[SOLVED] Huge latency Snakeoil ISO - 1.0.0

[orgi70]

Snakeoil ISO - 1.0.0.
Not suitable for high-quality playback now.
Huge latency MPD.Average about 2000 .For comparison TinyMPD not more than 10 latency!!!


Snakeoil ISO - 1.0.0.I even know why.There are constant disk access.Even when not playing a music file.To do the playback from RAM!!!

[agent_kith]

Snakeoil ISO - 1.0.0.
Not suitable for high-quality playback now.
Huge latency MPD.Average about 2000 .For comparison TinyMPD not more than 10 latency!!!


Snakeoil ISO - 1.0.0.I even know why.There are constant disk access.Even when not playing a music file.To do the playback from RAM!!!

What version of MPD are you using? If the MPD is still scanning the databases, make sure you only run cyclictest after that has finished to get a more accurate result.

Also, can you confirm if TinyMPD is showing the results as nanoseconds (ns) or microseconds (us)?. Snakeoil uses the following command line:

Code:

cyclictest -t2 -p 80 -N -n -i 10000 -l 10000 -q

Where -N is showing the numbers in nanoseconds. 
 

Code:

-N       --nsecs           print results in ns instead of us (default us)

So 2000 is 2000 ns (nanoseconds), equates to 2 us (microseconds) and that's 0.000002 s (seconds).

I don't think any CPUs right now have a latency of 20 ns? To confirm run the above command line on a machine installed with TinyMPD and see if you get 10 or 10000.

Edit: I'd add some clarification notice in the cyclictest output window that the numbers are in nanoseconds.

[orgi70]

You are right!
In tinympd is measured in microseconds.
This means that Snakeoil ISO very very good delay!!!
It seems to me that Snakeoil ISO now best by such characteristics as design,ease of use and sound!

[agent_kith]

You are right!
In tinympd is measured in microseconds.
This means that Snakeoil ISO very very good delay!!!
It seems to me that Snakeoil ISO now best by such characteristics as design,ease of use and sound!

And don't forget it's all free .

2000 ns is pretty good. The best I did with my previous setup was a max/avg/min 1 μs before I give up trying to strive for the absolute minimum latency (my current system has a latency of around 15 μs).

This is actually the second time somebody mentioned the test numbers are huge and therefore Snakeoil OS is no longer a good audiophile software. This is back with 0.1.5a when I fixed cyclictest to return nanoseconds (0.1.4 and before was returning the default microseconds by mistake). Didn't bother to respond the first time because I should have clarified the test numbers in the WebApp itself. It's not that easy to do that with the old software, but it's added now in the new Snakeoil Web App (1.0.2).

The reason for using nanoseconds as the output will be more interesting for people who are using computers with latencies hovering around the 1 μs mark, e.g. major's system here. Yes Snakeoil OS is this good we have to measure in ns.

If not wrong, the last time I read up on this the HPET timer has a resolution of 1 μs so values of 1000 to 2000 ns is about as good as the hardware limits can go - i.e. any value lower than 2000 should not be trusted (precise but not accurate). 

Precise because nanosecond gives you 3 extra decimal places. Not accurate because 2 μs is really at the lower limits of the HPET resolution and we are only running this test 10,000 times (You need to run this loop at least 100 million times to have a general idea, and 1 billion times to be accurate). Interesting numbers but ultimately IMO not useful because it will hover around the 1000 mark anyway. In a future release I will add a time history and histogram plots which will be a lot more useful. What matters is not low latency, but the distribution of each calculation. In other words, converting from a function of time to a function of frequency. Will write more I guess when I have the time to add those 2 plots.

Be mindful of bias, comparing a large number (2000) when expecting a small number (10) may introduce a bias. As an example if you see a huge number like 2000, that is already enough to make you believe Snakeoil OS is no longer good any more. The reality is nothing has changed, it's all in the mind. And there is every possibility a system with a latency of 10 μs can sound better than a latency of 2 μs). 

This is why I don't rely on measurements - always use your ears to judge how good Snakeoil OS is. Trust yourself, don't rely on anything or any one else (including me) to tell you what good sound is.

Finally, Snakeoil OS is the only software I'm aware of that allows you to tweak/tune your machine to play music the way you like it. If Snakeoil OS don't sound good, almost all the tools are already in place to fix it.

On the bright side, if you know anybody outside this forum claim Snakeoil OS is lousy because it has a high latency, know that he is a fraud and has no idea how to listen  .

[orgi70]

You are right!
In tinympd is measured in microseconds.
This means that Snakeoil ISO very very good delay!!!
It seems to me that Snakeoil ISO now best by such characteristics as design,ease of use and sound!

And don't forget it's all free .

2000 ns is pretty good. The best I did with my previous setup was a max/avg/min 1 μs before I give up trying to strive for the absolute minimum latency (my current system has a latency of around 15 μs).


On the bright side, if you know anybody outside this forum claim Snakeoil OS is lousy because it has a high latency, know that he is a fraud and has no idea how to listen  .  

Now after tweaking the MPD 0.20 when playing DSD about avg

1550 ns = 1.5 µs!





It is seen that works only 2 cores .I have a 4 core PC.Maybe 4 cores to configure Snakeoil?

2.Tell me why a hard drive in the NTFS format is now displayed like this -fuseblk?

[agent_kith]

It is seen that works only 2 cores .I have a 4 core PC.Maybe 4 cores to configure Snakeoil?

The command line uses 2 threads. The 2 threads can be run on any of the cores (e.g. both on Core 0, or on Core 1 and 3 and so on). That behaviour is non-deterministic if not wrong (depends on the CPU scheduler). Don't get too invested in this cyclictest, in future I'd have something better.
 

2.Tell me why a hard drive in the NTFS format is now displayed like this -fuseblk?

That looks like a bug.. I'd check to confirm, NTFS is always run in user space (FUSE), so that part is normal. Guessing Snakeoil looked up the wrong field and showed FUSE instead of NTFS.

[orgi70]

The name change happened after a reboot Snakeoil.
Snakeoil made this mistake.It is dangerous for the hard drive.
What needs to be done to clean up and show fuseblk to NTFS?

[agent_kith]

The name change happened after a reboot Snakeoil.
Snakeoil made this mistake.It is dangerous for the hard drive.
What needs to be done to clean up and show fuseblk to NTFS?

It's safe, the drive is mounted as NTFS and your data + hard drive is safe. The fuseblk text is merely a display issue so there's nothing to be dangerous about it. 

This display quirk will be fixed in the next patch update (1.0.3).

[orgi70]

The name change happened after a reboot Snakeoil.
Snakeoil made this mistake.It is dangerous for the hard drive.
What needs to be done to clean up and show fuseblk to NTFS?

It's safe, the drive is mounted as NTFS and your data + hard drive is safe. The fuseblk text is merely a display issue so there's nothing to be dangerous about it. 

This display quirk will be fixed in the next patch update (1.0.3).  

Thank you so much!
Snakeoil ISO - 1.0.2 the best!
Recommend!