XXHighEnd

Ultimate Audio Playback => XXHighEnd Support => Topic started by: PeterSt on November 26, 2007, 06:46:49 pm



Title: Description of Processor Core Appointment Schemes
Post by: PeterSt on November 26, 2007, 06:46:49 pm
To make good use of the Schemes, here's an attempt to make clear which Scheme does what (undemoed version only !).

No Appointment

Nothing is changed to a just booted system.
When one of the other Schemes have been active during this (boot) session, things might be different from before, because appointed other processes (i.e. by the OS) are explicitly set to "No Appointment". Chances that appointments have been performed by the OS are zero (I did not see them).

Other processes interfere with Audio playback, but the system as a whole performs best.
Note that for Audio playback not only the Audio processes are needed, but also various drivers; if they (the concerned) don't get the power they need, your Audio playback is in trouble just the same. Note that without tweaking drivers cannot be appointed. Keep this in mind for the below too.
Tweaking wasn't done.

Scheme-1

The most important Audio processes are appointed to core-2.
All of the other processes are appointed to core-1.
Nothing can interfere with the Audio processes.
If Audio processes run short of cpu power ... problem.
If supporting processes (think of driving SPDIF etc.) run short of cpu power ... problem.

For a 4-core system, the other processes are appointed to core-3 and core-4 as well.

Scheme-2

The most important Audio processes run on core-2, but can go to core-1 as well.
All of the other processes are appointed to core-1.
Theoretically, when the Audio processes run short of cpu power, they have more air now because of they can swap to core-1 as well. The disadvantage might be that at that particular moment other processes interfere.

For a 4-core system, the other processes are appointed to core-3 and core-4 as well.

Scheme-3

The most important Audio processes are appointed to core-2.
All of the other processes run on core-1, but when they are short of cpu power they can go to core-2.
When glitches, pops, etc. are supposedly caused by supporting processes running short of cpu power, now you're better of. But when this really is needed, it interferes with the Audio processes.

For a 4-core system all of the other processes will not go to core-2, but to core-3 and core-4 when needed.

Scheme-4

The Audio processes run on core-2, but can go to core-1 when needed.
All of the other processes run on core-1, but can go to core-2 when needed.
A system with crackles, glitches, pops, etc. may benefit most from this setting.
Apart from No Appointment it will also be the setting which is the most prone to have the SQ influenced by other processes.

For a 4-core system all of the other processes will not go to core-2, but to core-3 and core-4 when needed (note this is still different from Scheme-3).


Title: Re: Description of Processor Core Appointment Schemes
Post by: Marcin_gps on June 07, 2010, 06:48:53 pm
Peter, what about recommended setting for a single core CPU? I suppose I should set "No Appointment"?


Title: Re: Description of Processor Core Appointment Schemes
Post by: PeterSt on June 07, 2010, 10:27:39 pm
Marcin - It doesn't matter much because 1 cpu will be dealt with anyway. With results of ... no good ... :nea:


Title: Re: Description of Processor Core Appointment Schemes
Post by: Marcin_gps on June 07, 2010, 10:42:30 pm
Results are brilliant (soundwise). My only concern are those nasty spikes in CPU usage resulting in small pops. This is happening while XXHE reads data from disk (???) at the end of each track. I wish I could load entire playlist/album to memory. That would be the cure, wouldn't be?


Title: Re: Description of Processor Core Appointment Schemes
Post by: AUDIODIDAKT on June 07, 2010, 10:55:37 pm
Quote
I wish I could load entire playlist/album to memory. That would be the cure, wouldn't be?

This is called Unattended !?

But I agree on loading a complete playlist to C: at normal attended playback, should be an option.


Title: Re: Description of Processor Core Appointment Schemes
Post by: Marcin_gps on June 07, 2010, 11:47:32 pm
I play at unattended.


Title: Re: Description of Processor Core Appointment Schemes
Post by: PeterSt on June 08, 2010, 09:06:46 am
Roy - Marcin was talking about a complete *album* being in memeory;
Impossible.

Marcin - The only thing you can do with one processor core is make the Split File size setting smaller. So, this is the opposite of loading as much into memory as possible. By itself you will be observing more clicks because of it, but they will be smaller. Until - hopefully - the process becomes small enough that nothing can be heard hence nothing is the matter.
In order not to waste your time, start at the low end (12MB). If that already doesn't help, you'd have to give up. But notice the Priorities might do some job as well. The higher the Theard Prio and the lower the Player Prio the better the balance becomes for this ... until the track is loaded too late. Then you'll have a gap.

Big fun, ain't it ?
:swoon:
Peter


Title: Re: Description of Processor Core Appointment Schemes
Post by: Marcin_gps on June 08, 2010, 09:59:14 am
Roy - Marcin was talking about a complete *album* being in memeory;
Impossible.

Why not? Even with 8 or 16GB of RAM? Right now I am able to load Yello - Touch (24/48) to memory. I already played it few times and can confirm that absolutelty no pops occur :) But that's album-file. The problem is still a pain in the neck when I try to play few tracks. It would be great if you could add another option to memory management responsible for max number of tracks loaded into memory. Of course the total max size would be the border. That's the cure for all the folks having slow CPUs and not being able to play without  pops and clicks. Let's make use of that large memory modules, shall we?  :)

Now, back to the Core Appointment Schemes. They still affect SQ even on a single core CPU. The difference is very much audible with the same characteristics of sound as on a dual core processor. Will go back to this later.