HQ Player

[jabbr]

Yes, that's on the TODO-list now that latest CUDA should have the necessary feature set. You may still need to use Quadro-series card with Maxwell generation GPU though... (not sure yet)

 

 

 

I have Haswell-EP generation Xeon E5 workstation and it's not actually noisy at all. It is more about how the cooling is implemented. Many ordinary desktops/laptops are not designed to be running at full loads for extended periods and their thermal solutions become noisy in such situations. While workstations are designed for heavy loads and the thermal solutions are thus also optimized for sustained full loads.

 

Thermally: Yes I watercool even when not overclocking -- including GPU -- look at performance/db ratios -- also use fanless NAA for silence

 

Electrically: GPUs and 4k monitors can be very noisy. NAA isolates this very well.

 

Optically isolate both Ethernet and USB sides of NAA to firewall DAC from noisy workstation

[jabbr]

Has anyone written an OS optimizer (Fidelizer or CAD's script for Win 8, AO for WS2012) script for Win 10 yet? And if so has anyone used it with HQPlayer?

 

So none of this matters when using NAA ... my philosophy is to send the audio bits to another low powered machine which can be optically isolated if you want with an inexpensive FMC pair. The NAA should run minimal software.

[jabbr]

So you are recommending stock OS install with all its crazy extra processes..and that NAA is a panacea...even for performance (let alone sq)? I find that difficult to believe. Yes, NAA is a great FIFO buffer, but how is it a replacement for clutter management at the server end?

 

Pretty much. No need to optimize the machine HQPlayer is running on. You can optimize the NAA to your heart's content including LPS, PPA USB V2 card, Corning USB/regen etc etc. The noise from the HQPlayer machine is brickwalled from the NAA.

[jabbr]

I have all those things (JCAT USB card, separate linear or battery power for all SSD, card, mobo, etc) on my server. Are you saying my WS2012 server ought to be the NAA? Then use a puny Atom CAPS v2-ish HQPlayer machine? It will never do any upsampling, since even my i7 3770S server is brought to its knees at the max end (eq'd DSD128 to DSD256).

 

My understanding of the server vs NAA is quite different, with NAA allowing to be lightweight and the server doing the heavy (in my case very heavy) lifting. So move JCAT card to a small Windows NAA? It still doesn't answer my need to have a well oiled (optimized) heavy cpu machine to do heavy lifting. Am I missing something here (and I easily could be )?

 

The heavy lifting machine which does the upsampling/conversion/deconvolution etc. is likely to have high power requirements and will generally have an electrically noisy environment. My machine has a couple of GPUs and a 4k monitor which emits gobs of noise. It may also be physically noisy with fans etc. Let that machine do the math and send the bits out over the wire (Ethernet). HQPlayer needs to run on this machine because it has very high CPU requirements.

 

I have an NAA which is a DN2800MT board which is low power and silent. It is in a different room than my workstation. It receives the bits. I have a PPA USB V2 which I was getting ready to install when my Corning and regen arrived. So my NAA outputs via Corning to regen -> DAC (iFi iDSD micro which does DSD512 in my case). This is both an electrically and acoustically low noise environment (well except for my plasma TV but go with me here )

 

The point is that HQPlayer runs on the heavy lifting machine (in my case i7-4930k with 64gb) which I leave as a general purpose workstation which is electrically isolated from the NAA and consequently electrically isolates all the junk from the audio hardware. Right move JCAT card in your case to the NAA (the NAA talks to the DAC). The audio optimization stuff doesn't optimize the heavy lifting machine to do a better job at math. I believe the audio optimization scripts are designed to reduce the electrical noise of the machine. Since the NAA is the one connected to the DAC, it is the NAA which should be stripped down and optimized.

[jabbr]

There is a Windows version of the NAA software so you should be good to go. There are reports of issues with private connections (it wants DHCP). If Ethernet switches are a concern, or there is a concern about noise travelling down the Ethernet wire, I think the dual FMC optical isolation technique is a nice low cost solution. I'm in the process of evaluating SFP+ connections but nothing to report yet.

[jabbr]

I have tried with fmc's and linear supply on the pc running hq player and it still made a difference despite the naa pc already fitted with fmc's and linear supply. The hqplayer pc alone with fmc's and linear ps is improved by an naa pc with switching supply And without fmc's. Go figure huh..

 

What size LPS on the HQPlayer machine? What level DSD are you upsampling to? What are the specs of your HQPlayer machine?

[jabbr]

It was in response to this quote from jabbr: "No need to optimize the machine HQPlayer is running on. You can optimize the NAA to your heart's content including LPS, PPA USB V2 card, Corning USB/regen etc etc. The noise from the HQPlayer machine is brickwalled from the NAA."

 

My experience has been that improving the playback abilities of either machine in a dual-PC setup can have a positive impact on the system.

 

Perhaps I stated this too strongly, however

 

there is less need to sonically optimize the machine HQPlayer is playing on than the machine networkaudiod is playing on (NAA).

