Taiko Audio SGM Extreme : the Crème de la Crème

[Taiko Audio]

A few weeks back, I paid a visit to Emile at his workshop in Hengelo ... I couldn't have dreamt up a more enjoyable visit. Emile, thank you.

Thank you for your elaborate report and the positive commentary. I greatly enjoyed our time spend together as well. You placing an order for an Extreme, considering your legendary background and capabilities, is the biggest compliment I could have imagined receiving.

 

[Taiko Audio]

Would be interested to know how your extreme compares with the W20 considering one is half the price of other (?)?

It might be good to realise there is a very distinct difference between the various offerings of music servers / streamers.

The first foremost and largest difference can be found in processing power.

The highest current processing power offerings that I'm aware of are:
-The Taiko Audio SGM Extreme with 2*10 core Intel Xeon CPUs
-The Pink Faun 2.16 with 1*8 core AMD Ryzen CPU
-The Laufer Memory Player with 1* AMD Threadripper CPU, depending on the model between 8 and 32 cores

Some obvious differences between these 3 are: The Extreme and Pink Faun are both fanless cooled and linear powered. The Memory Player uses fans for cooling and is SMPS powered. The Extreme and Memory Player are both running on a Windows operating system, for music playback the Extreme uses Roon, the Memory Player uses Jriver Media Center. The Pink Faun runs on Audiophile Linux and uses Roon for music playback. The Memory Player is the only one offering built in CD-ripping to what appears to be a proprietary 64 bit format and it has some unique in house designed software solutions improving playback performance. The Pink Faun focuses on hardware clock quality where the Extreme and the Memory Player both address this by other means (neither of these benefit from upgrading clocks). The Extreme offers PCIe storage up to 24TB, I think The Memory Player does too now, The Pink Faun offers SSD storage up to 8TB. The Pink Faun can be stacked to improve performance, the Extreme and Memory Player do not benefit from being stacked. Pricing wise The Extreme is more expensive then the Pink Faun, unless you buy 2 and stack them. The Memory player ranges from being the cheapest of all 3 to being the most expensive depending on the model.

Then there is a range of "middle CPU power offerings", usually using a 4-6 core Intel I7 CPU, I'll list a few of the better known ones:
-Sound Galleries SGM 2015 / EVO
-Lampizator Super Komputer
-Antipodes CX/EX

And then we have a very large selection of low CPU power offerings, using Intel Atom or other low power embedded CPU types, they all run a variation of a Linux operating system and offer Roon, Linux MPD (Music Player Deamon), UPNP or their own proprietary playback solution. Some better known examples:
-Aurender
-Innuos
-Melco
-Auralic
-432 EVO

The high CPU power offerings tend to cost more because obviously the digital side of the hardware is more expensive, they need beefier power supplies, larger / more elaborate cooling solutions and larger cases to fit everything. As always in High End the chassis can be the single most expensive part of it all, and some consider its build and finish quality to be a fundamental part of why we call it High End.

 

[Alinous]

It might be good to realise there is a very distinct difference between the various offerings of music servers / streamers.

The first foremost and largest difference can be found in processing power.

The highest current processing power offerings that I'm aware of are:
-The Taiko Audio SGM Extreme with 2*10 core Intel Xeon CPUs
-The Pink Faun 2.16 with 1*8 core AMD Ryzen CPU
-The Laufer Memory Player with 1* AMD Threadripper CPU, depending on the model between 8 and 32 cores

Some obvious differences between these 3 are: The Extreme and Pink Faun are both fanless cooled and linear powered. The Memory Player uses fans for cooling and is SMPS powered. The Extreme and Memory Player are both running on a Windows operating system, for music playback the Extreme uses Roon, the Memory Player uses Jriver Media Center. The Pink Faun runs on Audiophile Linux and uses Roon for music playback. The Memory Player is the only one offering built in CD-ripping to what appears to be a proprietary 64 bit format and it has some unique in house designed software solutions improving playback performance. The Pink Faun focuses on hardware clock quality where the Extreme and the Memory Player both address this by other means (neither of these benefit from upgrading clocks). The Extreme offers PCIe storage up to 24TB, I think The Memory Player does too now, The Pink Faun offers SSD storage up to 8TB. The Pink Faun can be stacked to improve performance, the Extreme and Memory Player do not benefit from being stacked. Pricing wise The Extreme is more expensive then the Pink Faun, unless you buy 2 and stack them. The Memory player ranges from being the cheapest of all 3 to being the most expensive depending on the model.

