[wilsynet]

There are two reasons I can think of as to why 8GB is better than 2GB despite one never actually using the 2GB of physical memory.

1. The more physical memory headroom you have, the less likely there will be memory fragmentation.
2. The more physical memory headroom you have, the less likely the operating system will decide to opportunistically move things to swap.

Here's some useful background material:

http://en.wikipedia.org/wiki/Fragmentation_(computer)
http://en.wikipedia.org/wiki/Virtual_memory
https://help.ubuntu.com/community/SwapFaq

[JohnR]

Look for the checkbox in audio setup.

Oh there it is... grayed out. This is something to do with me using the Toslink output, no doubt. I'll look into it tomorrow.

Or... I could just listen! Because it's really not bad at all.

[JohnR]

There are two reasons I can think of as to why 8GB is better than 2GB despite one never actually using the 2GB of physical memory.

1. The more physical memory headroom you have, the less likely there will be memory fragmentation.
2. The more physical memory headroom you have, the less likely the operating system will decide to opportunistically move things to swap.

Here's some useful background material:

http://en.wikipedia.org/wiki/Fragmentation_(computer)
http://en.wikipedia.org/wiki/Virtual_memory
https://help.ubuntu.com/community/SwapFaq

OK, but this is pure speculation. Years ago, we would have scoffed at the very idea that you could do realtime audio with a general purpose OS. But the solution wouldn't have involved throwing more memory at it.

I can only think of three ways that sound quality could be affected if a processor was short on resources in some way. 1. Dropouts, 2. not doing all the processing in time, 3. not servicing an interrupt to provide a sample. I have no idea what hardware is in computers these days to help with 3 but I don't see how more memory is going to help, unless it sounds really horrible; 1 will be obvious; and 2 is a CPU issue. (I set the volume dithering fairly high which is why the load is showing 40+ percent, but that can be turned down again.)

[jtwrace]

Although the Hynes ps took the Alix to a new level (ahead of Macbook/PM 1.74) the release of Pure Music 1.8a and the nonmixable integer stream support for the Antelope leapfrogged the Alix/Antelope combo.  It's not a fair fight, as the Alix is all of $125, and the Hynes another $450+ (tour demo was used, and i put a deposit down on the SR3-18), but it was worth it to me.  I've now decided to use my "Hynes money" to go for his SR7-18v5 (Mac Mini external PS).

What year Mini do you have?

[wilsynet]

Yup, it's pure speculation.  I can't prove it. 

Having said that, it isn't a real time OS, and if the processor is spending more time moving memory around or having to perform more instructions to handle fragmented memory, then this would in fact impact performance, the impact likely not being dropouts but by introducing more jitter.  Which would manifest itself as minutely degraded sound quality (the kind that audiophiles notice).

But you know, 8GB of memory (from someone not Apple) is so incredibly affordable these days ...

[JohnR]

Cutting off my left toe and sticking it in my right ear is also affordable. I would still need a reason to do it.

On the jitter question, it seems that maybe some other approach to solving it would be better (given that we don't have an RTOS or proper hardware)

[ted_b]

What year Mini do you have?

2009 with SSD and 8gb.  External ps will be replaced by Hynes.

[jtwrace]

2009 with SSD and 8gb.  External ps will be replaced by Hynes.

Exactly what I have.  When do you get the Hynes PS?  Please let me / us know what you think.

[ted_b]

Exactly what I have.  When do you get the Hynes PS?  Please let me / us know what you think.

Paul got my money a couple weeks ago but is on holiday for 2 weeks, then I'm scheduled in his production run for another 2 weeks after that.  I'm guessing late September frankly.

[jtwrace]

Paul got my money a couple weeks ago but is on holiday for 2 weeks, then I'm scheduled in his production run for another 2 weeks after that.  I'm guessing late September frankly.

Can you share cost?

[ted_b]

Can you share cost?

I'd love to!!   .  Thanks!!

Oh, you meant can I share what I paid?   . Not really;  Paul worked out a deal with me as thank you for running his US demo tour, etc.  The list on it is about 580 pounds, and 90 pounds to insure and ship.

[wilsynet]

Cutting off my left toe and sticking it in my right ear is also affordable. I would still need a reason to do it.

US$59.99 for an 8GB kit for a Mac Mini from Crucial.  Much more affordable than cutting off your left toe.  Well, more affordable than the medical bills and pain medication anyway ...

On the jitter question, it seems that maybe some other approach to solving it would be better (given that we don't have an RTOS or proper hardware)

Given that we don't have a RTOS to work with here, would you agree that fragmentation and swappiness are things that would impact the real time requirements of audio transcoding?

If so, then I have the following post that gives an example of how much swap is used even when a system is not pressured for memory resources.

http://forums.macrumors.com/showthread.php?t=629752

One of the solutions offered (although with disdain) is to buy more memory.  Or one could try disabling swap altogether (I have not tested this).

http://wiki.summercode.com/how_to_disable_or_enable_swapping_in_mac_os_x

[skunark]

US$59.99 for an 8GB kit for a Mac Mini from Crucial.  Much more affordable than cutting off your left toe.  Well, more affordable than the medical bills and pain medication anyway ...

Given that we don't have a RTOS to work with here, would you agree that fragmentation and swappiness are things that would impact the real time requirements of audio transcoding?

