Audirvana Plus 2.5.1 with PCM->DSD Conversion

[OldBigEars]

Just to clarify on the 250-260% figure you see in activity monitor. The 100% value means a full utilization of a single CPU core. Our modern CPUs have multiple cores, and thus a value above 100% means that the computation is spanned accross multiple cores to make better use of the CPU hardware.

In this case, this means the equivalent of 2.5 cores of the CPU are used.

As you can see in the Activity Monitor screenshot below, the total use of my CPU (Core i7 2.3GHz) is only of 37% while upsampling to DSD256:

[ATTACH=CONFIG]25763[/ATTACH]

 

Damien, thanks for your information (as well as the incredible development work you put into A+. I remain a huge fan)

 

It sounds as if you really need a 4-core processor to use the DSD function. I have a 2012 MBP - its 2.5 GHz i5 processor has 2 cores. Therefore 250% means both cores are running well over 100%. I need to understand clearly what this kind of usage would do to my computer over many hours.

 

Is there anything else I might have missed in the settings that would explain why I am getting such high usage readings?

[Jud]

do you mean the graph showing (almost) no roll off before 20K means nothing ?

 

with a steepness of 5 looks like a cutoff of 1.14 is required to get no roll off before 20 and generates a lot of aliasing : guess I don't read right or there is more to the picture... could you please elaborate ?[ATTACH=CONFIG]25761[/ATTACH]

 

No, I'm not saying the graph of your settings means nothing. I am saying if I tried to use *your* cutoff of .95 with *my* steepness of 5, it would result in a high frequency cut that would begin at a lower frequency than yours does. Your cutoff works fine with your steepness. I don't use your cutoff because I want less steepness, since that causes less ringing.

 

Hope I have explained this understandably.

[Jud]

Is there a Linux version of this software?

 

No, though OS X is an offshoot of various BSD operating systems, which are closely related to Unix (as is Linux).

[mansr]

It sounds as if you really need a 4-core processor to use the DSD function. I have a 2012 MBP - its 2.5 GHz i5 processor has 2 cores. Therefore 250% means both cores are running well over 100%.

 

That's impossible. Your CPU probably has hyperthreading, so the software sees four virtual cores. This gives a bit more performance than two single-threaded cores, usually 20-30%. It also mean that total CPU utilisation figures are hard to interpret since they are calculated as if all cores were separate.

 

I need to understand clearly what this kind of usage would do to my computer over many hours.

 

The CPU itself should be fine. If the temperature inside the chassis rises significantly, it can accelerate the ageing of various components, especially electrolytic capacitors. The power supply is usually the first to fail in such situations.

[Jud]

Damien, thanks for your information (as well as the incredible development work you put into A+. I remain a huge fan)

 

It sounds as if you really need a 4-core processor to use the DSD function. I have a 2012 MBP - its 2.5 GHz i5 processor has 2 cores. Therefore 250% means both cores are running well over 100%. I need to understand clearly what this kind of usage would do to my computer over many hours.

 

Is there anything else I might have missed in the settings that would explain why I am getting such high usage readings?

 

The i5 in your computer has two real and two virtual cores, for a total of four. 250% means these four cores are running slightly less than 65% each.

 

Edit: I see mansr has already answered. Note also your fan. If you aren't getting a lot of fan noise, your computer isn't getting overly warm. If you *are* getting a lot of fan noise, I would expect that would make listening less than ideal, irrespective of how well the DSD conversion is working.

[Jud]

Any chance to get "replay gain" in upcoming betas or 2.5.1?

 

I believe Damien may have mentioned it was coming in 2.6. Can anyone confirm?

[MrTHX]

I'm also using and i5 mini ... Is here anything further that can be done to 'assist' keeping the temps down?

I'm using 16gb of RAM but leave Audiravana to set how much to use. Would an SSD help a little?

 

Can convert to DSD64 without much trouble but any higher and as said earlier the fan noise becomes a problem.

[OldBigEars]

That's impossible. Your CPU probably has hyperthreading, so the software sees four virtual cores. This gives a bit more performance than two single-threaded cores, usually 20-30%. It also mean that total CPU utilisation figures are hard to interpret since they are calculated as if all cores were separate.

 

Sorry mansr, you lost me with that one :-)

[OldBigEars]

Note also your fan. If you aren't getting a lot of fan noise, your computer isn't getting overly warm. If you *are* getting a lot of fan noise, I would expect that would make listening less than ideal, irrespective of how well the DSD conversion is working.