Then there is a range of "middle CPU power offerings", usually using a 4-6 core Intel I7 CPU, I'll list a few of the better known ones:
-Sound Galleries SGM 2015 / EVO
-Lampizator Super Komputer
-Antipodes CX/EX

And then we have a very large selection of low CPU power offerings, using Intel Atom or other low power embedded CPU types, they all run a variation of a Linux operating system and offer Roon, Linux MPD (Music Player Deamon), UPNP or their own proprietary playback solution. Some better known examples:
-Aurender
-Innuos
-Melco
-Auralic
-432 EVO

The high CPU power offerings tend to cost more because obviously the digital side of the hardware is more expensive, they need beefier power supplies, larger / more elaborate cooling solutions and larger cases to fit everything. As always in High End the chassis can be the single most expensive part of it all, and some consider its build and finish quality to be a fundamental part of why we call it High End.

It might be good to realise there is a very distinct difference between the various offerings of music servers / streamers.

The first foremost and largest difference can be found in processing power.

The highest current processing power offerings that I'm aware of are:
-The Taiko Audio SGM Extreme with 2*10 core Intel Xeon CPUs
-The Pink Faun 2.16 with 1*8 core AMD Ryzen CPU
-The Laufer Memory Player with 1* AMD Threadripper CPU, depending on the model between 8 and 32 cores

Some obvious differences between these 3 are: The Extreme and Pink Faun are both fanless cooled and linear powered. The Memory Player uses fans for cooling and is SMPS powered. The Extreme and Memory Player are both running on a Windows operating system, for music playback the Extreme uses Roon, the Memory Player uses Jriver Media Center. The Pink Faun runs on Audiophile Linux and uses Roon for music playback. The Memory Player is the only one offering built in CD-ripping to what appears to be a proprietary 64 bit format and it has some unique in house designed software solutions improving playback performance. The Pink Faun focuses on hardware clock quality where the Extreme and the Memory Player both address this by other means (neither of these benefit from upgrading clocks). The Extreme offers PCIe storage up to 24TB, I think The Memory Player does too now, The Pink Faun offers SSD storage up to 8TB. The Pink Faun can be stacked to improve performance, the Extreme and Memory Player do not benefit from being stacked. Pricing wise The Extreme is more expensive then the Pink Faun, unless you buy 2 and stack them. The Memory player ranges from being the cheapest of all 3 to being the most expensive depending on the model.

Then there is a range of "middle CPU power offerings", usually using a 4-6 core Intel I7 CPU, I'll list a few of the better known ones:
-Sound Galleries SGM 2015 / EVO
-Lampizator Super Komputer
-Antipodes CX/EX

And then we have a very large selection of low CPU power offerings, using Intel Atom or other low power embedded CPU types, they all run a variation of a Linux operating system and offer Roon, Linux MPD (Music Player Deamon), UPNP or their own proprietary playback solution. Some better known examples:
-Aurender
-Innuos
-Melco
-Auralic
-432 EVO

The high CPU power offerings tend to cost more because obviously the digital side of the hardware is more expensive, they need beefier power supplies, larger / more elaborate cooling solutions and larger cases to fit everything. As always in High End the chassis can be the single most expensive part of it all, and some consider its build and finish quality to be a fundamental part of why we call it High End.

Hello I would like to know the difference with the extreme server and SOTM SQ1000 with clock 10 Mhz, someone who made the comparison?
The Taiko extreme tempts me

 

[Taiko Audio]

Hello I would like to know the difference with the extreme server and SOTM SQ1000 with clock 10 Mhz, someone who made the comparison?
The Taiko extreme tempts me

I'm unaware of anybody comparing these head to head but I would not expect them to perform at the same level. It's quite a different price range as well.

