Need Advice - Dedicated PC Build

[glynnw]

Finally ready to build a PC dedicated to using XXHighEnd as my main source. and need advice.  I have never assembled a computer, but I built all my tube gear and speakers from Bottlehead kits so I can probably figure out the assembly part.  Looked with interest at SeVeReD's choices, but my ignorance keeps me from good understanding of what I need as a minimum.  (I am assuming I can always update gear later as needed).  Intend to continue to use current peripherals, so USB will be used for Hard Drive and DAC (Scott Nixon NOS).  Would like to place all in horizontal case and put on equipment rack where CD player used to reside, but am flexible if this is not a good idea.  Read that Peter says to use a Duo Core processor. 

So I guess first question is - what should I look for in a processor and motherboard?  Are there minimum requirements for each beyond which SQ is not influenced?

Second question - if unit is truly stand alone (not connected to internet or LAN), what is best method for getting new info into it, such as updates?

[PeterSt]

I guess the motherboard isn't that important when you don't have requirements on SataII connections (which as SeVeReD's base). The Core2Duo (mind the difference with a DuoCore !!) is the main requirement and the more Level2 cache it has on board, the better it is.

When the unit is not connected to the Internet, a USB stick suffices. Just download the update to the stick, stuff it into the audio PC and unzip it from there (or unzip first at the receiving PC, doesn't matter). Then copy the contents from the stick to whereever you intended as usual.

BUT :
The main mistake you make is having USB harddisks and a USB DAC at the same time. Besides USB is relatively (very) slow, harddisk acces has priority over the data going to the DAC. It depends a bit on what you do and when, but for sure it is asking for (playback) troubles.
When you use similar now, it is *not* said at all, that this will prevent you from the problems as implied on the new machine ...

Peter

[glynnw]

If when using XXHighEnd, the data playing is held in RAM, why would the computer need to get more info from the harddisk while sending a signal to the DAC?  While USB has limits, I am under the impression that, with the exception of the new feature that allows playback to begin while still loading info from harddisk to RAM ("start Engine 3 during conversion"), the 2 operations do not occur at the same time.

Sorry if my ignorance of all things computer is confusing me.

[PeterSt]

I unerstand your confusion;
When gapless is required (and we just have to state it is) there is a small overlap near the end of a track when the next track is loaded.
This is unrelated to starting playback during conversion (!!) of the other tracks right after the first track has completed conversion.

So, "conversion" happens in advance of everything, however, during that stage again the harddisk is used (excessively), and after that has been done, there is still the loading of the tracks from harddisk (during the end of the track playing) which doesn't need conversion anymore then. It is that stage (and at conversion when in order at the beginning) that the USB bus has priority over the data that has to be supplied to the DAC. This is when you may receive hiccups, ticks and glitches. May ... because when everyting is fast enough or happens at convenient times, nothing is the matter. This is related though to the Q1 settings, and you won't be able to set that as low as you may want for the best SQ.

Besides all (and this may be of importance) when you have some mass copies going over everyting but USB during playback, I kinfod guarantee you will never hear that to the sense of hiccups etc. But when you have even one going over USB I dare to guarantee that you will notice.
YMMV but this is just how the theory is. Also, I don't want to be right on this, but instead I just want to warn you. So just keep in mind : there is no way playback can be interrupted (Engine#3), unless you incur for stuff which has a higher priority which nothing can be done about. USB is one reason, challenging the swap file is another. I don't know of any other currently.

Hope this helps ,
Peter

[glynnw]

Thanks for the info.  So to get the best sound, I should either get a non-USB harddisk (which currently has about 300+ GB filled with music files) or a non-USB DAC.  Perhaps the best route may be to have 2 storage devices in the case - 1  small SSD for operating system and large harddisk for music files.  Then look to upgrade the DAC in the future if (when-hehe) I   feel it is needed.  So much to learn. And by the way, I really am looking forward to your remote system. Thanks - I am sure I will have more questions very soon.

[PeterSt]

Well, don't change your DAC because of this !

[SeVeReD]

Well, don't change your DAC because of this !

but do take your hard drive off of the usb.  Not good to run both dac and HDDs off the usb.  Listen, I got pretty good results using USB... but cable (I found no name 1m clear plastic silver-braid type that beat everything generic I threw at it, including belkin gold, did not try the really expensive ones... so gl), does make a difference AND you need to find the best sounding USB port on your machine you can find... trial/error, then shut off all the other USBs you can.

Also, everyone!!!!
I get more and more convinced that a single WAV file & CUE file is the best way to go,,, then you don't have to worry about gapless.  You CANNOT have your cake and eat it too.... meaning, you can't load different tracks from different albums and have it play the best presentation possible ... especially because the volume is going to change from track to track and there is no way to "level" the volume and get good playback. This is the only reason I can see for ripping a CD as individual tracks and not as a single wav file... am I missing something, cause I think you all might be?(little "drunk" sorry for the dramatics)

If you build my machine,,, well, I'm very very happy right now, course hehehe I'm "drunk"

[glynnw]

Concentrating on building a quiet PC, does the clock speed of the Intel Core 2 Duo make a big difference to XXHighEnd player?  All the e8000 series have 6MB of L2 cache - will the lower clock speed models run at lower temps, thus being easier to cool with less resources?

