Pimping my C.A.P.S.

[manisandher]

I think Chris’s original design is great – my C.A.P.S. looks gorgeous and really sounds superb (with a TI firewire card instead of a Lynx card). Nowadays, I listen primarily through my headphone rig and tend to sit right next to the C.A.P.S. – it’s deadly silent, which is quite eerie when you’re used to noisy PCs.

 

Now ordinarily, I wouldn’t want to change a thing about the C.A.P.S. However...

 

... A few days ago, PeterSt brought out the latest version of the XXHighEnd player. With Windows 7, there are two choices of sound ‘engine’: Engine#3 (WASAPI) and Engine#4 (Kernel Streaming). Both of these place a low demand on the CPU, and with its Atom processor, either is absolutely perfect for the C.A.P.S. If this were the end of the story, I’d be a happy chappie...

 

As well as running in a low resource-intensive ‘normal mode’, Engine#4 can now run in what Peter calls ‘Special Mode’. From the XXHighEnd website:

 

“When [the ‘Special Mode’ checkbox is] ticked you will switch to an ultra low latency mode, unknown to the Pro world of (computer aided) musicians. This is so low, that a sneeze may loose some samples, so to speak. The core2duo - 2.4GHz of your humble developer goes as low as .... 5 samples, with the data rate as implied by 32bits and 176.4KHz (@ 2 channels).”

 

I don’t exactly know the mechanics behind this ‘Special Mode’, but I do know that it gives the music an ethereal feel, which is totally addictive. (For those of you with an older PS Audio power regenerator, it’s a bit like choosing the P-1 MultiwaveII option.)

 

But of course such low latencies require computing power – certainly far more than the C.A.P.S. can provide. And hence, reluctantly, I’m going to pimp my C.A.P.S.

 

The original Intel D945GSEJT will be replaced with an Intel DG41MJ motherboard, with an E8600 3.33GHz processor and 4GB of RAM. Now, the PCI riser of the original design cannot be used with this board, so a card such as the Lynx AES16 (or the RME AES32 which I recently bought to try out in the C.A.P.S.) cannot be used. However, because the TI firewire card is a half-width PCI card that can be slotted vertically into the PCI board without the need of the riser, it will fit into the new board. The new board does not have a 12V power socket, but does have a 2x12 pin power connector, so I will have to use the DC-DC supply that came with the Origen M10 case.

 

Of course, I’m violating many of Chris’s original C.A.P.S. requirements... especially no.1. The new processor will need active cooling (I think), so there will be at least one fan somewhere in the case. I can do nothing more than try to make this run as quietly as possible.

 

That’s it. My pimped C.A.P.S., which I’m hoping to have all the parts for this weekend.

 

On a final note, for those of you who are thinking about putting a C.A.P.S. together but are perhaps a little unsure of the technical knowledge required... I had never touched the inside of a PC until I started out on C.A.P.S. In any event, I think there are plenty of people here on CA who would be more than willing to help you should any issues arise.

 

Mani.

 

 

[The Computer Audiophile]

Hi Mani - It's great to see you're tweaking the C.A.P.S. system to fit your needs. There is no such thing as one size fits all.

 

I've run both engine #3 and #4 and used special mode without any issues on the server. Maybe I'm missing something. Can you give me a little more info about the resource issues?

 

Thanks Mani.

 

[manisandher]

"I've run both engine #3 and #4 and used special mode without any issues on the server."

 

That's very interesting.

 

Chris, when you have a moment, could you check what the lowest Q1 value is that you can go down to in special mode, without any hiccups during playback and between track changes? With 16/44.1 material, I'm at 75-100. With 24/176.4 material, I'm at 300-500. Some users with more powerful PCs are at 5-20. One user is at 1 - yes, a latency of 1 sample!!!

 

If your C.A.P.S. with a Lynx card is really capable of these sorts of low latencies, then the villain in my case must be the firewire card and/or Weiss interface. But I'd love to know for sure...

 

Mani.

 

[manisandher]

There's been a screwup with the order of the new motherboard, RAM and processor for my 'pimped' C.A.P.S. Actually, I think this has been a blessing in disguise, as I've decided to cancel the order all together and leave the C.A.P.S. as it is.

 

Well almost.

