Why software sequencers suck

[DrBunsen]

Timing of software sequencers against hardware sequencers

Graphs show the difference in timing tightness/sloppiness or drift of software and hardware sequencers.

[OrganGrinder]

hi DrBunsen

i disagree with your conclusion

before condeming software sequencers you should consider what you are comparing.

Firstly the hardware sequencer is a dedicated piece of hardware designed for the specific task you are testing.  The hardware sequencer would not require a fast range of peripherals such as com ports, video, keyboard/mouse, hard disk controllers, usb etc.  and if any of these are provided they are on a secondary system so as not to interfere with the hardware sequencer.

A PC (or mac or any other) is a generic computing machine where the provision of millisecond precision i/o was not a primary design specification.  in reality the pc's i/o is shared between many systems including com ports, video, keyboard/mouse, hard disk controllers, usb etc...

so in the end, your comparision is like comparing who comes first in a race between a ferarii and a mac truck.  the ferarii may be faster but you cannot carry what a mac can.

similarly, a hardware sequencer is limited in what you can do with it, if is not in the design and final product, it will not do it (expansion modules only enhance the system so far).  on the other hand a software sequencer (especially if you can modify the source code) can be customised to do virtually anything you can concieve.

looking at your results, you should be amazed that a pc can achieve the results it did and still do everything else as well.  when comparing win98 and win2000 did you use the same pc to midi interface?  also what other hardware was the win2000 machine using in addition to what the win98 machine had?  add also the fact that win2000 is much more sophisticated (complex) to win98.

in conclusion, the better of the two will depend on what you want.  the hardware for a dedicated solution with limited expansion but very good precision.  the software for customisation, versitility and good percision (it had a variance of 1-2ms).

please don't take the above personally, i just believe that your conclusions didn't take all the factors into acount.

OrganGrinder

ps.  it might be interesting to see how a real time operating system which is dedicated for use with sequencers would compare.

[DrBunsen]

These are not my tests, my results, or my conclusions :)  I'm just passing on something interesting I found on the InterSurf.

Yes, I would certainly like to see the results from other non-Windows OSes, like Mac OS X, and perhaps the dedicated audio Linux distros like dyne:bolic, agnule, demudi etc.  From other things I've seen, it seems like OS X has much tighter timing and lower latency that Windows.

I thought it was more interesting for his testing methodology.

[johnh]

It would be very informative if we knew what type of MIDI interface was used on the PC.

[stryd_one]

I wouldn't trust someone else's specs on something like this purely because the configuration required to yield results that were worth a crap would be impractical and far too costly both in hardware and manhours

(like, comparing different MIDI interfaces, from diffrent brands, on different connectors, with different driver versions, with different BIOS setups fro different motherboard manufacturers - we're takling about 1000's of different setups and that's just the basic setup of the MIDI interface. I reckon I could easily come up with 100000 different variations)

But everyone's missing the point.

Computers.....crash.

And unpredictably. Well, that's not entirely true, they do tend to be a little predictable - they crash lots when you're on stage. If my MBSeq were to crash, I could find out why and fix the source. Try doing that with windows, OSX or linux, and the hundreds of proprietry drivers you run.

Course, the difference between an MBSeq and a PC is largely the OS/software. If you wanted to write a dedicated OS and seq app in x86 ASM, I'm sure it would ROCK. But AFAIK only one person ever tried it, and the job got too big :(

[moebius]

Well,

MIDI as a slow serial protocol sucks by itself-- but surey Microsofts vision of the modern operating system makes things whole lotta worse. The task scheduler doesn't really lend itself to provide realtime I/O and so called sophistication is actually bloating the system with unnecessary background tasks and add-ons not really needed for OS to function. You can also thank this sort of thinking for the many of the "Internet Threads" of today.

Operating System for Audio/MIDI use should be optimized-- but for those defending "General purpose" OS concept: What has gone wrong, if MIDI jitter of the 300MHz PII is as bad as Current machines running 10 times faster?

Moebius