Close

Page 1 of 4 123 ... LastLast
Results 1 to 10 of 34

Thread: My New Build

  1. #1

    Default My New Build

    I've been in the planning stages of building a new 64 bit SAC/SAW machine now for several months. It has been laboriously coming together in fits and starts and I finally have some results to report. The short of it is - THIS REALLY WORKS! It has exceeded my expectations so far and I have not found it's ceiling yet. I am so pleased to be able to say that.

    I built a new machine and bought new equipment to use with it so long as I was moving to 64 bit. For some years I've been using an i7-based PC with a Fireface 800, Windows XP and, of course, SAC/SAW 32. The combo has been problematic from the very beginning. I've used it a lot anyway but there have been many points along the way that I really wanted to ditch it. Or, maybe kick it - then ditch it! I haven't had those problems with my new rig and I have pushed it much further and harder than the old one.

    It mostly runs great although I still have one intermittent issue involving a conflict between SAW and my mouse that I need to resolve. In some situations, while doing heavy tracking, when I try to move my mouse it's being delayed so that nothing happens - and then the pointer apparently teleports to the other end of the screen. I can resolve it by plugging a different mouse into a different USB 2 at the same time. Somehow that causes the original to move over to use the other, non-offending, stream. Thereafter, I can even unplug the second mouse and the first still works correctly. I need a better solution than this, of course, but for now it's just a minor annoyance.

    The new hardware:

    Converters and preamps:
    RME Fireface UFX+
    Grace Design m108
    Startech Thunderbolt 3 to Thunderbolt adapter (TBT3TBTADAP)
    Motherboard:
    Gigabyte Z390 Designare
    Processor:
    Intel i5-8500 with stock cooler but Arctic Silver 5 processor grease
    RAM:
    32 GB - Corsair Vengeance LPX, 2x16, DDR4 2666 MHz
    Storage:
    Samsung 970 EVO Plus - 512 GB
    OS:
    Windows 10 LTSC

    Thank-you to all those who have given me advice, whether I took it or not.

    As this is a test build, I mounted it all in an old case I had which already had a DVD drive (handy for installing from the MB disk) and a power supply. I'm not overclocking. The i5-8500 runs stock at 3GHz.

    I chose the MB because it had Thunderbolt soldered into it - and secondarily because I like Gigabyte MBs. It may be that those MBs making use of other TB-access techniques will work just as well - but this seemed most likely not to have problems to me, so I went with it. I didn't want a 'oh - you got the kind with the exposed wire...' sort of problem.

    After setting up the RME stuff and plugging the UFX+ into the computer via Thunderbolt, and hanging the m108 off the UFX+ via ADAT, and setting up SAC and SAW - I started testing. I've done larger and larger tests over the past few days using successively smaller buffers - currently at 1-64 and 48k/24. Except for the mouse issue I haven't really had a problem. I haven't limited the number of cores nor given it instructions as to how to prioritize itself.

    My most recent test suite is 2 mics from the m108 providing input to 28 stereo input channels in SAC (thanks Uptildawn!). All of those channels have individual copies of FreeverbToo (thanks Dell!) running on them (so - 28 copies...). FOH is using up 4 output channels. In addition, I have 2 more monitor mixes using 1 stereo track each and each headed out to 2 outs (total of 8...). The 28 stereo inputs are being picked up by SAW, where they're being recorded to 28 stereo tracks. I've done test runs as long as 3 hours and 15 minutes. I've ganged-up all 28 SAC channels and then played Tarzan with the SAC controls - hopefully exercising them sufficiently - until I got bored doing it. That's been the big issue with testing so far - utter boredom. And you want that. So far that seems rock solid at about 20% processor usage.

    I figure this is equivalent to 60 mono signals running 28 VST effects plus 8 output channels while recording 56 mono signals. It's more than I'll ever use - or even have input equipment to provide. Down the road I want to re-purpose my Fireface 800 to become 8 more channels in, 4 more preamps, and 8 more out. That brings the total to 28 in and sort of explains why I doubled that to 28 stereo (to be sure...).

