Isadora, 2013 Mac Pro, dual GPUs and CPU isssues
jhoepffner last edited by
Sorry, I was too short about sound, my advice (at the moment)
The best way to drive properly sound with Windows is to use a good sound card (I recommend Motu or RME, Focusrite is the minimum, avoid M-audio or Behringer) with proper Asio drivers. Unfortunately, Isadora do not use Asio but use the windows sound system. You can try Asio4all but I have a mixed experience with and I disinstaled it.
Personally, I prefer to use proper sound software, mainly Ableton Live to manage all my sound and I use midi or OSC (via max4Live) to communicate. If there is sound with video, I route it from Isadora to Live. In Mac there is SoundFlower, in Windows at the moment I use Loop.be with mixed results. In Windows there is no aggregate devices, so it can be complicated, sometimes I prefer to output the sound from Isadora via the PC sound card and recapture it via my Motu and manage it in Ableton Live. There is a little loss in quality but you avoid all the windows problems concerning mixed sample and definition (48k vs 44.1k, 16bits vs 24bits) and mixed drivers. Its very difficult on windows to avoid glitch and scratch with sound. Principally avoid to use the same card with Asio drivers and DirectSound drivers, the results are completely unpredictable
On the performance I will play next Saturday in Lille, Video and light are managed from my MSI windows computer (no sound output) and the sound with my 13" MBP. All is working perfectly but there is no direct link between sound and images.
Hello Jacques and Fred, thank you for your suggestions. Now that I have looked at the results of the tests, I will definitely use a PC. A drawback to this PC is its very large size. How much slower would a PC laptop be with a GTX1080 compared to a desktop with a GTX1080Ti? I found laptops on Amazon with the GTX1080, but none with the GTX1080Ti. Also, how many video outputs can I get from a laptop with a GTX10180. I just spoke to Asus regarding their G703GI, and they tell me just one monitor.
Regarding audio, I am using a small external Behringer interface/mixer, but I also have a Focusrite that I will test. I have two Macbook pros running NI-mate in the setup, so I could also use them for audio. Jacques, are you sending OSC instructions from the PC to the Mac to control the audio? Also, how can I determine if there is “the crap installed inside gamer machines”?
NI-mate is working just fine for me with 3 Kinect XBox 360s. I spent a few hours with Touch Designer a few days ago, but I read it does not work with the Kinect Xbox 360, which is what I have. I am reluctant to use the newer Kinects because I was under the impression that there is a limitation in the cable cable lengths for that model. I am using 20m USB extensions with the Xbox 360s with no problem.
I have been running the 135 channel to PC test for 16 hours, and all is fine.
jhoepffner last edited by
Concerning video card, the difference is not so great between 1080 and 1080TI, principally video memory (11 Go vs 8 Go).
The question is more about laptop and desktop. Laptop with 1080 have a great problem of cooling, so you have the choice of ugly 17" gamer beast with huge fan output and noise, or system adapted for laptop, mainly 1070/1080 MaxQ with reduced capacity to avoid throttling.
On laptop you have also not 4 discrete output, on the better one, as my MSI, you have 1 4K HDMI, 1 4K mini-DP and 1 Thunderbolt 3 (with the possibility to plug another 4K screen). With a Matrox tripleHead2Go, I am able to play 5 HD screens + the main one on laptop (1 on HDMI, 3 on mini-DP through Matrox, 1 on HDMI through TB3 hub). I had a look to the Asus, beside the size, its quite the same connection as mine, so, potentially, you can hook the same number of displays. Materially, you are limited to 4 display but Matrox is considered as one display for the graphic card, so you can avoid this limitation.
Concerning audio, you have to test. Concerning Ni-Mate, its really demanding for the computer and I found a great delay for kinect analysis. I made last week, during a class, a test between Ni-mate and Processing for skeleton analysis and Processing was much better but for skeleton you need kinect V1…
Sometime I use OSC instruction from PC to Mac to control audio, mainly with a kinect on the PC running TouchDesigner, sending skeleton data to Isadora and Ableton Live (with Max4Live OSC reception) on the mac.
The crap are all the software installed by the hardware builder, try version of Office, Norton anti-virus, Nahimic sound enhancer, nVidia geforce experience and many other. You can reinstall windows from windows parameters, suppressing all the not windows software. The best is to find a person able to do that for you.
