Close

Page 3 of 4 FirstFirst 1234 LastLast
Results 21 to 30 of 34

Thread: My New Build

  1. #21

    Default Re: My New Build

    No. One more.

    Today I discovered something that I'll bet many of you already know: if you left click on the "Load" text readout a new window temporarily pops up that lets you know how many slipped (not dropped) buffers have occurred since SAC last went live. I am not the sort of user who reads the manual cover to cover. Rather, when I don't already know how to do something - I search the manual looking for it. I searched for 'dropped' buffers and didn't find anything.

    I did all my previous tests with only a subjective idea of whether there were dropped buffers. Just on a whim today I searched the SAC manual for 'slipped' buffers and now I know how to measure them. This is a very useful feature.

    As a result, I felt that I needed to do at least the last 119 channel/track test again. I did and noted that I had over 300 slipped buffers during an hour recording. This was a good enough reason to try to improve. I tried nearly every permutation of the processor limiting and etc. options (but not combinations of individual failures...). Ultimately, the one that worked the best for me was to "Force Single CPU". With that set I repeated the previous 119 track test, and it lowered the slipped buffers to 85 'corrected' buffers.

    I'm not really sure how missing data can be 'corrected' - or what the difference between one that has been corrected and one that hasn't is. It seems to me that the data is either there or not - but clearly it is more complicated than that. That's a point of confusion for me. Maybe it means that a blank sample has been inserted in place of the missing one so that the stream remains in synch at least. So, in that case data is either 'there', 'not there', or null and SAC has repaired the stream such that the sample is not there rather than null.

    But, at any rate, I believe that means that there was less than 1 slipped buffer per track per hour. I'm disappointed that it's not perfect. But - on the other hand, it is a lot of simultaneous channels too (119 x 3 x 2). Plus, I've listened, but I haven't actually heard a tick - which is what is most important to me. It could be that the distribution is such that they can't be heard. But, for that, I haven't listened to every channel either. Maybe a hideous crackle is hiding somewhere in one of them - or maybe there was one in one of the monitor mix channels - which were all headed to void. It puts a bit of a wet blanket on my capacity results. Ultimately, I'm still pleased - but most especially because I will never approach 119 channels in my own recording. Otherwise maybe I would overclock the processor and switch out the stock cpu cooler for a water cooler - or just get a faster processor.

    I'm realizing now that I probably need to back off to one 64-sample buffer and try that one more time. Maybe that will give me the perfect result I'm hoping for. And maybe I need to figure out how many perfect tracks are possible for me at 1-32 as well.

    I'm really looking forward to doing something other than capacity stress tests at this point. But, it would be a shame not to nail it down. And not much time left before Monday when the monitors get here. After that I need to shift focus.
    Last edited by John Ludlow; 07-25-2019 at 06:48 AM. Reason: changed 'fixed' to 'corrected'

  2. #22
    Join Date
    Jul 2006
    Location
    SF Bay Area
    Posts
    1,509

    Default Re: My New Build

    Quote Originally Posted by John Ludlow View Post
    No. One more.

    Today I discovered something that I'll bet many of you already know: if you left click on the "Load" text readout a new window temporarily pops up that lets you know how many slipped (not dropped) buffers have occurred since SAC last went live. I am not the sort of user who reads the manual cover to cover. Rather, when I don't already know how to do something - I search the manual looking for it. I searched for 'dropped' buffers and didn't find anything.

    I did all my previous tests with only a subjective idea of whether there were dropped buffers. Just on a whim today I searched the SAC manual for 'slipped' buffers and now I know how to measure them. This is a very useful feature.
    Yes, slipped buffers and dropped buffers for your purposes are the same thing.

    They occur when the audio engine doesn't complete processing of the buffer data, before it has to fetch the next buffer.

    Changing to 1x64 will likely get you to zero slipped/dropped buffers.

    I have personally found that shifting SACs processor affinity off of CPU core 1 to also have an impact on slipped buffers, as there are a lot of OS related processes vying for time on CPU 1.