    I'm plenty pleased enough at 1-64, but I'm hoping it will work as well at 1-32. And I still haven't tried it with a VSTi. I ran latency monitor (thanks Warren!). It didn't give me the dropped buffers report I was looking for, but it did give me the '...suitable for tracking...' message. I should probably keep investigating that. And - I still have that mouse issue to resolve. It may be that project lasso (thanks Philip!) will resolve that (or maybe I can just change which port I plug the mouse receiver into). I need to rebuild that machine in a rack case (and buy the case...) and move it and the rest of the stuff to mobile racks. Plus there's still lots of testing to do. But otherwise this is really coming along.

    So well, in fact, that I'm going to buy some new monitors for my home studio. There's no place close that I can listen to good ones so I'm going to do a shoot-out here with 3 that I'm interested in that I'm shipping in: Focal sm9, Adam s3h, and Genelec 8341a (sometimes known as 8341 SAM) while the end-of-fiscal-year sales are still going on. I plan to return 2 sets. The timing isn't good but the sales discount is a powerful motivator.

  2. #2

    Default Re: My New Build

    I installed Process Lasso tonight - just the free version for now. I didn't do anything special with it. Just caused it to load at boot. It seems to have mostly taken care of my mouse problem. There is still an issue if I open Task Manager while I'm running the stress test. But - not when I'm mousing around SAW anymore. I guess it's too early to say that PL has fixed it for sure. But, I did re-boot in between to re-set whatever algorithm the OS is using to assign processes, and I haven't had problems in SAC or SAW, whereas I would have had them in that situation before.

    Then I closed both and re-set the buffer to 1-32 in RME and then SAC. Oddly, SAW doesn't allow me to set it any lower than 64. It must double up I suppose. Then I decided to just go for it rather than slowly working my way up, and ran my most recent stress test again. After an hour I checked and there were no apparent issues.

    I've loaded Latency Monitor as well now and I'm doing another run at 32 samples 48k/24. LM continues to conclude that I can track successfully. I wish I had a scale to compare the delays LM notes against. I think 250 microseconds is the longest anything waited. That's one 4,000th of a second, I believe, which might well be hearable. But it didn't happen to either SAW or SAC.

    So, I guess the major thing left to test is a VSTi. I haven't replaced my 32 bit ones yet. I do have the Fishman Triple Play though - a guitar to MIDI converter. It also comes with some MIDI instruments that can be applied to the output of the converter (which I've never heard because they require a 64 bit OS...). Hopefully I can test that tomorrow.

  3. #3
    Join Date
    Jul 2006
    Location
    SF Bay Area
    Posts
    1,167

    Default Re: My New Build

    Quote Originally Posted by John Ludlow View Post
    I installed Process Lasso tonight - just the free version for now. I didn't do anything special with it. Just caused it to load at boot. It seems to have mostly taken care of my mouse problem. There is still an issue if I open Task Manager while I'm running the stress test. But - not when I'm mousing around SAW anymore. I guess it's too early to say that PL has fixed it for sure. But, I did re-boot in between to re-set whatever algorithm the OS is using to assign processes, and I haven't had problems in SAC or SAW, whereas I would have had them in that situation before.

    Then I closed both and re-set the buffer to 1-32 in RME and then SAC. Oddly, SAW doesn't allow me to set it any lower than 64. It must double up I suppose. Then I decided to just go for it rather than slowly working my way up, and ran my most recent stress test again. After an hour I checked and there were no apparent issues.

    I've loaded Latency Monitor as well now and I'm doing another run at 32 samples 48k/24. LM continues to conclude that I can track successfully. I wish I had a scale to compare the delays LM notes against. I think 250 microseconds is the longest anything waited. That's one 4,000th of a second, I believe, which might well be hearable. But it didn't happen to either SAW or SAC.