 

Over the weekend, I received my new RME AES-32 card and placed it into the C.A.P.S. (using the PCI riser card). I love the C.A.P.S. even more. Compared to using a firewire PCI card + Weiss AFI1, I'm getting lower latency, lower CPU usage and a much neater package to boot. And I'm getting SQ that is at least as good (it may even be better - the jury is still out).

 

If you're interested in putting a C.A.P.S. together, I would stongly recommend that stick with Chris's original design - it really is very well thought out. If you don't have a DAC, I would place a 24/192 capable PCI soundcard in the C.A.P.S. in lieu of the Lynx (or RME) card. I'm hearing good things about the ESI Juli@ and the Cantatis Overture 192...

 

Mani.

 

[The Computer Audiophile]

Thanks or the update Mani. I will still check out what you requested above on my C.A.P.S. server.

 

[blueixus]

I am about to embark on the C.A.P.S journey for two reasons.

 

1) I want a PCI equiped computer so I can use my Lynx L22, the apple solutions are for my money unacceptably big & noisy for my main listening room.

 

Currently the L22 is used in my study and I have always wondered what it would be like in my main system feeding my Audio Note Dac.

 

2) I have finally plucked up the courage to try XXHighend and after an inital period of pain and hair pulling I have been blown away by it, it seems like it was developed to sound perfect with my Audio Note Dac. ( I firmly believe that system synergy plays a big part in Computer Audio, Amarra seems to like Weiss etc etc)

 

I am keen to know if C.A.P.S is ideal for XXHighend? Any views on this would be gratefully received.

 

 

 

[manisandher]

I'm kind of reluctant to give a definite opinion until Chris has tried Engine#4 in 'Special Mode' on his C.A.P.S. (with Lynx card). I suspect that in this mode, XXHE is a little too CPU hungry for the C.A.P.S., but I would be so happy to be proved wrong. In any event, I will place as much pressure on PeterSt (I even know where he lives!) to try to lower the CPU demand in this mode. What I really want is the C.A.P.S. to be able to play a quad upsampled file with Q1 setting below 30. Then I'd be ecstatic.

 

But look, this mode is kind of 'experimental' at the moment. I'm quite happy to use XXHE in 'Normal Mode', in which case the C.A.P.S. is just perfect.

 

Mani.

 

[PeterSt]

Ok, allow me to jump in ...

 

Mani, no matter the ridiculous low latency which can be achieved here (as you know, like 22 samples and even lower as some down under crazies experienced), the cpu useage is still virtually zero. BUT :

 

It is the driver which goes bananas when not "treated right". Or as with Firewire ... in general (or IOW, don't play, just change the sample rate and/or buffer latency and there you have it).

At playing (and use Appointment Scheme 3 for it) you can see the driver operating in the first core, and the audio engine in the second. Now look at the intervals of the spikes ... both make no sense (because of the way way too long interval in relation to both the latency of the audio engine and that of the driver).

 

So, of course, XXHighEnd is incurring for it all (even without playing ? hehe), but it really is the drivers doing it to you (us).

 

FYI : Right now I am writing the release notes for 0.9y-7, with a new Kernel Streaming "Adaptive Mode" which approaches what we were doing with "Special Mode", but in a more friendly way to both you and the driver. You will notice it immediately because ticks/glitches will be out of your life, when following the rules. I assume this is not different for C.A.P.S. ...

 

bluexius ... I am really sorry for your hair, and I sure won't say it's worth exactly that. But I'm always working on improvement, and the more feedback, the better it gets. I just hope you don't end up like Mani when it comes to hair. :-))

 

Peter

 

 

 

[The Computer Audiophile]

HI Mani - I finally tested out the really low latency Q values on my C.A.P.S. system with Lynx card. I think my lowest values are pretty close to yours.

 

 

 

[manisandher]

blueixus,

 

PeterSt has just released the latest version of XXHE (0.9y-7). It has a new Kernel Streaming mode called 'Adaptive'. You have to play around a bit with your buffer setting, but it is definitely worth it.

 

With the C.A.P.S and XXHE's Adaptive Mode, I'm getting a beautifully natural sound, totally devoid of any digital nastiness. So much so that I can't listen to WASAPI any more.

 

I'm a very happy chappie, both with my C.A.P.S. (exactly as Chris designed it)... and of course XXHE!

 

Mani.

 

PS. Chris, thanks for trying 'Special Mode' for me.