    Also, again make sure that SAC is running in real-time mode. It won't do that by itself running on Win10 even if set to real-time mode in SAC menu unless SAC is running as admin, or you force it to real-time mode in Process Lasso.

    I personally have been doing that assignment in Process Lasso so I don't have to run SAC as admin.

    In my testing under Win10 it appears that you can launch multiple versions of SAW/SAC when run as admin, that doesn't occur when run as a standard user.
    Last edited by cgrafx; 07-25-2019 at 07:02 AM.
    ---------------------------------------
    Philip G.

  3. #23

    Wink Re: My New Build

    Quote Originally Posted by cgrafx View Post
    Yes, slipped buffers and dropped buffers for your purposes are the same thing.

    They occur when the audio engine doesn't complete processing of the buffer data, before it has to fetch the next buffer.

    Changing to 1x64 will likely get you to zero slipped/dropped buffers.

    I have personally found that shifting SACs processor affinity off of CPU core 1 to also have an impact on slipped buffers, as there are a lot of OS related processes vying for time on CPU 1.

    Also, again make sure that SAC is running in real-time mode. It won't do that by itself running on Win10 even if set to real-time mode in SAC menu unless SAC is running as admin, or you force it to real-time mode in Process Lasso.

    I personally have been doing that assignment in Process Lasso so I don't have to run SAC as admin.

    In my testing under Win10 it appears that you can launch multiple versions of SAW/SAC when run as admin, that doesn't occur when run as a standard user.
    Yes, I presumed that 'slipped' and 'dropped' were synonymous - but only after the fact. When I searched for 'dropped buffers' nothing came up. And when I searched for 'buffers' there were too many to be useful. I even used Google to do a site search and came up blank. I just got lucky there at the end.

    I haven't completed the latest round of tests, but I had already tested for an hour at 64 for 119 stereo x 3 (no reverb). There was one slip right at the beginning before I even turned on the recorder that was associated with arming SAC, and seven at the end when I pressed stop in SAW. Otherwise there was just one in between where it actually matters. That can't possibly be audible. So if someone was foolish enough to hire me to record a symphony orchestra (and I was foolish enough to take the job)I'd record it at 64.

    That's good intel on how to finesse Process Lasso for SAC/SAW. My instances are, indeed, running on core 1 (like '1', not '0'). I'll give it a try. So far all I've done is turn PL on. If it makes a big difference I'll feel internal pressure to re-do several of my tests - so part of me hopes it doesn't. I'll bet I have around 40 hours in stress testing by now. I have learned some new and useful things in the process though. You know, I've been using variations on SAW for over 20 years. And there have always been things it was capable of that were of no immediate use to me.

    Like - there was a guy who used it to cue fireworks ignitions to go along with music, with different delays built in such that there was time for the rockets to get into the air at different heights before they exploded at the right times to match the music. Pretty cool! But, I don't do that and so I've never been interested in knowing exactly how it was done. It's been like that for a lot of what SAC/SAW is capable of. And, besides, I've never had a good enough memory to retain all the little tricks (left-click while holding down the shift key, sticking out your tongue, and winking with your right eye...) Bob's products are capable of. So, I mostly just keep on doing the subset I've been doing with it that I can remember and that I use all the time. And - when I truly need to do something new, then I search for it. Like, 'remove reverb on 8 tracks' is too easy for me not to just do, whereas 'remove reverb on 119 tracks' makes adding a trick to the repertoire to accomplish worthwhile (and thanks). Same with left, then right, click on a slider to zero it. Sooooo valuable if you're setting up a large numbers of channels. Not so much for two more.

    Several versions of SAC/SAW at once? That makes my propeller spin - although I'm trying to think of the use-case for that and not coming up with it. Maybe when 120 channels are not enough? At one programming shop I worked at they used to call things like that 'dog polishers'. Cool, and yet...