    So, I guess the major thing left to test is a VSTi. I haven't replaced my 32 bit ones yet. I do have the Fishman Triple Play though - a guitar to MIDI converter. It also comes with some MIDI instruments that can be applied to the output of the converter (which I've never heard because they require a 64 bit OS...). Hopefully I can test that tomorrow.
    If you use SAC as the front end to SAW Studio you don't have to set SAW to a low latency buffer settings. SAC takes care of all of the heavy lifting.

    Its really a much more functional setup. You don't have to try and monitor through SAW or run SAW in live mode. It just takes care of recording and playback.

    Task manager will pretty much always cause buffer drops, its just the nature of the type of tool and what it needs to do in order to report on the low level status of CPU processes.
    ---------------------------------------
    Philip G.

  4. #4

    Default Re: My New Build

    Quote Originally Posted by cgrafx View Post
    If you use SAC as the front end to SAW Studio you don't have to set SAW to a low latency buffer settings. SAC takes care of all of the heavy lifting.
    Yeah, that's clearly true. I had presumed that I would need for it to match the RME declaration - but that's not the case. And, in as much as the possible settings are all multiples of 2, one need not be concerned with losing synch. I'll have to try tracking at 256, with SAC at 32, to see whether there is an advantage to the system in not having to work so hard.

    On the other hand, although I intend to increase the difficulty of the stress tests until it breaks, just to be thorough, it already runs quite a bit better than I'll ever actually need for it to. And it seems to do it effortlessly. I keep wondering whether I've missed something, made a presumption that is not going to translate to the real world. That is - I keep waiting for the other shoe. This has all just gone waaaay too easy. But, I haven't been able to come up with it if I have. As far as I can tell, SAC and SAW 64 are just monsters. Well done, Bob!

    In fact, I have Webroot loaded on this machine for situations in which I need (or want) to have access to the Internet. I had turned it off for the tests, of course. But, I accidentally neglected to shut it down again after my re-boot yesterday. So, those last 32 sample buffer tests happened with a live anti-virus running! No problem - lol. It will be interesting to see how far it can go - just to know. Very pleased.

  5. #5

    Default Re: My New Build

    Quote Originally Posted by John Ludlow View Post
    Yeah, that's clearly true. I had presumed that I would need for it to match the RME declaration - but that's not the case. And, in as much as the possible settings are all multiples of 2, one need not be concerned with losing synch. I'll have to try tracking at 256, with SAC at 32, to see whether there is an advantage to the system in not having to work so hard...
    When using the SAW/SAC link and SAC as a front end to SAW, the settings you have in SAW don't come into play. It won't matter what you set up in SAW in that case, so it's not worth testing for that specifically.
    SAW/SAC - finally retired a P4/2gHz after ten great years! Burnin' rubber in 2015 with an Intel Core2 Duo e4700/2.6GHz and the trusty RME Digiface/PCI pack.

    2013 Iowa City JazzFest sets recorded/mixed in SAW with SAC as a front-end mixer for live streaming radio!
    Dr Lonnie Smith Trio w/Jonathan Kreisberg on guitar
    Pharoah Sanders Quartet
    Fred Hersch Trio

    These are post mix sets to single camera video (by Rich Rauch).

  6. #6

    Default Re: My New Build

    I didn't get to Triple Play yesterday. Instead, I increased tracks.

    First I added 27 tracks to the first monitor (no effects). I left Webroot on. When that worked for a couple of minutes I started over and gave it about an hour. It increased the SAC 'Load' by about 5% but that was the only noticeable difference. So, I fleshed out the other monitor to 28 channels as well and when it worked for a couple of minutes I started over and ran it for around an hour. I expected for the Load to go up by another 5% - but it didn't really change. At the end of the hour it was still recording fine and playback was the usual.

    So, at this point, I had 28 stereo channels in (FOH), 28 channels in (Mon 1), 28 channels in (Mon 2), 28 copies of FreeverbToo, and altogether 8 channels out. And the only difference fleshing out the 2 monitors made was increasing Load by ~5%. So, I started adding input channels to FOH, Mon1, Mon2, with additional FreeverbToo for FOH, and tracks to SAW.