Concerning TD, it works perfectly with kinect Xbox 360 (I have one). My setting when I use multiple kinect is to use NUC computer with TD for each kinect, sending parameters as OSC through wired ethernet. Is very reactive and I can do what I want with the infos, in Isadora, Unity, Processing or TouchDesigner, depending of the project. Yes cable is a problem with kinect 2 because its USB3 but the precision is much better, deep measurement is on 4000 steps cs 255 and the skeleton is much pore defined.
But is your test is fine, what else?
All the best,
@bonemap Jacques, Fred and Bonemap, thanks again for your suggestions regarding the transition of my Isadora project from Mac to PC. I appreciate your help. After many weeks of trouble-shooting on various PCs, I now have a machine that is stable with a few problems that I am handling with workarounds. I recently switched to a Corsair One Pro with a GTX 1080, 4.7 GHz i7-8700K, GTX1080 and 3200 MHz DDR4 Ram, and it is performing well. The patch that was running at 11.5 fps on a 2013 Mac Pro (with D700 GPU) is now running at 28.5 fps.
A remaining problem is the patch takes about 90 seconds to open, and when saved the patch comes to a complete stop for about 90 seconds. The .izz file is 37 MB in size, contains about 2500 Isadora actors, and 200 media files. When I started this patch under OSX, it opened very quickly and saved without any interruption in the playback of the patch. Is this delay normal on a PC?
I also have a problem with the audio, which persists with the built-in sound card, an external Behringer 302USB, or a Scarlett 2i4. The problem is the Sound Player actor not working at startup. Sending a startup trigger from [Enter Scene trigger actor] to <start> in the [Sound Player actor] only works if the trigger is delayed at least 15 seconds after the patch is completely open. I determined the error is made only with this large patch. If I create a small patch with a few actors, Enter Scene trigger can trigger the start in a Sound Player actor with no problem. If the Enter Scene trigger is used with my large patch and the trigger is sent without a 15-second delay, all Sound Players become permanently disabled, including those that did not receive the enter scene trigger. This Corsair uses an MSI motherboard, but Nahimic is not installed. Do you have a solution to this audio problem? To my knowledge, all unnecessary audio software is turned off.
Fred last edited by
@dritter Hi, I am glad you managed to get that much of an increase in speed, 28fps should be enough to give a usable output. As for the time opening the patch I think this is annoying but workable. The thing you did not mention in your spec is what drives you have. I run NVMe drives for the OS (and where the patch is stored) that have very fast read and write times (4-5 times faster than a normal SSD). For media I use 4 raided SSDs which once raided have similar speeds to the NVMe boot drive. I would imagine this could speed up your patch opening times. How much ram did you put in the machine?
As for the audio I have seen similar issues with audio initialization for the sound player, for that reason I always use the sound movie player instead. I am not sure exactly how the sound player works under the hood. Maybe @mark can chime in, would any installation of codec packs like K-lite have an effect on this? I have also had the beginnings of files not play and have delays with the sound player actor. Is there features you need to use from the soundplayer that the movie sound player does not have?
Thanks Fred, I am using two drives, an internal NVMe with a read speed of 3.3 GB/s and an external SSD with a read speed of 400 MB/s. The PC has 32 GB DDR4 @ 3200 MHz. The Windows OS and a 4K Hap video are on the internal SSD and everything else is on the external SSD. On the previous PC's I was using for testing, Isadora was crashing when my media was on two different disks or when any media was stored on the internal drive. I placed the Hap video on the internal drive just 24 hours ago, so I am still checking if the patch is stable with that arrangement. I am using the Sound Player actor because I am still editing the audio files in Reaper, so it is more convenient if I can work with aiff or Wav files rather than mov. Also, some of the audio is interactive with its volume and pitch controlled by body motion, so I am assuming that the sound player actor is more responsive than the movie sound player actor. Do you know if that is correct?
Jacques, Fred and Bonemap, I have finally determined a workaround for being unable to save the large Isadora patch while editing. It is not related to having media on two different disks, as I thought, it happens when a background video is playing while I attempt a save or during an auto-save. This was an intermittent problem accompanied with one of two error messages(see attached). If I turn off the movie player and projector associated with the background video, I can save the patch. This is very strange because I have been saving a patch while a video plays to stage since I began using Isadora 12 years ago. This problem occurred on a 2017 Mac Book Pro, 2013 Mac Pro, and 3 different PCs.