Jump to content

Recommended Posts

How strange, you have something very fundamental mis-configured there. I'd be tempted to pull the graphics card and try again. It shouldn't matter but is the internal sound card disabled?

 

I've tried both with internal sound card disabled and abled and it's the same.

As performance seems to be working as it should with NAA - confirmed by Miska on the 'build a Roon/HQP server' thread - I'm sticking with that.

Share this post


Link to post
Share on other sites
But I think I'll give up and go back to the NAA arrangement as the 6900k is behaving consistently in failing to upsample smoothly 44.1-based rates to the 48-based 24M6 stereo. And CUDA offload makes little difference.

 

Just as a note, with the -2s filter variants there shouldn't be any problem doing that. But the non-2s seems to be too much.

 

For some reason, only the Tesla P100 seems to be putting out insane performance figures. Some claim that Nvidia is capping the double precision FP performance on purpose on GeForce to sell more Quadros, but based on the figures I've seen there doesn't seem to be such difference. Only for the biggest and most expensive Tesla.


Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Share this post


Link to post
Share on other sites

Well the P100 is a totally new chip, 15 billion transistors, the biggest GPU on the block. It's battling the Xeon Phi even while designed to work with it.

 

Sent from my Nexus 6P using Computer Audiophile mobile app


Mac Mini 2012 with 2.3 GHz i5 CPU and 16GB RAM running newest OS10.9x and Signalyst HQ Player software (occasionally JRMC), ethernet to Cisco SG100-08 GigE switch, ethernet to SOtM SMS100 Miniserver in audio room, sending via short 1/2 meter AQ Cinnamon USB to Oppo 105D, feeding balanced outputs to 2x Bel Canto S300 amps which vertically biamp ATC SCM20SL speakers, 2x Velodyne DD12+ subs. Each side is mounted vertically on 3-tiered Sound Anchor ADJ2 stands: ATC (top), amp (middle), sub (bottom), Mogami, Koala, Nordost, Mosaic cables, split at the preamp outputs with splitters. All transducers are thoroughly and lovingly time aligned for the listening position.

Share this post


Link to post
Share on other sites

The new beta of HQP is significantly slower than the older beta. I was able to upsample to DSD 512, non -2s filter from PCM 24/96 without sweat on my i7 4770T. And now, after downloading the newer HQP, even PCM 24/48 upsampled to DSD 512, even using -2s filter stutters a lot.

Share this post


Link to post
Share on other sites
Well the P100 is a totally new chip, 15 billion transistors, the biggest GPU on the block. It's battling the Xeon Phi even while designed to work with it.

 

All the Pascals are totally new, GP100 was the first in series followed by the other ones (GP102/GP104 with 12/7.5 billion transistors, and later stripped GP106/GP107 with 4.4/3.3 billion transistors and 14 nm process). If you look here:

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#Tesla

 

The only major difference between P40 and P100 is the memory subsystem HBM2 with 4096-bit bus vs GDDR5 with 384-bit wide bus. Then if we look at the performance figures, P40 is actually a bit faster on single precision FP while P100 is over 10x faster on double precision FP (used by HQPlayer).

 

P40 is not so different from the Titan X:

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#GeForce_10_series

And Quadro P6000 is practically same as P40:

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#Quadro_Pxxx_series

Which is clear because P40, P6000 and Titan X all use the same GP102 chip with slightly different configuration.

 

So is the P100's huge double precision FP performance just because of the integrated on-module RAM subsystem? It is not clear where the huge difference comes. Because the data size difference is only 2x, it doesn't alone explain 10+x speed difference.


Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Share this post


Link to post
Share on other sites

Thank you Jussi. I have no answer at all. I should have known you were following these various gpus. You've helped me out because some of the cards are just becoming available, and a project before me must make a decision. I'm neither a programmer nor physicist (yet) but am part of the team.

 

It seems to me that it's either very local architecture, such as larger registers within the cores, or the vertically attached RAM.

 

We're installing a developer system to give practice for physics students and faculty on both Nvidia CUDA coding and Intel Xeon Phi (exclusive) coding. It has a single Knights Landing cpu and will have one Nvidia card installed.

 

We won't be able to compare NVLink against Intels interconnect fabric, but will get some idea of CUDA vs Xeon Phi in programming efficiency. These new products and their libraries will inform our HPC decision in just a few months. Wish I knew more.


