable monitoring and... it is solved. I can have two monitors, drag windows over and across and no MIDI jumping!!!

It seems that, if there is an ASIO application running, MIDI is stable. Thank you AstroTom for the hint. And I have to try this on the Magmas too...

But this is something I never heard about... I am happy, but I am puzzled

I wish I only had to open sfp, though. It's a bit of a bummer to have to load another app, but hey, the CPU load should be minimal, since it's not doing anything, and the DSPs should take care of most stuff audio. Still, it's not a perfect solution.

Thanks for the help, anyway!

T

Post by **hubird** » Thu Dec 27, 2007 3:49 pm

that's good news!
Try also the ASIO2 modules, ASIO2 has a more complete implementation (it offers direct-monitoring and better sync if I'm right).
hope you have some holyday time left