 

 

Just because you have a different experience with different software and different hardware does not mean that your experience applies generally to all software and all hardware. The problem with tweaking as opposed to engineering is that tweaking is not generalizable. Moreover I don't consider HQPlayer/NAA to be a dual-PC setup. The two machines ideally are drastically different. In my opinion, one should be compute intensive and optimized thusly, and the other low power and electrically and acoustically quiet. YMMV

[jabbr]

@jabbr

300VA

Dsd128

Both hqp and naa machines i7 3770T

 

So I would do this very differently and thus I'm not sure your optimization strategy would work with the way I would do it:

 

I upsample to DSD512 which requires a computer needing more than 300VA power. It would be hard to get an LPS to power my workstation. Miska has published noise data regarding LPS and HQPlayer direct and I'm less sure that LPS on the HQPlayer machine would have an effect on noise at the DAC when going through NAA. I am not willing to drop down to DSD128 just to use an LPS. That said I am seriously considering the new crop of D-1540 machines which combine low power and 8 cores at 2ghz. Enough for DSD512???

 

I'd use a lower powered machine for NAA. Even consider ARM. I would tend to believe that if the ground plane optimization resources devoted to the USB Regen were devoted toward an audiophile ARM motherboard, that this might have a very nice sound.

[jabbr]

Jabbr, do I install Win 10 64 bit or x86 on my DN2800MT? It used to be a 32 bit board but lately I see 64 bit drivers available. Win 10 pre-release asks 64 bit or x86. And do I assume that you felt, then, no need to install a good 3rd party USB card (you PPA, me JCAT) once the Corning and Regen were available? Seems a good card would be icing, no?

 

And whata re folks doing with 10M of Corning USB when their DAC is sitting within a couple feet! I see no offering for a normal length Corning.

 

I have been using ubuntu on my NAA but do plan to try Windows 10. I do have a PPA V2 USB that I also plan to install -- icing agreed. I find that if I make too many changes to my system to quickly I have a hard time making heads or tails regarding what is an actual improvement. The current sound with the Corning and green Regen is remarkably good. I have an upgrade to Amber regen on order. I also plan to use Ethernet optical isolation with a pair of FMCs. So lots and lots and lots of variables.

[jabbr]

Appreciate the guidelines above. What is you experience with Xeon vs i7 ? Are you getting similar sound quality ?

 

Have you used Primo Ram disk for the Signalyst folder ? My concern with Xeon is that it's memory controller will negate Primo Ramdisk's advantage of writing data sequentially in memory. I understand Xeon memory control writes is data striped over several locations in a RAID like manner.

 

Also Xeon's have lower clock speeds than i7 for the same CPU benchmark scores

 

Things are way more complicated than they first appear. Sequential memory, and sequential processing is only an abstraction these days. For example, the new crop of Xeon E5's have a direct nework connection to processor cache and at 10Gbs network arguably the difference between RAMDisk and NAS is far far less than it used to be:

 

Intel® Data Direct I/O: Frequently Asked Questions

[jabbr]

I installed Win 10 on two computers, named HQPC and NAA, installed HQ Player on HQPC, DAC ASIO drivers on NAA. I cannot see anything under devices on HQPC when I go under settings. Backend is selected NetworkAudioAdapter. I think some kind of networking issue. Both machines are part of Workgroup. Someone said should be part of same domain, I thought you needed some kind of domain controller to do that. I can see both machines from each other. I can remote-in to them. I do not know what is going on. Let me know please. Thanks.

 

Try:

 

1. quit HQPlayer

2. unplug and replug DAC

3. reboot NAA

4. restart HQPlayer

[jabbr]

In the context of using HQ Player to do PCM to DSD, the Lampi B7 (DSD128) and the AZP with 10 M clock are quite close sonically. The AZP was not fully burnt in, so I'm not in the position to comment in detail.

 

the sonic difference between these DAC's rates about 2 points on a scale of ten, whereas the battery powered SSD rates about an 8

 

Ever try M.2? NVME?

[jabbr]

For those loving the HQP sound, especially when upsampling CD rips to DSD, can you please share what DAC you are using?

 

iDSD micro @ DSD512

Got DSC1 boards that will be built ... will modify PSU

Waiting for DSC2 which will also be build

[jabbr]

I am wondering if you think changing the configuration of my system would help.

 

I am currently running the following system:

External Hard Drive -(USB 3.0 connection)-> Late 2012 Mac Mini 2.6Ghz i7 4 core 16GB -(shared folder over wired ethernet connection though 2 routers & a switch)-> Late 2014 Mac Mini 2.6 i5 2 core 8GB running HQplayer outputting DSD128 -(USB Connection)-> Oppo 105D

 

What if I switched to the following setup?

