Zynewave.
acousmodHi Frits,
I have made some tests in order to verify the huge CPU usage of Podium, which is for me a real problem as I have told you in other posts.
I wanted to know if some problems are related to my audio setup (the two Multifaces) and my computer or more dependant to Podium itself.
I have made a small project in Podium and Nuendo 2 with a comparable setup.
Podium :
– 6 tracks with 16 channels audio files (16 bits / 48 khz)
– 6 16 i/o plugins with no automation
– one plugin on the master to route the 16 channels into the two RME cards
Nuendo :
– 8 tracks with the same 16 channels audio files, but due to the bus structure of Nuendo only the 12 first can be heared
– 8 16 i/o plugins (the sames) with no automation
– 4 tracks are send to one 12 channels output bus, and 4 to a second one, for a total of 16 outputs divided on the two RME (same outputs)
The ASIO buffer is set to 2048.
Results :
In Nuendo, the CPU indicator is stable around 7 %, the HD indicator is about 10 %. The Task Manager is around 12 %, shared equally between the two “processors” of the P4 HT.
In Podium, the CPU indicator moves between 50 / 60 %, and the Task Manager shows 50 %.
Bouncing is almost impossible inside Podium, since the CPU goes from time to time to 100 % and I get some heavy clicks.
How do you analyze this difference ?
Is there any hope ? (I am sure that the answer is YES !)
Thanks !
ZynewaveCould you email me the project file?
What GHz speed is your CPU?
acousmodI’ve made another test with the sames files and plugins inside Bidule :
– 6 file players with 16 channels audio files (the same ones)
– 6 16 i/o plugins (the sames too)
– one plugin on the master to route the 16 channels into the two RME cards
– same buffer at 2048
The results are very similar to those from Nuendo :
– DSP indicator : 18 % stable
– Drive indicator : 27 %
– the Task Manager varying around 12 %
What GHz speed is your CPU?
P4 3,2 GHz Hyperthreading.
I’ve ran Podium on only “one CPU thread” and it gives the same kind of results.
I have only noticed that when I run Nuendo or Bidule, the two CPU graphs show the same use, instead of when I run Podium the second one is a little higher then the firts one (10 % ?). It doesn’t seem to be significant.
Could you email me the project file?
Done.
acousmodYet another test on my laptop Centrino 1,3 GHz with ASIO4All (Realtek).
In Podium, Nuendo and Bidule :
– 4 tracks / files players with 5.1 audio files (16 bits / 48 kHz)
– 4 heavy 5.1 plugins (multiFX)
– output to a 5.1 bus
The results are less different because of the high CPU usage taken by the plugins.
– Podium : nearly frozen, the play head is updated at about a 2 seconds rate and the CPU indicator often goes to 100 %, which causes interrupts in the sound. Windows Task Manager is 100 %.
– Bidule : the DSP indicator is about 90 % but the files are played and processed without any dropout. The Task Manager is at 100 %.
Nuendo 2 : its Performance bar is around 80 % and it plays well, the Task Manager is about 90 %.
This suggest me that, perhaps, the huge difference which is shown with the other computer and project is due to the streaming of the audio files, since the plugins which are used are not so heavy ?
It would confirm what I have told you in another post, that drop outs occurs in Podium when playing 5 or 6 16 channel audio files, with no plugin, and when the CPU was correct (about 50 %).
I have also noted that these drop outs are “sequentially” shared between the tracks, like if a 500 ms hole was jumping from one track to another..
Perhaps this will help you to find the cause ?
ZynewaveThanks for the reports Jean-Marc,
I’ll start looking at optimizing the disk streaming in the near future.
acousmodHi Frits,
Sorry to repost on this subject, but it is actually for me rather impossible to bounce my work.
When Podium bounces, it takes a lot of more CPU power, and the passages that I can read but are about 80 % become completely crackly with intermitent aleatory silences on tracks when I try to bounce.
I’ve intended to make partial bounces for some parts only. But for some the bounce is only clean with 2 or 3 tracks (16 channels). Due to the complexity and the duration of the project it is very complicated to do it.
The limit to 2 Gb wave files is also a constraint.
I’ve tried also to process some individual 16 channels sequences with EnergyXT as a plugin to host several stereo plugins in parallel (I didn’t succeded to do it right inside Podium).