 

Thanks Jud - this is encouraging. My computer gets quite warm, but not so hot that it's uncomfortable to touch. I haven't noticed the fan working extra hard or making a noise.

[jeeb]

Long-time lurker and infrequent poster here. First of all, thanks to Damien for this great enhancement to a great product, one that I enjoy using almost every day.

 

I'm been trying this on two systems. In the music room, I have a mid-2011 Mac Mini (2.3GHz Intel Core i5, 16GB RAM, 240GB SSD; music is on a 6TB LaCie RAID array), connected via USB to the D3 DAC board in a Simaudio Moon Neo 340i integrated amp. For headphone listening, I have a Chord Mojo connected via USB to an early 2015 MacBook Pro (2.9GHz Intel Core i5, 8GB RAM, 512GB SSD).

 

My experience on both systems is much the same as others on this thread. On both systems when a track begins playing, CPU usage will shoot into the 240-260% range. After several minutes (it can vary from three to six), CPU usage settles down to a little over 2% on the Mac Mini, and a little under 2% on the MBP. The five-second delay I experienced with early builds is no longer an issue with version 2.5.0.9.

 

On the Mac Mini, fan noise as the system heats up is quite audible, enough to break the spell. This is reason enough for me to consider going back to PCM upconversion, or perhaps some other solution: getting a long USB cable and moving the Mini to a location where the noise is less bothersome, or when funds permit, upgrading to a Core i7-based system (assuming that would keep CPU usage at a level where fan noise isn't an issue. There's one occasional artifact: occasional, I'll get a weird stuttering, where the system overlays a portion of the music that played a few seconds ago onto the current portion: it's as if my DAC is composing a canon on the fly.

 

On the MBP, fan noise isn't an issue. The main problem is that I get momentary dropouts, which again break the spell.

[OldBigEars]

On both systems when a track begins playing, CPU usage will shoot into the 240-260% range. After several minutes (it can vary from three to six), CPU usage settles down to a little over 2% on the Mac Mini, and a little under 2% on the MBP.

Jeeb, my computer cycles. It peaks at 250-260 for a few minutes, then drops like yours for a few minutes, then goes back up again. etc etc.

 

Yours doesn't climb back up again?

[jeeb]

On both systems when a track begins playing, CPU usage will shoot into the 240-260% range. After several minutes (it can vary from three to six), CPU usage settles down to a little over 2% on the Mac Mini, and a little under 2% on the MBP.

Jeeb, my computer cycles. It peaks at 250-260 for a few minutes, then drops like yours for a few minutes, then goes back up again. etc etc.

 

Yours doesn't climb back up again?

 

My omission: yes both systems cycle. CPU usage will shoot up after several minutes on the Mac Mini, often when a new track is loaded. The MBP goes longer before shooting back up.

 

Quick question: any opinions as to optimum memory allocation for pre-loading tracks on a 16GB and 8GB system?

[tipunch]

Just to clarify on the 250-260% figure you see in activity monitor. The 100% value means a full utilization of a single CPU core. Our modern CPUs have multiple cores, and thus a value above 100% means that the computation is spanned accross multiple cores to make better use of the CPU hardware.

In this case, this means the equivalent of 2.5 cores of the CPU are used.

As you can see in the Activity Monitor screenshot below, the total use of my CPU (Core i7 2.3GHz) is only of 37% while upsampling to DSD256:

[ATTACH=CONFIG]25763[/ATTACH]

 

Hello,

 

Is the activity of the CPU normal for a 2012 Mac Mini Core i7 2,6 GHz 4 Cores in DSD128?

 

 

 

It seems to me a little lot compared with Damien (Core i7 2.3GHz Dual Cores in DSD256)

And the noise of the ventilator disturbs me a lot!

[OldBigEars]

Quick question: any opinions as to optimum memory allocation for pre-loading tracks on a 16GB and 8GB system?

 

Great question. I have 8GB and have set mine right in the middle @ 4000 MB. I tried reducing it way down to @ 500 but this didn't seem to do much about the CPU usage, which was surprising.

[pipis2010]

The new filters seem to be good, so now we're just waiting for Damien.

 

Αny chance these filters will end up in Daphile, or are you using completely different ones there?

[Jud]

Αny chance these filters will end up in Daphile, or are you using completely different ones there?

 

Following the commits to mansr's sox-with-DSD-conversion repository may be the best way for you to stay updated on that.

[RunHomeSlow]

Hello,

 

Is the activity of the CPU normal for a 2012 Mac Mini Core i7 2,6 GHz 4 Cores in DSD128?