    I decided on 8 additional stereo channels in, which brought everything up to 36: 36 FOH, 36 FreeverbToo, 36 Mon1, 36 Mon2, and 36 tracks recorded. After it worked for a minute, as per usual, I started over and intended to leave it for an hour. Load in SAC spiked up to 40%. I went upstairs to watch Seth Meyers from my recliner while the test ran. I got there in the middle of a musical number that included staff rather than guests, sat down, tried to puzzle out what was going on.

    And then suddenly it was 4:30 am and my neck and knees hurt from my launch position. I remembered I had a recording going and, after getting my legs to work reliably again, reluctantly walked back downstairs to the studio. It was still going. I spoke into one microphone and it reverberated through the monitors, so SAC was unaffected by the length of time - I always do that so there will be a strong signal right at the end of the recording. I stopped the recording and played back the last few minutes.

    This time it was weird. Instead of vaguely hearing me walk into the room and then speak into the mic, there was the distorted sound of a broadcast. But, of course, there had been no source for a broadcast. You can't hear the TV from downstairs. I was confused. I went back an hour. Just the periodic sound of the air conditioner turning on and air flowing. I slowly narrowed it down to the first moment the sound had appeared - which was just a few minutes before I got there. First there was a loud crackle, with a booming low tone. Feedback between the monitors and mics? - but then why had I heard nothing in the room when I first got there? Then the severely distorted sound of what I think was a commercial of some sort with a woman singing. But, I couldn't make out which product due to all the distortion. And it was like that all the way to the end of the recording, which did not include my voice. I told SAW I wanted to start over from the beginning, waited a few seconds for it to delete 150+ gigabytes of recording, and then closed SAC and SAW.

    What was left on the screen was the Process Lasso screen. It includes a running graph which I believe shows the degree of potential interference between processes over time - higher is more potential interference. Throughout all of my testing I've noticed that there is a periodic spike in potential interference that I have presumed was SAW writing to disk from buffer - and possibly Webroot checking it to make sure it's OK. In the past it's looked like a small, but continuous, bump. I noticed that at what must have been around the same time the strange sounds started, the graph went up a lot - and then there was a gap with nothing for several minutes at least, and then the graph re-started and went back down. I don't know for a fact, but I presume that the graph re-started when I stopped the recording.

    At this time I have no explanation as to the source of what was recorded toward the end. It is as if something in the hardware fried and began to pickup a radio station. Except that would have had to pass through SAC first, and so I would have heard it through the monitors when I first went down, and I would not have heard my own reverb-ed voice through the monitors when I spoke into one mic. So, I think whatever it was must be limited to SAW (and yet that seems impossible...). The machine is not connected to the Internet - but it was running Webroot at the time. I don't have other recordings I've made on this machine yet. I haven't moved them over and I delete my test recordings as soon as I'm done to conserve disk. I had a hard time getting back to sleep after shutting it all down, walking back upstairs, and going to bed.

    No, this was not a dream sequence. I checked this morning and I was definitely there: things were shut down and turned off.

    I'm going to re-run the 4+ hour test today. I guess I'll leave Webroot running to see if I can duplicate my result. If I can I'll do it again without Webroot.

  7. #7

    Default Re: My New Build

    Well, I let it go over 5 hours this time. And, as before, I spoke directly into one of the mics shortly before I stopped the recording. My voice, with reverb, echoed out into the room - so SAC was still working. Then I backed up about a minute and started listening. The result this time sounded like super-amplified background room noise, but with voices in the background. I backed off to 3 hours and it was relatively silent. So, I zeroed in on the point in time where the aberration began. At 4:08:33 there was a painfully loud digital ripping sound, along with a covered up sound of voices, and from then on it was different. And I was over an hour away from coming back downstairs.