  4. #24
    Join Date
    Jul 2006
    Location
    SF Bay Area
    Posts
    1,509

    Default Re: My New Build

    Quote Originally Posted by John Ludlow View Post
    That's good intel on how to finesse Process Lasso for SAC/SAW. My instances are, indeed, running on core 1 (like '1', not '0'). I'll give it a try. So far all I've done is turn PL on. If it makes a big difference I'll feel internal pressure to re-do several of my tests - so part of me hopes it doesn't. I'll bet I have around 40 hours in stress testing by now. I have learned some new and useful things in the process though. You know, I've been using variations on SAW for over 20 years. And there have always been things it was capable of that were of no immediate use to me.
    My bad, I forgot cores are numbered starting at zero (0)

    I was suggesting getting SAC off the first core (not the one labelled 1)

    In a SAC/SAW system, SAW doesn't need to be moved as all the heavy lifting is being done by SAC.
    Last edited by cgrafx; 07-25-2019 at 03:26 PM.
    ---------------------------------------
    Philip G.

  5. #25

    Default Re: My New Build

    I've spent a lot of time stress testing while changing parameters in Process Lasso. It has definitely made an improvement. The two areas of change within PL that Phillip suggested, in particular, were what I ended up with. Except - that when I engaged high priority in PL, I got stuttering in my recording and hundreds of slipped buffers. On the other hand, setting it to 'Above Normal' seemed to make the biggest difference. I also caused SAC to have 'affinity' with core 4, and SAW to have affinity with core 5. I tried leaving SAW out of the changes, but I had better performance when I was particular with it as well. I tried a number of other things, and I'm still experimenting, but those have given me the best results.

    In 1-32 for 84x3 channels and 84 recorded I had 2 slipped buffers total over an hour and a half.

    I ratcheted up to 119 channels for an hour then, and I had one great run in which there were no slipped buffers at all. Like Icarus flying too close to the sun, I decided to see whether realtime could improve that excellence - with the intention of adding reverb back in stages next. But, instead, it got much worse. I backed down to high priority and that resulted in only hundreds of slipped buffers. So, having already gotten excellent results from 'above normal', I downshifted again to there. But, I've never gotten zero slipped buffers again, in spite of many tries. The lowest I've gotten since then is about 30. Still - that's over 50% improved - almost 66% actually.

    At this point I'm wondering about putting both SAC and SAW into 'Game' mode (which puts the designated process into high priority as well as makes some other, unspecified, changes). And I'll try it, but high priority hasn't worked very well for me so I have doubts.

    There are also some memory controls in PL, which I haven't found documentation for, that I'm curious about.

    Meanwhile, on the bass trap track:

    We finished all the cages last night. So, today I was all set to get started putting fabric on the itch I ordered from GiK Acoustics last Tueday for delivery today (expensive...). But, I got a notice from UPS that there was trouble with an undeliverable address. The short version is that somebody at GiK typed my address wrong. They're blaming an intern. So now the 703 will be arriving the same day as the monitors which is... not ideal. Very disappointing, but nothing can be done now. The shootout will just have to be a little shorter.

  6. #26
    Join Date
    Jul 2006
    Location
    SF Bay Area
    Posts
    1,509

    Default Re: My New Build

    You need to set SAC priority to Real-Time, not high.
    ---------------------------------------
    Philip G.

  7. #27

    Default Re: My New Build

    Quote Originally Posted by cgrafx View Post
    You need to set SAC priority to Real-Time, not high.
    Tried it. See above.

    I think I might be on to something though. I'll let you know in about an hour.

  8. #28

    Default Re: My New Build

    Phillip - I recall that you had an issue with your machine going into screen saver mode that you couldn't run down. My machine just began to do that for the first time today. Today was also the first day I got a nag screen from them. I like their software, so I think I'll buy it. But it's quite a coincidence that they happened at the same time. I have screen saver in the OS shut off and I haven't changed it lately or even examined it. I wonder whether this could be Process Lasso in one of it's modes? In the setup file one of the last options has to do with power saving. It had been 'true', I manually changed it to 'false' - although it might have nothing to do with it. I can't do a decent test with the screen saver coming on though. It screws up my dropped buffer count.