 

[Alinous]

[QUOTE = "Taiko Audio, post: 614136, membre: 9216"] Je ne connais personne qui compare ces tête à tête mais je ne m'attendrais pas à ce qu'ils se produisent au même niveau. C'est également une gamme de prix très différente. [/ QUOTE]
Thank you, wanted to appreciate the benefit in terms of SQ. I understand that there is a difference in price and quality, I will make the investment, I first wait for the delivery of the select dac with the renderer.

 

[dminches]

It might be good to realise there is a very distinct difference between the various offerings of music servers / streamers.

The first foremost and largest difference can be found in processing power.

The highest current processing power offerings that I'm aware of are:
-The Taiko Audio SGM Extreme with 2*10 core Intel Xeon CPUs

Without giving away your design secrets, if the Extreme is just running Roon Server (and not HQ Player) why is there a need for so much computing power? I would think the processors are running well under 10% the majority of the time.

Thanks.

 

[howiebrou]

It might be good to realise there is a very distinct difference between the various offerings of music servers / streamers.

The first foremost and largest difference can be found in processing power.

The highest current processing power offerings that I'm aware of are:
-The Taiko Audio SGM Extreme with 2*10 core Intel Xeon CPUs
-The Pink Faun 2.16 with 1*8 core AMD Ryzen CPU
-The Laufer Memory Player with 1* AMD Threadripper CPU, depending on the model between 8 and 32 cores

Some obvious differences between these 3 are: The Extreme and Pink Faun are both fanless cooled and linear powered. The Memory Player uses fans for cooling and is SMPS powered. The Extreme and Memory Player are both running on a Windows operating system, for music playback the Extreme uses Roon, the Memory Player uses Jriver Media Center. The Pink Faun runs on Audiophile Linux and uses Roon for music playback. The Memory Player is the only one offering built in CD-ripping to what appears to be a proprietary 64 bit format and it has some unique in house designed software solutions improving playback performance. The Pink Faun focuses on hardware clock quality where the Extreme and the Memory Player both address this by other means (neither of these benefit from upgrading clocks). The Extreme offers PCIe storage up to 24TB, I think The Memory Player does too now, The Pink Faun offers SSD storage up to 8TB. The Pink Faun can be stacked to improve performance, the Extreme and Memory Player do not benefit from being stacked. Pricing wise The Extreme is more expensive then the Pink Faun, unless you buy 2 and stack them. The Memory player ranges from being the cheapest of all 3 to being the most expensive depending on the model.

Then there is a range of "middle CPU power offerings", usually using a 4-6 core Intel I7 CPU, I'll list a few of the better known ones:
-Sound Galleries SGM 2015 / EVO
-Lampizator Super Komputer
-Antipodes CX/EX

And then we have a very large selection of low CPU power offerings, using Intel Atom or other low power embedded CPU types, they all run a variation of a Linux operating system and offer Roon, Linux MPD (Music Player Deamon), UPNP or their own proprietary playback solution. Some better known examples:
-Aurender
-Innuos
-Melco
-Auralic
-432 EVO

The high CPU power offerings tend to cost more because obviously the digital side of the hardware is more expensive, they need beefier power supplies, larger / more elaborate cooling solutions and larger cases to fit everything. As always in High End the chassis can be the single most expensive part of it all, and some consider its build and finish quality to be a fundamental part of why we call it High End.

very interesting Emile and thank you for your description. It certainly helps me as often it is difficult to see the differences between the various offerings. Naturally sound quality, reliability, ease of use and value for money are important criteria too.

 

[joaovieira]

Without giving away your design secrets, if the Extreme is just running Roon Server (and not HQ Player) why is there a need for so much computing power? I would think the processors are running well under 10% the majority of the time.

Thanks.

Extreme runs JPlay as well and Roon.

 

[dminches]

Extreme runs JPlay as well and Roon.

Does jplay do upsampling and upconverting and require a lot of computing power?

 

[Taiko Audio]

Without giving away your design secrets, if the Extreme is just running Roon Server (and not HQ Player) why is there a need for so much computing power? I would think the processors are running well under 10% the majority of the time.

Thanks.