    I found that I could reduce the output gain in SAC to bring the result to a level that wouldn't threaten the life of my monitors. And then I listened more closely. This time there are two voices: one male, one female. It sounds a little like a broadcast interview. From the pace and inflection, the woman is asking brief questions and the man is giving long, involved, answers. But, the background is too loud to tell what they're saying. This time I've kept all 5+ hours, at least for now. I'm going to try to filter out the (louder) inflated room noise and then enhance the voices to see whether I can tell where it might be coming from.

    Oh - and I haven't been watching television or listening to the radio - or anything else for that matter. Whatever it is - is making it into a finished basement, with dirt on all sides, then concrete, then paneling, from the outside. And SAC seemingly doesn't know about it (unless maybe if I turned mic volumes up a huge amount maybe I'd hear voices too?). Possibly through house wiring?

    Or else it's the Ghost of Crowded Corner Studio...

  8. #8

    Default Re: My New Build

    Unfortunately, I presumed that SAW would save the tracks at the moment that I stopped recording it. If I had thought it through I would have realized this was not the case. Attempting to avoid saving hundreds of gigabytes of recording of the ambient sound in my studio, while the record panel was still open I made a copy of one of the tracks and then I told it I wanted to re-take, causing it to delete all 36 tracks for me. Then I opened a new recording session and a new SAC session and opened my saved track. All but the first hour and 20 minutes were gone - including the weirdness.

    I had to record it again. I reasoned that if the problem was the anti-virus I could eliminate it by turning it off - and I would be satisfied with that solution. Whereas if it still had the problem then I would know that it wasn't the anti-virus. So, I accomplished more by turning it off. I did the second recording without the anti-virus. Five hours later I talked into the mic and then stopped tracking. The short of it is that it has the same problem commencing at 4:08:34.

    I think this must be a bug in SAWStudio. SAC is not affected; and the problem happens repeatably at about the same moment in the recording, given the same tracking environment. I don't know what would happen if I only recorded 8 tracks, or had fewer reverbs, or no monitors. Maybe that would change the moment it started. Possibly it has something to do with the size of the buffer needed to store all but 1:2?:?? of over 4 hours of recording? I have virtual memory turned off because I have 32 gigs of RAM. Maybe that's involved? It just feels like a capacity problem to me.

    The reality is that I haven't recorded longer than 4 hours, in a real situation, ever. And I've never needed this many input signals, monitor tracks, and effects. And I don't have the input converters I would need to do it even if I wanted to. I happened to come across it only because I fell asleep while stress testing. This bug won't affect me. But it's there and I hereby report it's existence. If you record festivals continuously all day - beware. But, I'm going to let it go at that.

    But, it's time to increase the number of tracks in the stress test again. I'll just not fall asleep and I'll end the test at an hour or two. Oh - and I still want to determine the origin of that radio or TV program that appears on my recordings. It tells me I probably need to filter my power - and I already intend to let Furman or Black Lion do that for me on the 'production' system.
    Last edited by John Ludlow; 07-19-2019 at 11:50 PM.

  9. #9
    Join Date
    Jul 2006
    Location
    SF Bay Area
    Posts
    1,167

    Default Re: My New Build

    Its very unlikely to be a bug in SAW or SAC, but it could be a problem with a plugin.

    I had problems with Overloud TH2 when my system was left running for extended periods of time. It would create all kinds of horrific static noise.

    It also potentially sounds like a radio burst, only issue would be how its getting injected into record path signal chain.

    The system as listed should be able to record 28 tracks at a time without any issue and track data is written to disk as its being recorded, not when you hit save.
    ---------------------------------------
    Philip G.

  10. #10

    Default Re: My New Build

    The recorded files can only be limited to around 4 gig... in a live recording situation, you would need to start a new set of files to record past that limit.

    In general, this is around 4 hours max.

    You can track a timeline up just past 24 hours, but the max length of any file is around the 4 gig limit before the position pointer loops back afround to zero.

    Bob L

Posting Permissions

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