 

 

[ATTACH=CONFIG]25781[/ATTACH]

 

It seems to me a little lot compared with Damien (Core i7 2.3GHz Dual Cores in DSD256)

And the noise of the ventilator disturbs me a lot!

 

 

I'm around 210% also with a Late Mini 2012 i5/2.5 with 16gb ram.

 

Using filter B2 with DSD128. i've put A+ memory buffer to the max...

 

All plays well no bug yet. At last no more changes with sample rates. THANK YOU DAMIEN !

[mansr]

Αny chance these filters will end up in Daphile, or are you using completely different ones there?

 

Kimmo is aware of the updates. I don't know what he's planning.

[YashN]

Is this happening with the 2.5.0.9? Is it happening when you are not scrolling nor doing anything with the UI?

If yes, can you send me (by email) your iZotope settings, SDM filter used, DSD rate, and the file that shows the issue? And any other details you may think of to help me reproduce it.

 

Email sent to [email protected] with the Debug Log and pics of the relevant settings tabs.

 

The last glitch we heard yesterday was a Bee Gees AIFF/Redbook track, doing PCM -> DSD256 on the latest firmware iFi iDSD Nano.

[RunHomeSlow]

Hello again,

 

I've so no bugs, but it was for DSD upsampling...

 

I have the "loading first track" that appears often, had to stop than double click again,

and also the bug not counting the last track in an album as played... A+ 2.5.0.9

[YashN]

I forgot to mention another type of glitch we heard yesterday and I just heard it again just now:

 

This is still PCM -> DSD256 with a redbook AIFF (Goldfrapp - Crystalline Green today but yesterday it was a couple of other tracks).

 

At one point, one channel makes a very high-pitched 'tick' and then goes down. Sometimes it's the right channel which goes down, sometimes it's the left one.

 

I did not have Firefox running but I wasn't monitoring CPU usage either.

[orgel]

There's one occasional artifact: occasional, I'll get a weird stuttering, where the system overlays a portion of the music that played a few seconds ago onto the current portion: it's as if my DAC is composing a canon on the fly.

 

I heard this one time with 2.5.0.9 — good description — but only for a couple of seconds. I haven't been able to log a lot of time with this version, though, so it could happen again.

 

Mark me down with the others who are getting a lot of fan activity on the Mac mini (mid-2011 i5, 2.3GHz (I think), 8GB RAM) when usampling to DSD128. The mini doesn't seem to be overheating, but the fan noise is noticeable when playing music in my normal volume range.

 

Early days yet on the Filter Types, but I'm currently favoring B3.

 

--David

[Le Concombre Masqué]

based on the good looks I 'll also try 131/0.99

I tried some other settings…

 

[ATTACH=CONFIG]25762[/ATTACH]

 

 

But I find the settings "Le Concombre Masqué" more pleasant, more organic in my ears and on my system with also loudspeakers of the 90s (See signature).

Maybe that the loudspeakers of the 90s were a little better balanced ;-)

 

Best regards

[Le Concombre Masqué]

yes ! thank you Jud

No, I'm not saying the graph of your settings means nothing. I am saying if I tried to use *your* cutoff of .95 with *my* steepness of 5, it would result in a high frequency cut that would begin at a lower frequency than yours does. Your cutoff works fine with your steepness. I don't use your cutoff because I want less steepness, since that causes less ringing.

 

Hope I have explained this understandably.

[church_mouse]

Late 2012 Mac Mini i5 8gb running Yosemite. Prior to v2.5.0.9, my Mytek Manhattan would not output any sound via firewire. US2.0 worked, but even only converting RB AIFF to 128 I was getting a very unpleasant click every couple of minutes.

 

Today, on v2.5.0.9 things are much better with the Mytek. Firewire is working well (obviously limited to converting to 128). 3 hours of testing this afternoon with not a single issue. Yesterday I dropped memory loading down to 2gb and I have stuck with that today but I will try increasing to my original value (3.5gb). CPU activity is highish for the first 2 mins of each track load (but activity monitor shows no core getting even close to full load and no temperature issues at all), and then minimal CPU usage until the next section loads in to memory.

 

I have not really had chance to compare filters (save going from C to A0 - C seemed "closed-in" and less detailed to my unreliable ears), nor have I done any comparing with A+ as I previously used it (upsampling in PCM) - that said, I have been really enjoying the music all afternoon (which must be a good sign?). Actually, it was because I was getting lost in the music that I forgot I was supposed to be testing!