I'll refer to 2 earlier posts made in this thread:

The issue here is the "chip" is never idling along but in fact quite busy, all the time. Memory chips are entirely refreshed every 7.8 microseconds (JEDEC standard), more frequently at higher temperatures, this is also when memory draws its maximum current, just as an example to highlight it will never be idle, as next to this it will constantly perform read/write operations anyway. The cpu is constantly executing instructions. The heavier the OS, the fancier the user interface, the busier it will be. Therefor reducing OS footprint reduces activity, by reducing Memory read/writes, general I/O and cpu queues. And trust me, those queues are never empty. Executing instructions also comes with delays (wait states) (latency) which means more "chip" active time. The fancier the music playback software, Roon being about the fanciest currently, the more processing required. Next to reducing OS overhead from the software side, you can also increase memory bandwidth, cpu processing power and storage speed. Counter intuitively this actually lowers the amount of Active Time and the actual amount of RFI/EMI noise versus time, but the current consumption peaks are much higher which requires a beefier power supply (expensive). This is where you enter the trade off grey area, where you have low computing power systems with high queue depths versus high computing power systems with low queue depths. How this equates into the overall noise level of the system is quite dependant on the power supply design, radiation patterns and filtering.

Hi Bill,
It’s a different concept. In the original SGM we focused on creating a machine excelling in upsampling. You could view upsampling as a way to reduce the influence of DAC filters. For example DSD upsampling with the good old T+A as you experienced would push low pass filter requirements up to very high frequencies.
As Edward pointed out DAC quality (and filter quality) had improved significantly, to the point there are now quite a number of DAC’s not benefitting from upsampling.
For this generation of DAC’s the Extreme is the better partner, it’s focus is having as low processing queue depths as possible, with heavy cpu load upsampling queues will always be high which leads to different design choices.
For example the often discussed latency. This is obviously not the latency between pressing play and hearing sound, but a latency existing in computer systems being defined by simple copper trace lengths (distance), time to enter specific hardware states, time it takes to get access to section of memory or disk, etc etc. Every instruction or set of instructions being executed takes an amount of time which you can (simplified) define as latency+execution time.
With bit perfect playback execution times and data reads/writes are low and latencies become significant.
With heavy upsampling execution times are high and latency becomes a relatively small factor.
Now please note that latencies are not all of a sudden the only factor in play, theyre just relatively more important now, even more so when not doing heavy processing.
So why are we using such a heavy, high power machine which can provide an entire medium sized company’s IT services for bit perfect playback?
Because it’s able to maintain low queue depths everywhere (cpu/ram/drives) all the time, even when it is running Roon software which is not very light at all, the amount of processing being executed is minimal.
This leads to a very even, almost constant, current draw, ofcourse the massive powersupply helps smoothing this out even more. Current fluctuations come with dips, overshoots, ringing etc. This means the server’s influence on its surroundings is reduced in many ways, powerline ringing and harmonics, EMI and RFI emissions.
In fact system influence has always been important to us, from the first SGM we have always made sure introducing one into your system does not degrade other system components. But now taken to the Extreme as in it will not degrade your other system components, no matter how high level they are, while playing music.
I’m a bit short on time but will try to elaborate on this subject more extensively in the next few days.

 

[dminches]

Thanks Emile. I did read those before but they didn’t register until now.

 

[Taiko Audio]

very interesting Emile and thank you for your description. It certainly helps me as often it is difficult to see the differences between the various offerings. Naturally sound quality, reliability, ease of use and value for money are important criteria too.

Sonic differences for higher processing power systems can be found in:
-Increased soundstage size in both depth and width, extension beyond the boundaries set by the loudspeakers and room / development / envelopment / immersion.
-Overall higher detail levels, more resolving power, especially when there's a lot going on at the same time.
-Authority, bass depth and slam, frequency extreme extension.

It interestingly sounds very similar to moving up to a higher powered amplifier, or from stereo to mono blocks.

I'd say the difference is most pronounced with large scale complex orchestra. If you have a small room and mainly listen to small ensembles the difference is going to be less obvious.

 

[Taiko Audio]

Thank you, wanted to appreciate the benefit in terms of SQ. I understand that there is a difference in price and quality, I will make the investment, I first wait for the delivery of the select dac with the renderer.

