Re: [Jack-Devel] Can't bring F/P below 128

PrevNext  Index
DateMon, 19 Dec 2011 16:10:17 +0100
From Robin Gareus <[hidden] at gareus dot org>
ToNikos Chantziaras <[hidden] at arcor dot de>
Cc[hidden] at jackaudio dot org
In-Reply-ToNikos Chantziaras Re: [Jack-Devel] Can't bring F/P below 128
Follow-UpNikos Chantziaras Re: [Jack-Devel] Can't bring F/P below 128
On 12/19/2011 03:14 PM, Nikos Chantziaras wrote:
> On 12/19/2011 03:31 PM, Paul Davis wrote:
>> On Mon, Dec 19, 2011 at 1:41 AM, Nikos Chantziaras<[hidden]>
wrote:

[..]
>>> People expect to be able to play notes on their MIDI keyboard without
>>> audio delays on their non-specialized computers.
>>
>> people expect a lot of different things. that doesn't make their
>> expectations correct.

Amen.

> What are we arguing about here exactly?  My point was that an RT kernel
> might have some negative effects on average desktop systems, affecting
> performance and throughput. 

It not only /might/. It does. The reliability of RT kernel comes with
drawbacks. There certainly is additional overhead. `man cyclictest`..
run it.

You're BFS or vanilla kernel will both have a lower minimum latency as
well as a higher maxium latency compared to RT-linux.

> I've lost the ball by now :-P  Are you
> saying that pretty much everyone should be using an RT kernel?  Or that
> I should be rebooting to an RT kernel when I need to work with audio?

You should use (or reboot into) RT when you require absolute certainty
(Well, nothing can prevent hardware failure or PEBKAC): i.e. If you're
performing on stage or run an [expensive] recording-session.

> If yes, then I disagree.  I think you can have both a stellar desktop
> performance *and* low latencies with the same kernel (which is the case
> with BFS.)  

You can have stellar performance, but not reliability.

I tend to agree that for a /pro-sumer/ who occasionally connects her
MIDI keyboard to the PC at home, PREEMT_RT is overkill. She'll probably
be less annoyed by a handful of x-runs compared to occasional desktop
delays.

> But my issue is not a matter of vanilla kernel vs RT vs BFS.
>  It's about the on-board sound chip being quite lame latency-wise, while
> a cheap-ass, ancient PCI card I have lying around unexpectedly turns out
> to be quite the low-latency monster.

Are you really surprised by that?

The evolution of hardware in recent years has mainly been in the
Marketing and P&R department. In general, the quality -esp. onboard
devices- is going to the dogs.

2c,
robin
PrevNext  Index

1324307444.16571_0.ltw:2,a <4EEF53D9.20905 at gareus dot org>