Mac Mini 2012 with 2.3 GHz i5 CPU and 16GB RAM running newest OS10.9x and Signalyst HQ Player software (occasionally JRMC), ethernet to Cisco SG100-08 GigE switch, ethernet to SOtM SMS100 Miniserver in audio room, sending via short 1/2 meter AQ Cinnamon USB to Oppo 105D, feeding balanced outputs to 2x Bel Canto S300 amps which vertically biamp ATC SCM20SL speakers, 2x Velodyne DD12+ subs. Each side is mounted vertically on 3-tiered Sound Anchor ADJ2 stands: ATC (top), amp (middle), sub (bottom), Mogami, Koala, Nordost, Mosaic cables, split at the preamp outputs with splitters. All transducers are thoroughly and lovingly time aligned for the listening position.

Share this post


Link to post
Share on other sites
The new beta of HQP is significantly slower than the older beta. I was able to upsample to DSD 512, non -2s filter from PCM 24/96 without sweat on my i7 4770T. And now, after downloading the newer HQP, even PCM 24/48 upsampled to DSD 512, even using -2s filter stutters a lot.

 

What are the exact versions you are comparing? >=3.14 is somewhat slower than <3.14, due to some algorithm changes to accommodate the AMSDM7 modulator and some things coming later (internal experimental testing stage). But there are also some new optimizations that largely offset this difference.

 

So it could be something else. Have you checked the power profile you have in Windows? On my new machine I needed to change it to "Performance" because the "Balanced" was throttling the CPU too much.


Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Share this post


Link to post
Share on other sites

When will the new version of HQ Player be released ? I'm asking this because I don't want to buy the current one now and a couple of weeks later to have to buy the new one.. :)

 

Since I'm building my Music Server, please tell me if I should get an i7 6700 (not the 6700K, since my motherboard can't do OC) or will an i5 6600 be able to upsample PCM 96khz to DSD 512 with all those nice sounding filters ?

Share this post


Link to post
Share on other sites
When will the new version of HQ Player be released ? I'm asking this because I don't want to buy the current one now and a couple of weeks later to have to buy the new one.. :)

 

Since I'm building my Music Server, please tell me if I should get an i7 6700 (not the 6700K, since my motherboard can't do OC) or will an i5 6600 be able to upsample PCM 96khz to DSD 512 with all those nice sounding filters ?

 

Doesn't matter when you buy since I don't know if Miska ever charges for upgrades, but he hasn't for as long as I can remember.

 

Not sure what the 6700K has to do with overclocking - best to get the most capable CPU you can afford.

 

 

Sent from my iPhone using Computer Audiophile


One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> wi-fi to router -> EtherREGEN -> microRendu -> USPCB -> ISO Regen (powered by LPS-1) -> USPCB -> Pro-Ject Pre Box S2 DAC -> Spectral DMC-12 & DMA-150 -> Vandersteen 3A Signature.

Share this post


Link to post
Share on other sites

Since I'm building my Music Server, please tell me if I should get an i7 6700 (not the 6700K, since my motherboard can't do OC) or will an i5 6600 be able to upsample PCM 96khz to DSD 512 with all those nice sounding filters ?

 

I would go for the i7 6700 wirhout K, as I use with succes.

My Roon-HQPlayer Windows 10 PC does DSD512 with all filters. If you want to help the CPU further, you can always add a Cuda graphics card.

The i5 will most probably not suffice to do DSD512 on its own.

Share this post


Link to post
Share on other sites
I would go for the i7 6700 wirhout K, as I use with succes.

My Roon-HQPlayer Windows 10 PC does DSD512 with all filters. If you want to help the CPU further, you can always add a Cuda graphics card.

The i5 will most probably not suffice to do DSD512 on its own.

 

Just to clarify: I presume you mean 'it does DSD512 with all filters AS LONG AS YOU DON'T TRY YO UPSAMPLE 44.1k BASED RATES TO 24M6'. The 6700k stutters with that conversion (though not with 48k based rates to 22M6), and CUDA offload up to a 1080 GPU doesn't help, so I assume 6700 stutters with it too? So you need to have auto-rate family checked.

 

 

Sent from my iPhone using Computer Audiophile

Share this post


Link to post
Share on other sites
What are the exact versions you are comparing? >=3.14 is somewhat slower than <3.14, due to some algorithm changes to accommodate the AMSDM7 modulator and some things coming later (internal experimental testing stage). But there are also some new optimizations that largely offset this difference.

 

So it could be something else. Have you checked the power profile you have in Windows? On my new machine I needed to change it to "Performance" because the "Balanced" was throttling the CPU too much.