If I'm allowed to make a suggestion there I'd consider getting the MSB Pro USB module to go with that

 

[Blackmorec]

Sonic differences for higher processing power systems can be found in:
-Increased soundstage size in both depth and width, extension beyond the boundaries set by the loudspeakers and room / development / envelopment / immersion.
-Overall higher detail levels, more resolving power, especially when there's a lot going on at the same time.
-Authority, bass depth and slam, frequency extreme extension.

It interestingly sounds very similar to moving up to a higher powered amplifier, or from stereo to mono blocks.

I'd say the difference is most pronounced with large scale complex orchestra. If you have a small room and mainly listen to small ensembles the difference is going to be less obvious.

Doesn’t the result also depend on which operating system you’re running?

I would have thought that an OS designed to run efficiently on a lower power CPU (assuming there is such an OS) may generate less power supply demand and develop less noise from the CPU? I‘m simply basing this on the logic that you only need a hefty CPU if that’s what the OS requires. Am I wrong?

 

[Taiko Audio]

Doesn’t the result also depend on which operating system you’re running?

I would have thought that an OS designed to run efficiently on a lower power CPU (assuming there is such an OS) may generate less power supply demand and develop less noise from the CPU? I‘m simply basing this on the logic that you only need a hefty CPU if that’s what the OS requires. Am I wrong?

It's very likely all server / streamer vendors tailor the OS they run to the hardware used.

 

[Blackmorec]

It's very likely all server / streamer vendors tailor the OS they run to the hardware used.

Or maybe the other way around.....select the hardware according to the OS they run? Wouldn’t that make more sense In terms of matching exactly what the OS needs and obtaining optimum SQ. For example would it be sensible to upgrade the power supply to run a very powerful CPU if the OS used runs very efficiently on a lower power CPU? Would such an OS benefit from a beefier CPU if it wasn’t particularly demanding of CPU resources? I guess what I’m asking is does the higher power CPU deliver anything in terms of SQ enhancements if the extra power isn’t actually needed....will the improvements in CPU power offset the extra noise and heat generated if the extra power isn’t utilized at all.

 

[nonesup]

Emile, I understand that you bought an EtherREGEN from an owner who did not like it. Are you already testing with him?

 

[Taiko Audio]

Emile, I understand that you bought an EtherREGEN from an owner who did not like it. Are you already testing with him?

That owner has send it to Amir who wanted to measure one for the AS forum, it has now returned and he wants to give it one more try before shipping it to me. If he decides to keep it after all I'm going to have to wait for that December batch to ship.

 

[Taiko Audio]

Or maybe the other way around.....select the hardware according to the OS they run? Wouldn’t that make more sense In terms of matching exactly what the OS needs and obtaining optimum SQ. For example would it be sensible to upgrade the power supply to run a very powerful CPU if the OS used runs very efficiently on a lower power CPU? Would such an OS benefit from a beefier CPU if it wasn’t particularly demanding of CPU resources? I guess what I’m asking is does the higher power CPU deliver anything in terms of SQ enhancements if the extra power isn’t actually needed....will the improvements in CPU power offset the extra noise and heat generated if the extra power isn’t utilized at all.

There is a limit to which extent you can reduce the OS footprint. You will need basic functionality, networking, drivers for all hardware components, music playback software etc. There is quite a bit going on, drivers create interrupts demanding CPU time to execute instructions, the wait time depends on priority assigned to the processes and availability, all processes are spending a certain amount of time waiting for a CPU core to become available to execute instructions. There are a lot of instructions being executed all the time, even on a minimalist OS.

One thing where you can notice clear differences is sensitivity to the networking environment, just mentioning that as it's a hot topic lately. Networking adapters can offload a certain amount of processing to the CPU, you can manage this, choose what processing to do where, at what time, at which priority. It's a balancing act. More CPU capacity means instructions are executed faster, you can have higher cpu speeds, for a higher serial processing capacity or more cores for higher parallel processing capacity, the latter being generally more favourable, but harder to implement properly. A few practical examples of what increased CPU capacity can facilitate is better Qobuz/Tidal quality relative to local file playback, inaudibility of having your Roon remote app open / browsing albums while listening. It generally can reduce the influence of your networking environment, if implemented correctly.

 

[Taiko Audio]

Just for general interest a display of Ethernet error correction, note these stages of error correction are not all executed on all network adapters / devices.