External Hard Drive -(USB 3.0 connection) -> Late 2012 Mac Mini 2.6Ghz i7 4 core 16GB running HQplayer outputting DSD 128 -(Wired ethernet connection through 2 routers & a switch) -> Late 2014 Mac Mini 2.6 i5 2 core 8GB running NAA -(USB connection)-> Oppo 105D

 

Is there any reason to believe this setup might sound better? If so why do you think it might?

 

You should get rid of your external USB disk drive. There are a whole variety of better options. If you must direct-attach then use Thunderbolt. Otherwise a NAS is an excellent general option.

[jabbr]

Yes, another chapter in the "go figure" book of audiophilio. However, a caveat is that Miska assumed the music coming from an internal drive would benefit from RAM disk. My music is already fairly well handled via my NAS path and doesn't live on internal drives. Maybe worth something.

 

Consider this: HQPlayer on a diskless node connected via 10g optical to NAS. Now RAM disk does exactly what? The point being that one can completely eliminate disc based noise by removing all discs from the system and booting over the network (PXE boot).

[jabbr]

[jabbr]

Thanks Ted

 

I have moved the JCat card over and gone single machine for now.

 

My router is maxed out and sending data streams through a router plus a switch probably wouldn't show off NAA at its best

I've seen the fibre networking threads and will keep that in mind

 

I will be interested in your eventual verdict on one machine vs two machines

 

Thanks

 

You should not be sending traffic between HQPlayer and NAA through router just switch. Yes a fiber network will improve sound (reduce noise) and the switch will actually help by re clocking and reconstituting the Ethernet signal.

[jabbr]

Is there a document which describes the XML configuration? I am trying to name my NAA but having trouble changing it from "default" -- not sure why. Its probably easier for me to just edit the XML itself ...

 

Thanks,

 

Jon

[jabbr]

beta2 working much better.

 

I can: closed form, ADSMV7, DSD512 whether or not I GPU off load to 760

also poly-sinc-mp-2s either way but:

poly-sinc-mp stutters at DSD512 regardless of GPU offload or not, buffer times, etc. CPU load roughly 220% regardless. this load seems nicely better than v11 regardless

 

Jon

[jabbr]

I've thought about it, but it doesn't actually fit there. On some OS and some display/font size settings it would fit, but for example for me on Linux with 4K display it doesn't fit and I'm reluctant to make the information boxes larger because it immediately cuts out one line of space from library or playlist views. I rather just fit it on the same format line.

 

If you use latest Roon builds with HQPlayer, the active rate is actually reported there...

 

 

Edit: I changed it to look like this a bit of fitting exercise, is this better than before?

[ATTACH=CONFIG]24014[/ATTACH]

 

I've just started using Roon integration and the biggest thing it does for me is convert my old ALAC encoded iTunes library.

 

I also very much want to see what the actual output bitrate is. This looks good.

 

Available for OS X?

[jabbr]

That wouldn't be called HQPlayer, it would be called something else... I don't want one thing that tries to be be little bit of everything but just mediocre on every single thing. I rather have one or more separate things that are good on precisely one thing.

 

Perhaps you could trick HQPlayer into handling this by wrapping the digital stream in a file interface -- HQPlayer might want a file length? That could be made arbitrary.

[jabbr]

It doesn't need length, for example with Roon there's no length. I just don't like doing that kind of trick and trying to bend HQPlayer to do something it is not designed to do. It doesn't mean I couldn't design something for that purpose too. I'm just not particularly fond of multi-purpose tools.

 

Fair enough, though in Linux everything is a file

 

I've been considering a multichannel variant of DSC1 along with kernel filters that would make a very slick crossover in a multiamp situation. I think that's within your target intentions though and that said I'm not going back to vinyl.

 

I think that someone who really wanted that feature could do it without your assistance by dropping the appropriate "/dev" file into HQPlayer

[jabbr]

I've loaded HQPlayer 3.14.4 onto my son's Windows 10 gaming computer so I can test out CUDA offload.

 

We built an i7-6850/GTX1080, and this machine is connected to our network via 802.11ac/AC1900. So far so good. Seems to do DSD512 using poly-sinc -> NAA ... few stutters but I haven't tried to optimize. I'm controlling via Roon that sits on my OS X machine. All good except that the Volume control doesn't work while music is playing -- Roon says volume is fixed, and the volume dial doesn't work on HQPlayer itself ... needless to say its really hard to compare sound.

 

If I'm going to do this for awhile I'll need a separate Windows license for HQPlayer...

[jabbr]

... was that pesky "DirectSDM" setting that's hidden away in the "other" menu... was enabled by default. From a UI standpoint, consider moving that to main settings because it has effects even when (trying to) DSD64->DSD512

[jabbr]

Could dual-boot Linux (e.g., Lubuntu) - would give you an OS that's pretty well optimized out of the box, and you can use your Mac license.

 

I've considered that but ... want to run HQP essentially as a service: as a startup item and hidden :)

 

(upsampling to DSD512 uses 20% cpu on this machine)