 

This is happening using the latest beta (3.14beta8). The performance degradation (in terms of CPU load) is very noticeable. Between my last version of HQP and the newest beta 8, I haven't made any single change to my OS. Btw, I am using Windows Server 2012 R2 with Audiophile Optimizer. Not sure about the power profile.

Share this post


Link to post
Share on other sites
Guys, do the balanced outputs sound better than the single-ended ones of the T+A DAC8 DSD ?

 

Can both be used at the same time?

 

- Richard


LMS on Odroid XU4; HQPlayer on i7-8700; iFi iGalvanic; T+A DAC 8 DSD; Benchmark AHB2; Quad ESL 2805s + two Acoustic Energy subs.

Share this post


Link to post
Share on other sites

Thanks for your inputs guys. I'm gonna get the i7 6700 and I will keep auto-rate family checked, hopefully that will do the trick.

Share this post


Link to post
Share on other sites
This is happening using the latest beta (3.14beta8). The performance degradation (in terms of CPU load) is very noticeable. Between my last version of HQP and the newest beta 8, I haven't made any single change to my OS. Btw, I am using Windows Server 2012 R2 with Audiophile Optimizer. Not sure about the power profile.

 

You should be using latest release version which is 3.14.3...

 

I have not noticed anything significant.


Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Share this post


Link to post
Share on other sites
Yea, but there is said to be a SQ discount on both outputs when used simultaneously.

 

The Exasound explicitly says you can use both, which I assume means that for Exasound there is no SQ penalty. Can anyone confirm this? There is also said to be a penalty on the MDAC, which I'm currently using.

 

For me, It's a question of finding a DAC that I can also easily feed the L & R channels into a couple of active subs.

 

- Richard.


LMS on Odroid XU4; HQPlayer on i7-8700; iFi iGalvanic; T+A DAC 8 DSD; Benchmark AHB2; Quad ESL 2805s + two Acoustic Energy subs.

Share this post


Link to post
Share on other sites

Does the T+A DAC8 DSD benefit from an Intona or a Regen? Did anyone try these on the T+A?

I have a new PC on i's way to me and am anticipating...

could just buy both and try for myself but...


Synology 916+/Tidal --> fiber --> Er --> HDplex lps powered 6700K with JCAT NET running HQPe on Euphony—> Pink Faun Ultra OCXO I2s--> Pink Faun DAC 2.32 (32 x PCM1795 & dual power supplies)--> Ayre AX5 --> Paradigm S8 (Nordost SPM & TYR cables)

Share this post


Link to post
Share on other sites

I found my T+A immune from my regen, and took the regen out of the chain.


CAPS Pipeline with HDPlex Linear PSU running Win10 64 bit, AO 2.0, RoonServer, HQPlayer -> T+A DAC8 DSD -> Linear Tube Audio's MicroZOTL2 Headphone Amp with Mojo Audio's Illuminati Linear PSU -> Focal Utopia/Audeze LCD-3

Share this post


Link to post
Share on other sites

If you have a standard CAPS-like computer with linear power and audio grade USB card, Intona is the way to go, imho.

I compared the usual AQJB/Regen combo against Intona and in my system the Intona was the winner.

The T+A is less affected by poor sources, but not immune. Only if you bring on the big boys like an SGM, no additional devices are needed.

As a side note, Intona, AQJB and Regen together was not better than Intona alone, in my system at least.


Consultant to Taiko Audio

Share this post


Link to post
Share on other sites
If you have a standard CAPS-like computer with linear power and audio grade USB card, Intona is the way to go, imho.

I compared the usual AQJB/Regen combo against Intona and in my system the Intona was the winner.

The T+A is less affected by poor sources, but not immune. Only if you bring on the big boys like an SGM, no additional devices are needed.

As a side note, Intona, AQJB and Regen together was not better than Intona alone, in my system at least.

 

Will give my opinion when I get my cables. I have Curious cables on the way and have JCAT version of Intona. Looking forward to it. I really like this DAC!! My headphone system is really coming together and I have to say is quite remarkable. There is a definite burn in time necessary and I really recommend anyone evaluating this DAC give it a couple of hundred hours. Mine has transformed from a "yeah its pretty good" to a "yes very good" to an "outstanding DAC and now I understand what so many find remarkable about it" and that is just straight out of my Northwest Falcon Tiki with standard USB no high end power conditioner, etc. My Cavalli Liquid Carbon definitely "HELPED" but that is another subject. I am definitely getting my money's worth.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...