Now that we are going to be having multi core CPUs up to 16 to 18 cores I would imagine if mixbus would be able to run more efficiently. Will and can mixbus be more optimized for these next generation processors
Sent from my iPhone using Tapatalk
Expensive CPU, expensive water cooling. How many people will buy in?
(06-14-2017, 09:29 PM)retro Wrote: [ -> ]Expensive CPU, expensive water cooling. How many people will buy in?
The water cooling will be temporary I think. They will come up with a better cooling plan soon, I am certain, because you are right. I do not want to fsck with water cooling. Been there, done that. No thanks!
In my experience the CPU has not been the limiting factor since at least 5 years. System-boards, peripherals, drivers constrain rt-performance.
Anyway, Mixbus can use multiple cores already. In general the schema is: process all tracks in parallel and process all mixbusses in parallel.
It won't scale linearly with number of cores though: more cores more overhead, particularly for low-latency apps (It can even go backwards -- but for time-slices in the low ms range it probably won't).
(06-15-2017, 06:13 AM)x42 Wrote: [ -> ]In my experience the CPU has not been the limiting factor since at least 5 years. System-boards, peripherals, drivers constrain rt-performance.
Anyway, Mixbus can use multiple cores already. In general the schema is: process all tracks in parallel and process all mixbusses in parallel.
It won't scale linearly with number of cores though: more cores more overhead, particularly for low-latency apps (It can even go backwards -- but for time-slices in the low ms range it probably won't).
So is it safe to say that CPU power nowadays will be no real benefit then what's already available
Sent from my iPhone using Tapatalk
I would not generalize it that way.
It depends what you do. FX and synths become more complex and more CPU hungry, particularly physical modeling. If you run many instances of such plugins more CPU cores will help.
Also if you're interested in fast exports of long material (e.g. faster than realtime export of a 4 hour show), more raw CPU power is beneficial.
Those multicore high-end CPUs are mainly for the benefit of crunching huge amounts of data and are optimized for bandwidth. Could be games, video or neural networks. Audio on the other hand is not a lot of data: 187 kB/sec for a mono stream at 48KSPS and throughput is important. Most of todays quad-cores are more than sufficient to process 100+ tracks with Mixbus. As long as Mixbus isn't CPU limited, upgrading the CPU won't make any difference. So if you record+mix and don't hit CPU limits, your money may be better spent for on a good mic or speakers for example.
Check your system's CPU load meter if any cores max-out. Also compare it with Mixbus' DSP load indicator in the status-bar (which displays worst-case of the last seconds, very conservative, not averaged).
and also experiment Preferences > General > DSP CPU Utilization, by default Mixbus reserves one core for GUI work, but PCs are complex systems, only way to know what's best is to test. On a 12 Core machine Mixbus can use them to process all 12 Mixbuses in parallel and those cores may still idle 80% of the time. If you haven't seen it:
https://www.youtube.com/watch?v=GUsLLEkswzE is a good intro (although somewhat long).
Anyway to answer your question: Mixbus is already multi-core ready, there may well be tweaks to streamline some details down the road but no magic bullet. The most benefit will probably come from compilers supporting some CPUs specifics and the CPU itself.
(06-15-2017, 07:48 PM)x42 Wrote: [ -> ]I would not generalize it that way.
It depends what you do. FX and synths become more complex and more CPU hungry, particularly physical modeling. If you run many instances of such plugins more CPU cores will help.
Also if you're interested in fast exports of long material (e.g. faster than realtime export of a 4 hour show), more raw CPU power is beneficial.
Those multicore high-end CPUs are mainly for the benefit of crunching huge amounts of data and are optimized for bandwidth. Could be games, video or neural networks. Audio on the other hand is not a lot of data: 187 kB/sec for a mono stream at 48KSPS and throughput is important. Most of todays quad-cores are more than sufficient to process 100+ tracks with Mixbus. As long as Mixbus isn't CPU limited, upgrading the CPU won't make any difference. So if you record+mix and don't hit CPU limits, your money may be better spent for on a good mic or speakers for example.
Check your system's CPU load meter if any cores max-out. Also compare it with Mixbus' DSP load indicator in the status-bar (which displays worst-case of the last seconds, very conservative, not averaged).
and also experiment Preferences > General > DSP CPU Utilization, by default Mixbus reserves one core for GUI work, but PCs are complex systems, only way to know what's best is to test. On a 12 Core machine Mixbus can use them to process all 12 Mixbuses in parallel and those cores may still idle 80% of the time. If you haven't seen it: https://www.youtube.com/watch?v=GUsLLEkswzE is a good intro (although somewhat long).
Anyway to answer your question: Mixbus is already multi-core ready, there may well be tweaks to streamline some details down the road but no magic bullet. The most benefit will probably come from compilers supporting some CPUs specifics and the CPU itself.
I agree for video or 3D rendering the more cores the better I guess with CPU and mixbus a high end CPU might not matter I guess unless I work with a lot of virtual instruments which I don't, I guess other things are more important like having a fast ssd drive and enough ram etc
Sent from my iPhone using Tapatalk
The ryzen range is presented as expensive here, but aren't they really quite affordable? The Ryzen CPU's seem to be outperforming far more expensive Intel cpus in multithreaded tasks. Would mixbus be characterised as a multithreaded task or does it benefit more from high single core clock speed?
In my experience, your performance is more likely determined by the realtime-reliability of your system, rather than the CPU itself. The video that x42 posted explains this further.
Issues like PCI/bus sharing, memory access speed, memory sharing across cores, motherboard interrupt architecture, BIOS/firmware settings & bugs, and of course your OS's tuning and realtime performance are pretty hard, if not impossible, for users to determine before they buy a computer.
For our high-end console systems we sometimes have to test 10 or more motherboards before one is found that can maximize the realtime performance of the CPU.
Presumably some of the expensive "tuned for audio" computers are pre-tested to provide good performance.
This is one reason why macs are popular for audio applications: you can count on them working pretty well out-of-the-box.
-Ben
Would a high end Cpu be good for running lots of plugins. I know it may not be nesseaary for Mixbus to function but I use plugins that are often Cpu intense and have features such as oversampling. It would be nice to use them however I want. I find I’m always having to sacrifice plugins for Cpu resources on my MacBook Pro
Sent from my iPhone using Tapatalk