Lynx AES16 Sound Card: PCI verses PCIe

[jm5546]

After reading the comments and impressions of many individuals, I have decided to purchase a Lynx AES16 sound card for a music server I am building (my first). What I have not been able to determine, is which version to go with (i.e. PCI or PCIe).

 

My motherboard can support either, so sound quality is the primary factor. Which would you recommend? Also, what are the best driver versions to use with each card. I see references to "legacy drivers", yet the Lynx site offers only current drivers.

 

Thank you,

 

Frank

 

[The Computer Audiophile]

Hi Frank - Right now I am using Windows 7 with the Lynx AES16 on one machine and a AES16e on the another. I recommend the AES16 with the newest drivers and firmware if you use Windows 7.

 

If you use Windows XP then older drivers may sound best.

 

[Audio_ELF]

Chris...

 

Have you done comparisons with the RME HDSPe AES (PCIe) and HDSP AES-32 (PCI) cards?

 

Eloise

 

[bordin]

My Lynx AES16e sounds the best on Mac OSX.

 

[Audio_ELF]

Bit of a short reply ... the Lynx AES16e sounds best compared with what?

 

Obviously not compared with a Lynx AES16 as MacPro has only PCIe slots.

 

Eloise

 

[The Computer Audiophile]

Hi Eloise - I've only used the RME 9632 PCI card in my system. I'll be testing the AES-32 sometime soon.

 

Another thing people may want to consider is the longevity of the PCI interface and the ability to purchase appropriate computers with PCI slots. Some potentially good computers or motherboards may only have PCIe slots because these are the "latest and greatest."

 

[Audio_ELF]

That was my thought about PCI - how long will we be seeing PCs with them included. As you obviously know, Apple have already dropped PCI from their MacPro machines. Given that - I hope you'll be testing the HDSPe AES PCIe version - things I've read imply this may be better than the Lynx AES16e though they the Lynx PCI version was better than the RME equivalent. The RME have the advantage of allowing clock multiplying on their Word Clock input as well.

 

I think the RME card may be more expensive in USA though prices in UK are about equal (German vs USA company IIRC).

 

[manisandher]

I had a quick chat with the RME guys at the Munich High End show.

 

Apparently the PCIe version of the AES-32 has a latency of 24 samples, compared to 32 samples for the PCI version.

 

I don't believe there are any other differences in performance between the cards.

 

In any event, either of these cards will have a lower latency than any USB/firewire DAC/interface. Some people believe this is important...

 

Mani.

 

[blueixus]

There seems an inevitable tide toward PCIe. I have just put my Lynx L22 PCI card up on ebay for just this reason.

 

The Lynx products have excellent longevity so even if your current machine does have PCI the next may not.. that's progress for you!

 

[bordin]

Hi Eloise,

 

I compared my AES16e on my XP, Vista, and OSX machines and found the best performance, i.e. the lowest jitter level, on the OSX one. I tested it with both the OMZ DAC's SPDIF and the Metric Halo ULN-2's AES/EBU.

 

[Audio_ELF]

Not sure if I quite understand Bordin...

 

Are you just saying that you have a Lynx AES16e and that the best sound quality you obtained was using Mac OS X? Or did you compare a number of sound cards and the Lynx AES16e came out on top?

 

Eloise

 

[bordin]

Sorry for the confusion Eloise. I tested my AES16e card on different machines with different operating systems, namely Windows XP, Vista and Mac OS X 10.5.8 and found the AES16e on the OSX machine delivers the best sounds, compared to the other two settings.

 

Thus, not only which card to choose but also which operation system to pair with the card. IMHO, For getting the best performance out of a Lynx AES card, a Mac machine is preferable.

 

If you ask about a comparison among AES/EBU/SPDIF interface cards, then I would suggest the Weiss INT202 or the Metric Halo ULN-2 (Firewire-to-AES/EBU) over the AES16. I had a chance to do a comparison between them recently and ranked the AES16 (on OSX) at the bottom.

 

 

[bordin]

> Are you just saying that you have a Lynx AES16e and

> that the best sound quality you obtained was using Mac OS X?

 

Yes I am. My AES16e on Windows sounds muddier.

 

> Or did you compare a number of sound cards and the Lynx AES16e came out on top?

 

The Metric Halo ULN-2 (Firewire-2-SPDIF) and the Weiss INT202 came out on top. IMHO, ones of the best in the market.