If so, then I have the following post that gives an example of how much swap is used even when a system is not pressured for memory resources.

http://forums.macrumors.com/showthread.php?t=629752

One of the solutions offered (although with disdain) is to buy more memory.  Or one could try disabling swap altogether (I have not tested this).

http://wiki.summercode.com/how_to_disable_or_enable_swapping_in_mac_os_x

RTOS can swap too as they schedule tasks just like a normal OS, just it is just time based.     If you feel your computer is underflowing the sound card (optical) or USB DAC, then I would probably look at removing applications that might be running in the background.   

[jtwrace]

I'd love to!!   .  Thanks!!

Oh, you meant can I share what I paid?   . Not really;  Paul worked out a deal with me as thank you for running his US demo tour, etc.  The list on it is about 580 pounds, and 90 pounds to insure and ship.

WOW!  I'll be curious to hear what you have to say being that it cost more then the Mini.   

[ted_b]

WOW!  I'll be curious to hear what you have to say being that it cost more then the Mini.   

Yeah, but as I said, I got a good deal from Paul, plus....I pay that and more for a cable or set of cables to improve things..and the improvement I heard with the Alix was night and day, like major equipment upgrades.  It's not really surprising, I guess, given that a much better power supply is the centerpiece of many great system improvements.  YMMV.

[wilsynet]

RTOS can swap too as they schedule tasks just like a normal OS, just it is just time based.

No argument that a real time capable OS is able to swap.  Only thing I'm saying here is that if you cannot guarantee jitter free audio processing due to the vagaries of CPU scheduling, then processes like paging to swap may impact the timeliness of processing audio streams.

Adding memory (ie. increasing headroom) should minimize swappiness and memory fragmentation, although it is presumably not the only answer.

As for underflowing, you can introduce jitter without underflow.  And underflow you would presumably notice as audio drops.

[skunark]

No argument that a real time capable OS is able to swap.  Only thing I'm saying here is that if you cannot guarantee jitter free audio processing due to the vagaries of CPU scheduling, then processes like paging to swap may impact the timeliness of processing audio streams.

Adding memory (ie. increasing headroom) should minimize swappiness and memory fragmentation, although it is presumably not the only answer.

As for underflowing, you can introduce jitter without underflow.  And underflow you would presumably notice as audio drops.

What is your background here?   Most sound cards use a buffer in memory where the CPU converts the data to a supported codec, almost always PCM.  Then the sound card's DMA controller starts to read the data from system memory to the local buffer and other than PCIe latency and utilization, jitter is introduced by the sound card clock that is used to push the data to the DAC or SPDIF output.   

But we are probably talking USB here, which is another bucket of worms, but at the end of the day it's the same thing, the driver converts the data to PCM stores it in system memory and the DMA controller for the USB device will push the data to the USB DAC.   Sadly the for a synchronous USB DACs the USB hub can introduce data with the start-of-packet message (or whatever that thing is called) that indicates the next 1ms audio block is coming.   

Sure a RTOS will do a better job of not oversubscribing the main data paths, but I don't see how an OS vs RTOS can influence jitter on a sound card.  Dropouts sure, jitter, just not seeing it.   

[wilsynet]

If jitter were only introduced by soundcards and USB transceivers, then why does iTunes sound different than Amarra which sounds different than Pure Music, etc, but all produce a bit perfect stream of audio?

Why is integer mode support in these more boutique players desirable?

[JohnR]

If jitter were only introduced by soundcards and USB transceivers, then why does iTunes sound different than Amarra which sounds different than Pure Music, etc, but all produce a bit perfect stream of audio?

Why is integer mode support in these more boutique players desirable?

Those seem related to numerical processing to me. I don't think a player has much control over whether or when the OS decides to swap it in or out. I've got PM upsampling to 96k, with high-order noise shaping on the volume control. The bits will not be the same as what a different player will generate.

You might have to tell me what integer mode support means - I'm guessing it is so that the application decides what calculations are done on the bits, without having the OS do a floating point to integer conversion.

With regard to VM:

Given that we don't have a RTOS to work with here, would you agree that fragmentation and swappiness are things that would impact the real time requirements of audio transcoding?

Only in an extreme sense, like dropouts. Otherwise, I can't see how it can make much of a difference, except possibly through a path like modulation of the power supply.

However, you are talking as though the mini is thrashing the disk paging things in and out of swap. But it isn't! It's just sitting there processing. Here's the bottom of the activity monitor window (and fair enough, I didn't show this before):



178 MB of free RAM, and a teeny 15 MB swap file. There is very little paging going on here!! I watched it for a while and every now and then the page-in bandwidth blips, one page is getting loaded in for some reason. For all I know it could be Activity Monitor doing that.

I suppose disabling swapping would be one way to be sure, but it seems completely unnecessary based on what I'm observing here. You are right, memory is not that expensive - I don't think I can get it for that price but it's still a fair bit cheaper than I thought it would be. Nonetheless, I'm still scratching my head at the contention that it's necessary.

[wilsynet]

Yes, of course, the application has no control whatsoever as to whether the OS decides to page, or whether memory is ultimately fragmented or not.

I don't have upsampling turned on in any way, yet Pure Music versus Amarra versus iTunes all sound different.  So bits are the same, the sound is different across different players.