Taiko Audio SGM Extreme : the Crème de la Crème

[Taiko Audio]

I could find that there is a Refresh Period (tREF) in the memory setting, does it help?

It can go by different descriptions depending on bios / motherboard manufacturer.
The relevant settings are DRAM REF Cycle Time and DRAM Refresh Interval.

But I cannot recommend just changing them without a very solid testing protocol. As modifying these can easily lead to data corruption and/or increased error rates. Your motherboard executes something called a "MCH FULL CHECK" when you install new memory modules in your system and sets these values automatically based on your memory module capabilities. We select our memory modules and test them extensively for lowest overall latency and error rates while modifying these parameters. For the daring hobbyist its a way to very significantly (more then altering CAS/RAS latencies) impact overall memory access latency and power usage. But AGAIN, you can easily make things worse in stead of better, worst case being a complete operating system reinstall.

 

[darkfrank]

It can go by different descriptions depending on bios / motherboard manufacturer.
The relevant settings are DRAM REF Cycle Time and DRAM Refresh Interval.

But I cannot recommend just changing them without a very solid testing protocol. As modifying these can easily lead to data corruption and/or increased error rates. Your motherboard executes something called a "MCH FULL CHECK" when you install new memory modules in your system and sets these values automatically based on your memory module capabilities. We select our memory modules and test them extensively for lowest overall latency and error rates while modifying these parameters. For the daring hobbyist its a way to very significantly (more then altering CAS/RAS latencies) impact overall memory access latency and power usage. But AGAIN, you can easily make things worse in stead of better, worst case being a complete operating system reinstall.

Thanks very much for the detail explanation.
I wonder those audiophile highly recommended Elpidia chips DDR3 memory back in the days also have such good characteristics of low cycle length and long refresh interval.

 

[Taiko Audio]

Thanks very much for the detail explanation.
I wonder those audiophile highly recommended Elpidia chips DDR3 memory back in the days also have such good characteristics of low cycle length and long refresh interval.

Could be!

Another interesting feature is an option called "underfill" where they fill the void in between memory chips and the pcb with some kind of epoxy. This increases shock resistance and susceptibility to vibrations for use in extreme environments. We consider our audio systems to be harsh environments

 

[Bodhi]

In case of a networked connection there are several options, the most convenient one being through Roon.

Ok, good to know. Though what would be the highest resolution/fidelity option? I would have thought a wired connection over ethernet to a dac/renderer? But that's where it gets confusing, as most high end dacs don't actually function as a renderer.

 

[Taiko Audio]

Ok, good to know. Though what would be the highest resolution/fidelity option? I would have thought a wired connection over ethernet to a dac/renderer? But that's where it gets confusing, as most high end dacs don't actually function as a renderer.

It really depends on the implementation, feedback on network connected MSB dacs is very good, competitive to USB, on lampizator the USB input seems to be preferred so far. But their ethernet module is 100Mbit which makes network bridging difficult as bridging requires the same ethernet speeds. To avoid another tiresome ethernet vs USB discussion: YMMV

 

[microstrip]

JPLAY Femto referenced to above is playback software, you can use it as a local output for Roon (however you will not extract the maximum potential from it then), or standalone as a very minimalistic UPNP library manager and renderer. You can control it with UPNP control apps, the best functioning one is Bubble upnp (on android). It performs especially well with XMOS based USB receivers. Its for the purist who favours sound quality over user interface and does not mind tweaking a little.
Easily confused for JCAT Femto addon cards

Thanks - are you saying that member Esotar has JPlay Femto but none of the cards as he posted?

 

[Taiko Audio]

Thanks - are you saying that member Esotar has JPlay Femto but none of the cards as he posted?

He has both

 

[microstrip]

He has both

So my question is pertinent - do you endorse using these cards in the Extreme? Do they improve sound quality in your opinion?

 

[Bodhi]

It really depends on the implementation, feedback on network connected MSB dacs is very good, competitive to USB, on lampizator the USB input seems to be preferred so far. But their ethernet module is 100Mbit which makes network bridging difficult as bridging requires the same ethernet speeds. To avoid another tiresome ethernet vs USB discussion: YMMV

That is helpful, thanks. Carte blanche, I think I would pair a Boulder 2120 with that server, as I haven't found a better dac/streamer optimized for UPnP/DLNA protocol, nor for that matter have I seen a better color screen/UI on any other similar product. As a pure dac though, I agree MSB is very good.

 

[Taiko Audio]

So my question is pertinent - do you endorse using these cards in the Extreme? Do they improve sound quality in your opinion?

For USB it was more of a sideways move, mainly a different sonic character, BUT JCAT has released a firmware update for their USB card that I have not been able to test yet, supposedly the SQ increase is significant.
For networking I do get the best results by far using STP when using the right SFP modules. The JCAT card could be an upgrade for directly connecting to an ethernet renderer. But I'm very sorry to have to tell you again that I still have to test that.

This is one of the reason's though that we made the Extreme so versatile in expendability. Anyone capable of removing and/or inserting PCIe cards can add or remove expansions. We also offer a try out service, we can fit and/or ship a card for you to try and you can return it if it does not work for you.

 

[Taiko Audio]

That is helpful, thanks. Carte blanche, I think I would pair a Boulder 2120 with that server, as I haven't found a better dac/streamer optimized for UPnP/DLNA protocol, nor for that matter have I seen a better color screen/UI on any other similar product. As a pure dac though, I agree MSB is very good.

I'm unfamiliar with it but it sure looks the part!