 

[manisandher]

Bordin, that's interesting.

 

FWIW, I actually prefer my RME AES-32 (PCI not PCIe) to my Weiss AFI1 - both being slaved to the DAC. To me, the RME (set to 'Pro' not 'Consumer' output voltage - i.e. 5V not 2V) sounds smooth, dark and dry. The Weiss sounds edgy in comparison.

 

You might read in other threads that latency is not an issue for 'simple' 2-channel playback. But with XXHighEnd, it definitely is... and the lower, the better. Here, a PCI/PCIe interface simply trounces any USB or firewire interface - I can set the PCI AES-32 to a much lower latency than I can the firewire AFI1.

 

Mani.

 

[bordin]

Mani,

 

Thanks for the note. I wish to listen a RME card.

 

Regarding the latency effect, I still can't imagine how latency would affect S/PDIF signal generation. Bits must be perfect so that the only factor is the jitter level.

 

I guess XXHighEnd talks directly with the driver of the audio device, bypassing the kernel routines. So, it can sustain bit stream at a constant rate.

 

However, I still wonder how a smaller buffer setting makes better sounding. There must be from some properties of electronic components of the audio interface.

 

My belief is that the latency effect would exist only in the AES/EBU or S/PDIF signal in which clock is extracted from clock-embedded serial data.

 

[manisandher]

Yes, things get a little strange when you get into the realm of the ultra-low latencies that XXHE is capable of - we're talking a few samples (one user has a latency of one sample!).

 

But, the sound improves! Don't ask me how/why, but it does. One of the most striking differences (for me, at least) is that everything becomes way more 3D - the stage stretches right back, and everything becomes super focused. Cymbals sound oh so sweet. Now normally, I don't really care too much about 'sound stages' - I'm far more interested in timing/coherence and harmonics. But XXHE in ultra-low latency mode(s) is hard not to fall in love with.

 

I cannot get an ultra-low latency with either a USB or firewire interface. I can with a PCI interface.

 

Mani.

 

[hifitubes]

Does Lynx "buffer" = latency settings. When it is lowered too low playback suffers skipping? It must be as I don't see anywhere else to define latency/buffer in the Lynx control panels.

 

I too, thought latency is for recording etc.

 

Are you talking about system latency i.e. when you run the DPC Latency Checker?

 

http://www.thesycon.de/deu/latency_check.shtml

 

 

My belief is that the latency effect would exist only in the AES/EBU or S/PDIF signal in which clock is extracted from clock-embedded serial data.

 

Very interesting point, but wouldn't designs compensate for this?

 

If try playback via Kernel, WASPAI , or AISO in JRMC15 for example and lower the buffer and set your latency at 34 etc. are you getting the lowest latency possible? If yes, well listen to the results, there is a reason JRiver recommends buffers of at least .50 seconds (IIRC). Otherwise the result on most systems is unlistenable.

 

Recently they had a bug in JRMC (maybe 20 builds back?) where the audio engine would not use the large hardware buffer (for ASIO) and everyone was very confused when their audio sounded horribly choppy.

 

 

[manisandher]

hifitubes,

 

I'm not sure how things work with the Lynx card.

 

My comments apply strictly to the RME AES-32 PCI card and especially to the XXHighEnd player, which is a law unto itself. With a 'powerful' PC, you can set the latency very low, without any skipping/glitches whatsoever. (FWIW, with the buffer set to 32 samples in the RME software and 92 samples in XXHE, I can play up to 24/192 material with no glitches. Other XXHE users are able to set their buffers to significantly smaller amounts.)

 

System latency is another thing altogether and is not what we're talking about here.

 

Mani.

 

[hifitubes]

My understanding is that w/ a more powerful PC you are offloading to the CPU when you lower the buffer, these system are designed with buffers, however small, in mind. Lynx CS informed me that the buffer will have no adverse affect on an embedded clock in serial data.

 

Now with an ATOM for example, buffer underuns occur on my system if I lower my sample buffer, but I have had pretty decent cards in the past on quad-core systems. I will play around with the Lynx control panel tonight.

 

Buffering is good imo. I'm not sure why you would lower buffers and tend towards the edge of uncertainty where buffer underuns (which can be VERY subtle) may occur; sees counter-intuitive to me.