XXHighEnd

Hi Peter,

I use Scheme1 core appointment option, but I looked at core affinity for Engine3 in task manager and it shows CPU0 and CPU1, is this ok? I thought it should use CPU1 only...

Greets.
Marcin

Marcin, with 4 cores it uses (the 2nd core,) core-1 yes, but now I don't know whether that's different for a 2 core CPU (it is even difficult to derive this from my own coding). But I assume it's the same for 2 and 4 core CPUs.

For sure nothing changed from 0.9z-2 to z-3, or it is by accident (bug).

Peter

Now I'm using quad-core CPU and for Scheme-1 I see that XXHE works on CPU0 and CPU1. Any advices how to solve this?

Greets,
Marcin

Seems impossible to me ...
Therefore no ideas at this moment.

I assume you can get it to 1 (and 1 only of course) manually (TaskManager) ?

Or starting off with all 4 and you appoint #1 manually, does it do that ?

Yes, I can do this manually, but it's very annoying...

Here's how it looks. Maybe if I switch between different Schemes...

:wtf: :oops: :wtf:

I now see that it's not set back after a trackload. And in your case with SFS=12 it will keep in that stage.

Careful now, because this may the "your" reason why SFS=12 sounds better.

Thanks for noticing this Marcin !
Peter

Always at your services :)

I noticed sth strange again, if I set the SFS size to 200MB, then the Engine3 is on CPU1, as it should be. I went back to 12MB and it's on CPU0 and CPU1 again.

That's what I said :


But the remainder is a little different from what I thought;
It is set back allright, but with such a low SFS it immediately gets back in the "track load state". But ...

This resets when the new part plays, and not when the load finished. :fool:
So, with some tricks I just managed to change it back after the load finishes and you won't be able to see it at both cores now (not with TaskManager and "Set Affinity").

All 'n all this always has been so, and since 14 seconds is reserved for the track load, it always lasted 14 seconds (now only the 0.5 secs or whatever it is for SFS=12 of the real load).

So it was not the best eh ? :nea:
Peter

It's fine now I prefer high SFS values. dCS sounds like cr*p compared to my Cantatis, haha

Peter, is there any way it can be fixed - I mean Scheme-1 staying at CPU1 after conversion? (for low SFS values)
It's the same for SFS below 50 MB or so, even if I manually assign the Engine3 to CPU1, it goes back to CPU0 and CPU1 after few seconds.

Greets,
Marcin

Oh, didn't I say it was fixed already ? The only thing is you don't have that version yet (and I am late, as usual :)).

:)

THe same thing applies to No-Appointment setting. Eventhough I assign it to CPU1, it goes back to CPU0 and CPU1.

 :offtopic:

Marcin, how does the dCS compare with the Cantatis now that the dCS in being fed by an RME AES-32? Just interested...

(I've noticed that Peter is getting a bit grumpy in his old age, so if you have a lot to say then maybe you should start a new 'PC/DAC interface' topic or something.)

Mani.

Puh, watch me !                                                                                                                       

Haha! Mani, dCS sounded wrong fed by Cantatis (coaxial or optical) and Cantatis was better as a DAC than as a tranport combined with dCS - strange, but true. The same was with Sony drive, but with RME - oh boy, Cantatis sounds like cr*p compared to dCS, so you see, nothing is plain, but I'm sure of one thing - a DAC should be fed by an outside source, totally isolated from the PC (Firewire?). I experimented one more time with undervolting and underclocking of CPU, but I accomplished that with an app, so I got to compare it right away to normal and overclocked settings. I underclocked my CPU to 400MHz (for each core) and lowered the voltage to 0.450V!!! (1.325 is default, and 1.5V overclocked). I also changed CPU coligation for all processes to CPU2 (third core), XXHE scheme 1 (processing on CPU0, 'sound' on CPU1), cpu affinities for all non-audio devices on CPU2, RME, PCI Bus and PCI I/0 Communications BUS on CPU3. That was the best sound I have ever heard from my system and so much better than default or overclocked setting. So I think our pc-audio playback is going to be better and better every year, because of increasing performance of new CPUs and their smaller TDPs and better architectures.

But nothing is perfect right? With 4x400MHz and quad upsampling (not QAP) I get a random click from time to time, because there is lack of horsepower on CPU0, where all the processing/loading is done. If the processing was spread around all of the cores (except CPU1, where is playback engine), then I guess it wouldn't be a problem. The app I used for underclocking/undervolting is called K10Stat and allows for much better adjustment than you have available in BIOS. Normally I couldn't go below 0.85V, even running a single core! But I guess that the app is for AMD CPUs only...

Let me know what you think
Best regards,
Marcin

I can affirm this. I used a RME HDSP9632 as spdif source some years ago. Sounded like cr*p. A decent CDP was faaaar better but now with highface and the new version of xxHE PC has become better. Don't know where this will go with the Legato but I expect a significant step up (still waiting for it to come).

Marcin, I think Peter is rubbing off on you - you're starting to talk double-Dutch :)

So, you're convinced that the DAC should be fed by an outside source (e.g. firewire), and yet the source that you say sounds great (the RME) is internal (PCI). Hmmm... Care to explain???

FWIW, I can feed my ADC/DAC with an internal (RME AES-32 - PCI) or external (Weiss AFI1 - firewire) source. Both provide a dual-wire AES input to, and take a BNC wordclock from, the ADC/DAC. I'm not sure that I hear any obvious advantage of the external over the internal... What should I be listening out for?

Mani.

Mani,

RME and Cantatis are PCI-based, that's true and RME is much much better via dual AES, but it could be miles better if you undervolt your CPU by a large margin, like I did. The conclusion is very obvious here - the SQ depends on PC's EMI/RFI levels or whatever it is that is caused by high frequencies and high voltage of CPU (high TDP). I have tried it in the past and had the same results. I am 100% sure that you can't have the best quality using internal/PCI-based interface, unless you do some serious undervolting/underclocking of your CPU or provide external PSU to that interface and isolate it from motherboard's power, if that's possible. Mani, you should try a FW cable without a power leg in order to eliminate the power transmission from PC to DAC. It will cope with potential ground loop issues, but won't address noise issues, so it's not a perfect solution, but I guess it's the best at the moment. The bottom line is that PC audio is going to better and better every year, because CPUs and other components are getting faster and require less power (introduce less interference) at the same time, but I wouldn't build a high-end system based on internal interface - which is fed by a noisy computer power. That's why RME and dCS are going bye-bye.

Cheers,
Marcin

Yep, that's what I'm doing with the Weiss - I use a 6pin-4pin adaptor. I agree that this doesn't address the noise issues, but that's exactly where AES3 comes in - it should provide common mode noise attenuation (30dB, if I understand the AES3 standards correctly).


Ah, OK. To be replaced by what?

Mani.

Weiss DAC202? I don't see any alternatives right now.


Still, PC affects SQ the most - try the undervolting/underclocking yourself and you will know what I'm talking about. The lowest Vcore - the best SQ.