[PeterSt]

Generally, yes. But of more importance is the size of the fan (a tad larger makes a huge difference in neededs revs) or just water cooling.
Without notice, there are fans for the case (could be two), the cpu and the PSU. When the cpu is not stressed, they run on the lowest rev. However, since all is pro-active (what is the current temp ?) the smaller the fans, the more they have to work to make the needed flow.

Just some strange advise : nowadays, as it turns out, just leaving the case open (leave off the lid) just makes the difference needed; It is quite impossible to incur for a higher environmental temperature, impeeded by cpu or hard disks. Just my own experience.

Peter

[glynnw]

Thanks Peter - I am spending a lot of time (retirement is a good thing) at the various silent PC sites and will construct with much of what I am learning.  But since I am building from scratch, I want to start with the components that will make silencing the PC a bit easier.  If it makes no difference to performance with XXHighEnd, I will use a CPU with a lower clock speed to save the minute amounts of heat I may incure with a higher clock speed.  So I guess I am asking is there any advantage with XXHighEnd to using a CPU with clock speed of 3.16  GHz vs 2.66 GHz?

[PeterSt]

Certainly not, meaning : the only difference you will notice once you use the "proper" (zee below) cpu is at conversions which are very cpu bound, and the faster the cpu the earlier they are behind you. But also read the Release Notes on 0.9w-2 on this matter, because it might free you from clicks and such once the cpu is faster. This goes way beyon the few lines spent in there, but it comes to this.
On the other hand, my Core2Duo is 2.4GHz only, and if one never has problems, it's me. But that's me ...
With the latter I mean, once you kind of know what you're doing, you will be able to avoid problems. When ot or less, you can buy them off with, well, money. Mind you, in general. But many many things are speed incurred, and the sooner the problem is over, the less you will notice it, or even - occur.

Personally I wouldn't bother (at all) for higher clock speeds and the necessity for more cooling, as long as you don't over clock. Again, leaving off the lid already solves all, although it may be a strange solution (and sight of course). When leaving the lid on, it is just (and only that) of the utmost importance that you keep the PC (hence motherboards, PSU inlets, CPU cooling lid) clean of dust. This is just a matter of blowing through it with a compressor which every car repair as and will be glad to use (on you). It is too less realized that this does most of the job, while at the same time it *is* realized that when the PC was new it was so quiet ...

In any case, do not give way to the cheaper cpu's like Celeron, because they *will* matter on sound and everything else. And as a major rule : the 2L cache is what heals all. The more you have of it, the exponentially better things work out.

Hope this helps a bit.
Peter


PS: A nicely water cooled system looks way "cool" of course. Must admit that ...
PPS: Do not make the mistake that XX needs the clock cycles. It is the other (environmental) processes which do and which can disturb. Like a disk, a cpu can trash too (for sure a multi core -> dying from overhead at switching shared memory). This is the first thing to avoid. But hey, back in my old days I claimed to better have a 12MHz than an 8 MHz cpu. Now what.

[andy74]

PPS: Do not make the mistake that XX needs the clock cycles. It is the other (environmental) processes which do and which can disturb. Like a disk, a cpu can trash too (for sure a multi core -> dying from overhead at switching shared memory). This is the first thing to avoid. But hey, back in my old days I claimed to better have a 12MHz than an 8 MHz cpu. Now what.

Peter,

what exactly do you mean by "(for sure a multi core -> dying from overhead at switching shared memory)"
Is it bad to have multi-core for XX or not. Can you clarify.

Andrey

[PeterSt]

Hi Andrey,

As said, this is about the other processes, because XX will arrange it for herself, and in the mean time tries to arrange it for her environment as good as possible.

This can be seen / understood probably the most easy by searching Google for Quad core problems, and the unextpected relatively low performance of those, opposed to the dual cores. In general 2 x 2 = 4, but the performance of the Quad stays at just over 2 (it is maybe 8 months ago I investigated things around this).
Note that these problems emerge(d) at the digital audio workstation softare (Cubase et al), which softare similarly to XX tries to utilize the processor to the max, for obvious reasons in there.

I too have tried to utilize the 4 cores, but failed. I mean, there is no (to me ovbious) means to spread unknown tasks over 4 cores. And the contrary, with two cores this is easy : XX on one, the rest on the other.

When this is out of proper control, the OS determines which process (thread) goes to where, and this is just based on capacity. It does, however, *not* include the calculation of the overhead at switching a process from one core to the other, or better : moving a process out of focus (time slicing) in order to let another in. And now the big trick : when swithing happens, the processors' own memory has to be replaced as well (the memory is decidated to the process running), and it is there where it goes wrong. Careful : my thoughts, and possibly nothing you can read somewhere.

All together I am not sure how it exactly works, but when e.g. two processes need the cpu, the two will take longer when running together opposed to run them subsequently. Compare with two processes needing near 100% disk I/O (not SSD) : each process needs its own head position, and the overhead of that determines the throughput time. Run the processes after eachother, and they can be 100 times more fast easily.
The positioning of the disk head can be compared to the copying of the processor's memory. It takes time and that time is overhead.
"Trashing" would be the situation that the overhead becomes larger than the net time needed without the overhead and can't catch up.

On the latter : Keep in mind that XX dedicates a core to herself, therewith giving the other processes twice as less change (dual core) to come through within the time needed (whatever that is, but the "swapfile" thread is an example of this, solved by temporarily allowing that process to use two cores).

Ah, hard to explain it all for me, but I'm sure you know what I mean.
Peter

[andy74]

Thanks Peter,

So you meant 4 cores, while I had only 2 in mind. It makes sense now.