 

[microstrip]

For USB it was more of a sideways move, mainly a different sonic character, BUT JCAT has released a firmware update for their USB card that I have not been able to test yet, supposedly the SQ increase is significant.
For networking I do get the best results by far using STP when using the right SFP modules. The JCAT card could be an upgrade for directly connecting to an ethernet renderer. But I'm very sorry to have to tell you again that I still have to test that.

This is one of the reason's though that we made the Extreme so versatile in expendability. Anyone capable of removing and/or inserting PCIe cards can add or remove expansions. We also offer a try out service, we can fit and/or ship a card for you to try and you can return it if it does not work for you.

My main interest is just on the direct connection to an ethernet renderer - the Vivaldi upsampler.

Could you decode "For networking I do get the best results by far using STP when using the right SFP modules. ", if possible giving an example? Thanks!

 

[nonesup]

As I told, I sold 7 EVO and 11 Extreme in Korea.
Korea is very very very poor country.

I guess you're kidding, right?

 

[Taiko Audio]

My main interest is just on the direct connection to an ethernet renderer - the Vivaldi upsampler.

Could you decode "For networking I do get the best results by far using STP when using the right SFP modules. ", if possible giving an example? Thanks!

Single mode fibre, 1550nm, 80km or even further reach sound best. I do prefer using regular ethernet over using SFP with multi mode or single mode short distance SFP's. I find those dynamically restricted sounding.

 

[Esotar]

I guess you're kidding, right?

No, it's really true~!!!

Trust me~!!!!

 

[microstrip]

Single mode fibre, 1550nm, 80km or even further reach sound best. I do prefer using regular ethernet over using SFP with multi mode or single mode short distance SFP's. I find those dynamically restricted sounding.

Interesting. Are you addressing this type of connection between the server and the DAC? https://fibertronics-store.com/SFP-...bps-Singlemode-1550nm-80km-SFP-ZX-SM-0180.htm Using long fiber links is really advantageous?

 

[Taiko Audio]

Interesting. Are you addressing this type of connection between the server and the DAC? https://fibertronics-store.com/SFP-...bps-Singlemode-1550nm-80km-SFP-ZX-SM-0180.htm Using long fiber links is really advantageous?

I use these:
https://www.startech.com/Networking-IO/sfp/modules/1000base-zx-sfp-module~SFP1000ZXST

Fibertronics store sfp’s are being used with satisfactory results too.

I personally prefer USB or AES/EBU server to dac connections, ymmv Using STP networking between dac and server requires the use of a FMC with powersupply, high quality fmc’s and powersupplies dont come cheap.

Unless I missed something, are there dac’s offering direct stp in?

 

[Folsom]

Back in 2015 I was sceptical too. How can digital equipment delivering an un-corrupted data stream, buffered and timed with precision, sound any different ?

The interesting consistent empirical observation is that the listener heard sound quality coming from digital equipment is very vibration sensitive and of a similar order of magnitude to the vibration sensitivity of analog equipment.

Another consistent observation is the sound quality is heavily influenced by OS activity and processes, the delivered bits are the same, they are all buffered and timed with precision so what’s going on ?

3 years ago Emile and I invested in acquiring quite a bit of RF detection and spectrum analysis gear from these folks https://www.aaronia.com/products/antennas/

What we can observe is that the RF soup coming from all of the clocks and oscillators gets everywhere and is very difficult to dramatically attenuate. The RF noise is transmitted over the air and piggybacks on every cable, conductor and ground plane. We can also observe that CPU and RAM activity changes the observable RF emissions.

I would humbly propose that the sound quality differences we are hearing are being transmitted by RF noise emissions emanating from the myriad clocks and oscillators in digital equipment, and being detected by the analog components and cabling in the playback chain.

All of the hardware and software features which Emile has implemented deliver sonic differences that are easily perceivable and consistent with a mechanism of RF transmission of noise, both SQ enhancing, and SQ degrading. The design strategy is to shape the noise, minimizing the SQ degrading RF noise and to amplify the SQ enhancing RF noise.

I am more than happy to have a phone call with you to further discuss transmission mechanisms, but let's keep it off line and not clutter this thread with theoretical discussions

Oh, I don't doubt it sounds different at all. My point is that everytime something like this gets measured, it shows that it has problems that change the signal to produce that difference in sound, not retained more of it.

The simple answer here is to use a fiber card out to skip all the interference that's parasitic and doesn't exist prior to mixing with the output. The thing to fix is where it's delivered. If the maker of this wanted to work on a refined receiver for fiber, that'd be a great place to start IMO if you want the original signal. (I'd be willing to work on that project)

From what I can tell however what I'm reading about all the other things, is that DSD is a pretty weak format because it's so subject to computer problems. I know people like it, but I'm still yet to be sold.

Why not set the CPU clock to the most cycles with lower bandwidth and voltage, if you're already trying to get memory refresh times set? You may even be able to use a cheaper CPU to get more cycles.

If you really are experiencing so many changes in sound, I would also suggest doing field control inside the unit. There's also numerous other things that can be done. IMO a lot of the approaches here are "hail-marrys" because even a lot of engineers aren't really sure what's going on fully.

 

[adamaley]

No, it's really true~!!!

Trust me~!!!!

Great to see that dealers in North Korea are able to get these luxury goods to the people without incurring the wrath of the communist party ;-)

 

[nonesup]

Great to see that dealers in North Korea are able to get these luxury goods to the people without incurring the wrath of the communist party ;-)

Je, je, je