When I play, the CPU indicator shows an average of 50 %, but when I bounce it goes to 90 / 100 % with crakles. So it is impossible to record.
I’ve made the same config inside Bidule : 32 % when play and 34 % when I record with its internal multichannel file recorder.
I don’t understant why Podium takes so much CPU while bouncing, where others takes nearly nothing.
Without an offline bounce option (which will be the final solution ?), I dont’ see how I will end my work…
I know that you are working on audio optimizations and that you always do your best, and I am sorry to bore you with my little personal problems…
Have you see yourself, or other people, this increase of resources when bouncing ?
If yes, do you think that the bounce process have a chance to be improved ?
Or is there another solution, workaround ?
The support for Hyperthreading has not change anything in this domain for me.
Thank you !
Jean-Marc
ZynewaveHi Jean-Marc,
Sorry I haven’t worked on optimizing disk streaming yet. Just a quick suggestion: Maybe you can optimize the bouncing process if you solo the track that you are bounce recording.
I’ll hopefully start work on offline rendering soon.
super_crunchy@Zynewave wrote:
Just a quick suggestion: Maybe you can optimize the bouncing process if you solo the track that you are bounce recording.
this is what I always do when bouncing – it’s good practice to do so, no matter what host you use, in my opinion
Soloing tracks then bouncing them is working great for me. Whenever I start to run out of CPU I bounce the instruments in the arrangement which are unlikely to change, and I always keep the original midi tracks (although muted) in case I need to revisit sequences. I don’t even care about offline bouncing because the current bounce works fine for my needs 8) But I understand we all work differently 😉
acousmodJust a quick suggestion: Maybe you can optimize the bouncing process if you solo the track that you are bounce recording.
Yes, of course ! I always mute all the tracks that are not in use for the bounce.
(the exemple that I have done with EnergyXT as plugin was made with only one track…)
But my actual problem is to bounce the tracks of my project.
Even with two or three active tracks, there is sometimes drop outs during bounces (with a correct average CPU).
To mix all the tracks I might make several series of partial bounces, then make new partial bounces with them to be able to finaly mix three tracks… And all of, this section by section (limit of 2 Gb files).
I had a hope by disabling the card outputs in the preferences, I thought that this way Podium could bounce without auditionning to obtain better performances, but they are not recorded in the bounce… No chance.
The only solution I think that is possible for me, is to use Bidule as a VST to record only two partial mixes of the total length in separate mono files (each less than 2 Gb), and after to synchronize and mix the 32 mono files in Podium (ah, a MIDI external sync would be helpful…).
I’ll hopefully start work on offline rendering soon.
This is a really good new…
Nevertheless, I hope that you will have some time also to look at the “streaming issue” for normal playback.
I have tried all sort of buffer settings, but it doesn’t seem to change anything.
It is logical : it only changes the CPU spikes and crakles, not the “random holes” in the tracks.
I also defragment the drive very often, but I’ve not noticed a difference between before and after.
Thanks !
acousmodHi Frits,
Any chance for the offline bounce feature in November ?
I have not succeded to bounce my project in Bidule since there are always some parts that Podium can’t read properly through the 35′ duration of the project.
The offline bounce is now my last hope.
It will be a disaster for me if I can’t finish my work in time…
Sorry to bore you with this 🙁
Jean-Marc
ZynewaveAny chance for the offline bounce feature in November ?
Offline bounce is one of those features I hope to start on in January. There is a better chance that I may be able to speed-optimize disk streaming in November. Would that make it possible to work with your project?
acousmodThere is a better chance that I may be able to speed-optimize disk streaming in November. Would that make it possible to work with your project?
Yes, I think that it will be enough, since there are only two or three passages that have the problem in playing.
I will be very greatfull !
Thanks.
ZynewaveHi Jean-Marc,
Podium 1.43 is just released, with CPU optimizations of the audio streaming and processing in the engine. Note that this does not cover optimizations of disk streaming. With the help of CPU profiling tools I found that there were things in the way that audio is routed and mixed in the engine that could be optimized. If you only have a few tracks and plugins you won’t notice much improvement, but if you have many audio channels (like you have with the multi-channel files), the performance gain can be 30%. I tested this in an arrangement with only one stereo wave file phantom-copied onto 200 tracks (equals 400 channels) and using no plugins.
I’m interested in hearing whether these optimizations have improved the performance of your arrangements.
Frits