  9. #29

    Default Re: My New Build

    I figured out how to avoid Process Lasso powering things down when no-one has touched them for awhile. And I got a 3 slipped buffer run at 119 stereo x 3.

    To stop PL doing a screen saver, et. al., with PL on:
    start SAC,
    find the SAC listing in the PL list of processes
    right click on it
    select prevent sleep
    click on all three options

    My machine has stayed awake and alert ever since and lowered my slipped buffers. It will do so as long as SAC is open and running.

    Besides the above, here is the list of things I did before I got my 3 slipped buffer run:

    Both SAW and SAC set to be limited to 1 cpu.

    From PL:
    Options->IO Priorities->add SAW->critical
    Options->cpu priority->add SAC->above normal
    Options->cpu priority->add SAW->above normal
    Options->cpu affinities->add SAC->core 4
    Options->cpu affinities->add SAW->core 5

    Find SAC in the list of processes->
    Right click on SAC->More->memory->normal

    Some of the names above are incomplete but hopefully it will be obvious when you see them. That machine is downstairs without internet access. This one is upstairs without Process Laso. It could well be that some of my changes were superfluous. But... it works.

    This is good enough. I'm really going to stop testing now. I have a 28 input system that my new machine is way more than capable of serving.

    Phillip, I'd like to thank you for sharing your results way back when. Your kindness in doing so is the reason I began to feel confident that SAC/SAW 64 worked with a setup as large as 24 channels. Also, thank you for steering me to Process Lasso, which has made a large difference in my results, and Windows 10 LTSC, which has also made a big difference. And, finally, for all your suggestions in the meantime.

    Uptildawn, I'd like to thank you as well. It was kind of you to go out of your way to give me tips and set me straight.

    On the shootout front - the 703 arrived after all! I was surprised to find it leaning up against my garage door when I got back from lunch today. There has been a lot of scurrying to put together bass traps since then. I'm not sure I'll get all of them up before the monitors arrive on Monday. My friend who has been helping me re-hurt his shoulder and may be out. Putting the parts together is sort of a two man job. But - some traps are done at this point and there's still another day to go. Maybe I'll make it.
    Last edited by John Ludlow; 07-28-2019 at 08:48 AM. Reason: Only 1 cpu in SAC and SAW

  10. #30

    Default Re: My New Build

    I thought I had this all figured out. Then I installed the Adam s3h monitors and played some music through them to loosen them up. For convenience, I created a SAW track with all the wavs from Bela Fleck - Flight of the Cosmic Hippo and ran it out through SAC. I wasn't really listening to them yet, since they weren't broken in. But then a friend came over and I took him downstairs to check them out. As they were playing, with the Adams in AES mode, I became a little critical and noticed frequent crackling right away. What? Dropped buffers?! What happened? I thought I got rid of all those!

    So, I checked SAC's slipped buffer indicator. No slipped buffers.

    I reasoned that it might be Adam's d-a converters. So, I connected the monitors analog and tested that instead. Still there. But, of course Adam digitizes the analog signal coming in - and then uses that same d-a converter going out. So, I listened with my (analog) headphones. Still crackling. Finally, in disbelief, I backed off to 1-64. Crackle gone.

    This seems to call into question the results of the last few days of my testing since I was relying upon SAC to indicate to me when slipped buffers had occurred. I'm trying to wrap my head around where the problem happened.

    Presuming for the moment that SAC is accurate for itself, where else could slipped buffers be occurring in the signal chain? What if SAC wasn't doing all the heavy lifting after all? I boosted SAW's buffers to 1-256 but that didn't help - so presumably the slipped buffers aren't happening in SAW. What does that leave? RME hardware or Totalmix, right? Or else SAC is creating slipped buffers but isn't counting them. I think it has to be one or the other.

    Bob - could you weigh in? How reliable is SACs slipped buffer count, and where do you think the ones I heard could be coming from? Or - is it something completely else?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •