Here is a comprehensive stacked report from multiple write-ups by Romaz from the WTBF tread about the Taiko Audio's Extreme music server, that is simply too good to pass...
Romaz writes. "I received my Extreme exactly 2 weeks ago and I thought I'd provide some personal observations and a few comparative assessments. As always, the following comments represent opinions based on my own personal sensitivities and biases and so, of course, YMMV. Because of the posting limitations imposed by WBF, this reporting will be divided into multiple parts.
As most of my components are black, when I first contemplated ordering an Extreme, I was of the mindset that I would want a black one but during my visit to Emile's factory in Hengelo, I found myself infatuated with the silver chassis. With the contrasting black heatsinks making up the sides of the chassis, I found the silver Extreme strikingly elegant and so I went with it. It turns out there are 2 varieties of silver available (brushed vs bead blasted) and while I could live with either one, as Emile told me the brushed one hid scratches better, I went with brushed (the one on the left):
Here is a photo of the internals of Extreme that I grabbed from Taiko Audio's website that many have seen. No commentary is necessary. Quite simply, it is a work of industrial art:
Here is a photo of the internals of the SGM 2015. I will be kind and not show photos of the internals of servers from the Extreme's competition as they are really no better than this one:
Here is a photo of my Extreme alongside my latest server build which is housed in a comparatively diminutive Streacom FC-9 Alpha chassis:
As I have been doing with certain cherished components recently, I asked Emile if he could sign the chassis of my Extreme. Apparently, being the rock star that he is, he has been asked this before and rather than just signing my chassis, he had his signature laser engraved onto it and so as you can see, I have the SGM Extreme - Signature Edition 
:
:
Many are aware of the numerous ventilation holes drilled on top of the Extreme that function as "waveguides" which, according to Emile, function also to shield EMI emissions (from both entering and leaving) by 81dB.
What I didn't realize is that there are nearly as many holes drilled onto the bottom of the Extreme. While it is not apparent to me whether these holes also function as EMI blocking "waveguides," this is some serious labor that has gone into just the Extreme's chassis:
In the above photo, the integrated footers that you see are a hybrid design that sandwiches a block of German Panzerholz wood between a metal foot and the Extreme's chassis. The type of Panzerholz wood that Emile uses is the same type used in the soundboards of pianos and are favored for their ability to translate high frequency energy into a rich resonant sound. You can see this Panzerholz wood in the following photo:
Panzerholz is further used quite strategically within the Extreme. Also note the extensive use of expensive copper which is known not only for its superior heat dissipating properties but also its more musically resonant qualities:
Copper is used even in the standoffs. Further note how cleanly the DC cabling has been soldered onto the PCB. Not just the design but the quality of workmanship is incredibly meticulous.
Flanking the dual row of Mundorf capacitors are a pair of Dueland capacitors ($1,000 each). Emile has stated that the Extreme represents his maximum effort which is to say that the parts used in the Extreme were carefully vetted through a combination of measurements and extensive listening tests and that no expense was spared in using the best available parts.
One of the advantages of being so heavily entrenched in DIY is that over time, you get a real sense of the SQ contribution of individual components. Over the years, I have tested various SSDs (SATA II vs SATA III, SLC vs MLC), storage transmission buses (SATA vs PCIe) and even SATA cables and have found that they each have a particular sonic signature that is audible. The fastest SSDs have the lowest latencies but also generate a high frequency noise that I find fatiguing with long term listening and cannot be completely ameliorated with SATA filters or by cleanly powering the SSD. Just because you power an SSD with a high quality linear power supply doesn't prevent that SSD from generating high frequency noise. This is apparently not noticeable by all given how commonly SSDs are used in music servers today but I have purposely avoided using an SSD as an OS drive (which is incessantly being accessed by the OS even during music playback) in any of my builds since late 2018. I have even gone so far as to use a compact flash as an OS drive but as these are very high latency storage devices, they create as many problems as they fix and proved to be a less than ideal solution. In late 2018, through the reporting of others, I began using Optane drives for the operating system and these drives have proven to be the ideal solution. Optane drives have very low "RAM like" latencies but also have a noise spectrum that is more like RAM than an SSD which is to say they do not cause the same fatiguing HF noise. It was quite refreshing to find that the Extreme uses an Optane drive for its OS.
What is another dream come true to this DYIer is the Extreme's 6 PCIe slots. One is occupied by the Optane card and one is occupied by 8TB worth of storage for my music files. One slot is also occupied by the Startech SFP card which leaves 3 free slots. This means excellent future expansion capabilities without the need to use riser cables.
Lastly, there is the OS. For years, there has been a contentious battle regarding which OS is best for audio playback: Windows vs MacOS vs Linux? MacOS is perhaps the most user friendly, Windows has the broadest compatibility, while proponents of Linux suggest that Windows and MacOS are too bloated whereas with Linux, you start with a minimalistic OS and add only what you need. Windows users have countered Linux through various tools like Audiophile Optimizer and Fidelizer Pro which are software designed to shut down unnecessary Windows processes. In its most extreme form, this meant using a specialized version of Windows designed for servers without a graphical interface that is not for the faint of heart. Having gone deep into this type of Windows, I got very good results but ultimately, I found my best SQ with a version of Linux called Euphony combined with their proprietary software player called Stylus and my last few builds have utilized Stylus over Roon. While I dearly missed the library management capabilities of Roon, I could never get it to sound as good as Stylus and so I have been using Stylus almost exclusively with my builds for the past 6 months.
What is unique with Emile's version of Windows 10 is that he does not use Audiophile Optimizer or Fidelizer Pro. Instead of working hard to shut down unnecessary Windows processes, he instead modified the installation media to install only the subsets of Windows he wanted which is really the more elegant solution. While his version of Windows 10 is not as stripped down as it could be and includes Windows' graphical user interface, this is where having 20-cores over 2 CPUs come into play. Having the luxury of this much processing power and with the aid of Process Lasso and JPLAY allowed him to render the SQ penalties of both Windows and Roon moot. If you assess what's going on through Process Lasso, what you'll see during playback is that CPU load and queue depths spread across so many cores and so much RAM are basically nil. Rather than software waiting for the Extreme, more often than not, the Extreme is waiting for the software leading to a level of SQ that I will describe in a separate post. Here is a screen shot of the Extreme's graphical interface.
This screen is probably familiar to all Windows users and is a joy to navigate compared to a command line. Without having to ever attach the Extreme to a monitor and keyboard, the Extreme has been designed so the user can easily remote into their Extreme via VNC using a PC, MAC, tablet, or smartphone. Of course, the Extreme was designed to be plug and play and so this is only necessary in order to make changes to the OS such as configuring network settings, downloading drivers, or launching or stopping a piece of software like HQPlayer. When necessary, TeamViewer has also been installed which allows Emile to remote into your server. For those who feel uncomfortable with tasks like updating drivers or installing a new piece of software like Spotify (as an example), Emile has said that he is willing to provide this service. Ultimately, that is the strength of Windows is that it is broadly compatible with much more software than Linux and even MacOS. It is also why Emile believes Windows inherently has the ability to sound better than Linux. With Linux, you are forced to use generic USB, network, and sound drivers whereas with Windows, you have many options leading to a greater level of control. Having tried various ASIO and Wasapi output driver options in Roon with my Extreme including from my DAC's manufacturer, I did find JPLAY to sound best of all although this does not happen by chance as Emile specifically optimizes JPLAY to run with your DAC once you receive it. Emile is even developing his own ASIO driver which makes sense as this would give him much more complete control of the playback process.
Listening tests:
Those who have read my post about my visit to Emile's workshop in Hengelo in November know that I was quite impressed by the Extreme when I first heard it, impressed enough that I went ahead with my order for my own Extreme. Of course, not having had the opportunity to directly A/B the Extreme against my own best effort, I had some doubts. Even some of my DIY friends wondered what I would do if the Extreme was only marginally better than my own server because how could "marginally better" ever be worth Extreme-level of dollars?
For those not familiar with my latest build, it is comprised of highly curated parts based on things I had learned from prior builds and centered around a 12-core/24-thread AMD Ryzen 3900X CPU, 8GB of Apacer ECC DDR4 memory, Optane memory for OS storage, and a JCAT Femto Network and JCAT Femto USB PCIe card with each card powered externally by my Paul Hynes SR7. Using what I considered to be an excellent 800w DC-ATX converter powered by an SR7 from a different Paul Hynes supply along with a specialized DC harness comprised of high purity OCC copper, this server had the advantage of being specifically tuned to my tastes with the aid of Euphony OS and Stylus and has been easily superior to anything I have previously had in my home for comparison including Aurender's latest W20SE. As a high power server, I would happily pit it against any low power server (that means just about all other commercial music servers out there today) and would feel confident that I would prefer it. Unlike some who purchased an Extreme because they were unhappy with the SQ from their Mac Mini or NUC, I was quite pleased with this server and often spent many hours into the night listening because of an already high level of engagement. My wife, who also enjoyed our listening sessions wondered aloud why I felt I needed an Extreme as she knew I had sold every other server I had previously purchased after I was able to build something better on my own. My only response was that I felt compelled "to be done." I had been building servers long enough to know that I could forever chase the next big CPU, motherboard, etc, that people were talking about and that this would lead to incremental improvements but despite my best efforts, at least on paper, I knew I could never match Emile's efforts. That's what I kept telling myself.
Because I had asked Emile for a few minor modifications and because of the Christmas holidays, it took a while longer to receive my Extreme and so by the time my Extreme arrived to me 2 weeks ago, it had already accumulated 600 hours of burn-in time according to Emile. For the first 7 days that I had my Extreme, I kept it powered almost 24/7 and ran it in stock configuration using the stock footers atop my Synergistic Research Tranquility Base, an Intona Ultimate USB cable, and HFC Pro power cord and so this 7-day period gave me a pretty solid idea of how a stock Extreme sounds using my reference ancillary gear that I was well acquainted with. Using the same power cord and USB cable on my server along with a quad of CS2 footers from Critical Mass Systems and using a different Tranquility Base as a foundation, I was able to fairly easily go back and forth. Both sighted and blind testing were performed but there really was no point to blind testing as the differences were so obvious, even to casual listeners who came by.
What I thought would take a few hours and possibly a few days to convincingly answer actually took only 30 seconds. The first thing I assess when I listen to a piece of equipment is not tonal quality or imaging or sound stage or dynamics but rather transient response. It is what I am most sensitive to based on the type of music I listen to the most and if it is not at least better than what I have already, then there's really no point going forward. The problem with transient response is that it is predicated on timing, speed and control but overly fast and overly damped presentations have a tendency to frequently sound thin and wiry and unnatural and so that isn't good transient response. Case in point, shortly after purchasing my Wilson Alexia 2s, I felt compelled to try every amplifier I could get my hands on and I had the good fortune of being able to audition Merrill Audio's latest Element 118 monoblocks, a $36k class D design based on gallium nitride transistors known for their incredible speed (in the gigahertz range). Indeed, these made any other amplifier I had previously heard sound slow and so I heard things with these amps that I had never previously heard. The problem with these amps is that the presentation was indeed thin and did not sound natural to my ears and performed just as badly as too slow of an amplifier.
This is where I know the Extreme's 20-cores come into play because the following are character traits of high power servers that I have consistently observed but never before to this extent. The Extreme's presentation is very very fast, almost gallium nitride fast. As Emile likes to say, it boogies. It is also very well controlled and very agile. It starts and stops and starts again like nothing else that I've heard resulting in unbelievable transient clarity. In comparison, transients on my server sound smeared. Because it is so fast and so well controlled, there is incredible separation, focus, and staging. Also, bass dynamics are startling good. With Stravinksy's "The Rite of Spring," I know exactly what will happen and when and yet, the Extreme has the capacity to frighten. But these traits are not what's most important None of these things would matter if the presentation were to sound thin and unnatural. Good transient response is as much predicated on naturalness as it is on the other qualities we normally think of and the Extreme pulls this off better than anything I have yet heard. Throw what you will at it, whether it be slow, sultry vocals or complex, large orchestral music and the Extreme maintains its effortless composure.
I was convinced that there was no way my best server effort would get embarrassed and in a way, it was not because ultimately, what sounds good sounds good but if I am to be honest, the Extreme is quite a big step better. I was given 10 days to give the Extreme a try and if I didn't like it, I knew I could return it. By now, as this marks 2 weeks of owning the Extreme, it's clear I've decided to keep it. The decision was actually quite easy and I knew it after the first 30-seconds. With respect to a music server, I am indeed done.
During week 2, I knew I needed to push the Extreme's boundaries. Could it be improved with different power cords, USB cables, footers, fuses, and JCAT cards? Could it be improved when used with a separate endpoint or when used as an endpoint? What about network switches and routers? We've heard Emile's take on some of these things and they contrast with @CKKeung's group and so obviously, we're not talking about improvements on an absolute level but rather improvements based on personal preference and so I'm happy to provide my own observations. To answer these questions will require a separate post and so more to follow...
Network
It has actually now been nearly 4 weeks since I have had my Extreme but due to a busy work schedule, it has been difficult to find time to post until now. My observations on the impact of the network with regards to the Extreme remains a work in progress as Emile has my dual SOtM switches in his possession. Marcin has also agreed to send me the M12 Gold for evaluation. Based on @hols glowing report on the Melco S100, I now feel compelled to also evaluate this switch and so consider what I say to be very preliminary. As always, YMMV.
I agree with Emile in that the impact of the network seems less with the Extreme than with other servers but I also agree with the findings of @CKKeung , @hols, and others that the impact of the network is still quite significant and worth pursuing. To offer perspective, I am finding the impact of my network tweaks thus far to roughly equal the improvement I am getting from a set of CS2 1.5 footers. In other words, ignore the network at your own peril.
Router
I have explored routers a fair amount and while I have not yet tried the Ubiquiti EdgeRouters that @nenon reported on, thus far, I have found greater impact by replacing not just the router but my entire modem/router combo. After trying a few that were compatible with my ISP provider (Comcast/XFinity), I settled initially on the Netgear N300 back in late 2017. It has dual gigabit ports, used a low power (and presumably low noise) CPU, and could be powered with a 12V/1.5A external PSU. I sent it to SOtM and had its main clock and Ethernet clock replaced and you can see the 25MHz and 54MHz clock taps on the back of this router/modem in the following photo :
Powered by my Paul Hynes SR7, grounded by a Setchi grounding block (via USB), and with the RG-6 internet feed transformer-coupled, I found the improvement to be modest (not large) and believed its performance was likely being compromised by the nearly 50 feet of Ethernet cabling that separated the modem/router from my server. I tried wifi and with an LPS-1.2 powering a wifi receiver, this resulted in an improvement confirming my suspicions about the harm that the 50 feet of cheap Ethernet cabling was causing. I found a bigger improvement by bringing my router to much closer proximity of my server (within 3 feet). To accomplish this, I had to run RG6 cable from my equipment closet to where my server is. It was not hard to do and the improvement was worthwhile but the overall improvement, even with this modem/router connected to my Mutec REF 10 master clock, was still not dramatic.
Recently, I upgraded my internet bandwidth from 200Mb/s to a full 1000Mb/s which forced me to upgrade to a different modem/router since this router maxed out at 300Mb/s of bandwidth. I purchased the following ARRIS SURFboard SBG8300 modem/router which is capable of a full gigabit of Internet bandwidth but it was with reluctance since my old modem/router had its clocks replaced and this one didn't:
My fears were unfounded because even without any clocks replaced, the SQ improvement with this modem/router is quite a bit better than the Netgear. Improved dynamics and a much bigger sound stage are what really stand out. Powered once again by my SR7, the improvement here is not so subtle.
Network Switch
With this ARRIS modem/router in close proximity to the Extreme and with this modem/router powered so cleanly, does an intermediary switch still matter? The only intermediary switch I have on hand at the present time worth commenting on is the EtherREGEN from Uptone Audio.
With the ER mounted on top of a G5 titanium disk and with a Setchi panzerholz grounding box sitting on top to further mitigate vibration, like Emile found, I did not find the ER to improve that much with vibration control. I tried connecting this Setchi to the ER via the ER's ground lug and did notice minor improvement with a copper connection to the Extreme but no discernible improvement when fiber is used.
Regarding power, I agree with what others have stated. The cheap SMPS that comes with the ER is not the best way to power this switch and it may not be a noise issue but rather a voltage sag issue. With the SMPS, dynamics takes a hit. Power it with an LPS-1.2 and there is only minimal (if any) improvement and I believe I know why. Power this switch with a 12V DR rail from an SR7 and this switch improves quite dramatically. Dynamics are much improved with greater separation but with a caveat which I will explain later.
As for why the LPS-1.2 doesn't perform well with this switch (and most switches), my guess is it is because this switch draws too much current (0.8A @ 12V) and while this is within the capability of the LPS-1.2, it is not within its optimum power range. It's like powering low efficiency speakers with your amplifier cranked up to max (or close to max). There's simply no headroom and the LPS-1.2 is probably struggling to maintain voltage (again, this is my guess but that's how it sounds to me). I have experienced this before with the LPS-1.2 With components that draw less than 500mA, the LPS-1.2 works VERY well but more than this and especially as you approach 1A, you are almost better off with any other PSU (imo). If you are an Uptone fan (as I am), you would be better off to power this switch with their JS-2. Another excellent option is the new SR4 Turbo from Paul Hynes. At 400 GBP, according to Paul, it has roughly 60% of the performance of a DR SR7 and has a continuous 2A rating and so it is well suited to drive a switch. Just like the LPS-1.2, however, if your component draws anywhere close to the continuous 2A rating of the SR4, the SR4 isn't going to perform very well.
Now for the caveat. When I first got the ER, like everyone else, my router was feeding the ER via the "A" side (1000Mbit) and the ER was feeding the Extreme via the much vaunted "B" side (100Mbit). To be honest, even with the ER powered by my SR7, I was a bit underwhelmed. Midrange was rendered beautifully and there was better smoothness but transients were soft, the attack was quite dull, and the sound stage felt considerably smaller. Ok for vocals recorded in a studio but not ok for large orchestral music. To be bluntly honest, I was hearing more good things coming from a Sablon Ethernet cable than I was from the ER and I was preferring going direct from my modem/router to the Extreme via the Sablon cable.
When Emile told me his ER was sounding better than my dual SOtM switches that he had in his possession, I was quite surprised but then he also told me he didn't care for the "B" side of the ER at all and much preferred using ONLY the "A" side. He was further connecting to the Extreme with fiber using these new SFP transceivers that he found to be much better than the Startech transceivers. I went ahead and switched to the "A" side and agree fully with Emile on this. At least with the Extreme, the "A" side, even when using purely copper (and not fiber) sounds better. It is a touch less warm and more coarse than the "B" side but transients are more fully and dynamically expressed. The problem that remained is the sound stage still sounded more confined compared against directly connecting the modem/router to the Extreme.
Before fully giving up on the ER, I decided to give these new SFP transceivers that Emile liked a try and I believe I have found them. This is the link that Emile provided and neither of us were sure these were the actual transceivers he likes because these are quite a bit less expensive at only $30 each. (mgb-tlx-mini-gbic-lx-module)-
I was skeptical that these inexpensive transceivers could rescue the ER but somehow, they do. They are indeed more dynamic. They also sound stage larger than the Startechs and yet, they don't sound mechanical which is how Emile described the Finisar 10G SFPs that I had also considered trying. With these transceivers, I now am preferring having the ER in the chain as opposed to going directly from modem/router to Extreme.
As has been reported by others, even with the ER connected to the Extreme via fiber, the quality of the copper Ethernet cable that connects the modem/router to the ER still makes a difference. An inexpensive Blue Jeans Cables CAT6A cable sounds fine in isolation but in comparison to either a SOtM dCBL-CAT7 cable or the Sablon Ethernet cable, the BJC's midrange sounds recessed and treble sounds smeared. A Ghent Belden CAT6 with JSSG360 shield fares only slightly better than the BJC. A WireWorld Starlight CAT8 is the worst of the bunch as this cable sounds fatiguingly bright. The SOtM CAT7 sounds very liquid and smooth and casts a spacious sound stage while the Sablon Ethernet renders beautiful textures and is the most resolving of all the cables I have here. Among these cables, I find the Sablon to sound the most natural but also the most resolving.
JCAT cards
As previously stated, one of the luxuries that the Extreme provides is that it has several free PCIe slots allowing for future expansion as necessary. This also allows for easy A/B testing.
According to Emile, slots 1, 2, 6, and 7 are tied to CPU 1 while slots 3, 4, and 5 are tied to CPU 2. In his listening tests, he found that the Optane card is best suited for slot 3 (CPU 2) from a SQ perspective and so I did not attempt to move that card. He also found that the SFP network card sounded best when tied to CPU 1. I tried moving this SFP card from slot 2 (CPU 1) to slot 4 (CPU 2) and I have to agree, when tied to CPU 1, the SFP card sounds smoother and less coarse and so credit once again to Emile for leaving no stone left unturned. When I compared network cards, I kept them tied to CPU 1. Same thing with USB cards.
As previously reported, the stock dual RJ45 Ethernet ports on the Extreme sound very good. In the absence of any comparisons, I could happily live with the SQ from these ports and so I don't blame anyone who doesn't wish to go beyond but what I will say is that the Extreme is capable of much more. With the EtherREGEN connected to the SFP network card using the inexpensive SFP transceivers versus the EtherRegen connected to either of these RJ45 Ethernet ports even with the Sablon Ethernet cable, I am finding the fiber connection to be superior and easily worth the small asking price of the ER + cheap SFP transceivers.
JCAT Net Card FEMTO
How about the JCAT Net Card Femto? The answer is more complicated. Compared against the stock Ethernet ports of the Extreme, with the JCAT card powered by an SR7, no question, the JCAT card is better and the improvement is not so subtle to my ears. It is more dynamic, more expansive sounding, and timber is smoother and richer sounding. To all that I have subjected to A/B testing, with no exceptions, everyone has preferred the presentation of the JCAT card.
With JCAT vs fiber, this is where things get tricky. The ER along with these cheap SFP transceivers feeding the SFP card is sounding very satisfying at this time. Against the JCAT card being fed by either the ER or directly by my modem/router, the SFP connection has the lower noise floor and better resolution while the JCAT card sounds airier, has more tonal body, and has the bigger sound stage, especially when fed directly by the modem/router.
With Mahler's 1st Symphony, with the JCAT card, I am really liking Thierry Fischer's DXD version:
This is a modern recording that I just love. It is clean, dynamic, and with a very accurate sense of scale and was originally recorded using a DXD recorder and so this DXD file is essentially an unprocessed version of the master. These types of recordings, imo, represent the future.
With fiber, I am preferring Bruno Walter's version more:
Ignore the smaller size of the above thumbnail photo as that has no relevance. Bruno is the naturally more expressive conductor but casting that aside for the moment, the Bruno Walter version (1961) is a tape transfer to DXD by HDTT that has a very dense tonal quality to it that sounds a touch slow and strident on the JCAT card but nicely adds body to the fiber card. All things considered, this presentation on the fiber card is "to die for." You could see where this leads. Certain recordings may sound better one way versus the other and so there is no clear winner here.
Adding to the complexity, because I have no idea how things will sound with the SOtM sNH-10G, M12 Gold, or S100, all of this will all have to be assessed again at a later time.
JCAT USB Card FEMTO
The stock USB port on the Extreme is probably the best sounding stock USB port I have heard. According to Emile, this port is based off of the ASM3142 chipset that he really likes. In any of my previous builds, especially my last build utilizing a certain ASUS mini-ITX board, the stock USB ports sounded atrocious making something like the JCAT USB card a necessity. With the Extreme, I could once again see why no one would care to try and improve upon this. Here, the JCAT USB card provides an alternative voicing that may not be to everyone's taste but it is to mine. The stock USB port on the Extreme sounds more incisive where the JCAT card sounds fuller and is harmonically richer. The LPS-1.2 does a better job powering this card than the JCAT NET Card FEMTO and the standard SR4 by Paul Hynes is a step better but it is with a DR SR7 rail where the benefits of the JCAT USB card are most fully realized.
USB cables
I'll cut to the chase here. Not surprisingly, USB cables make a big difference. Having been part of many USB cable shootouts, I found the Intona Ultimate USB cable to be the highest resolution cable I had yet heard with no apparent harshness. Some found this cable to sound a touch sterile or mechanical but not me. I really like this cable.
Due to @spiritofmusic's prodding, I reached out to Mark at Sablon and asked to demo his new USB cable and I was quite surprised to find that his new cable somehow matched the resolution of the Intona Ultimate (not a small feat) while providing texture and air that the Intona lacked (although I hadn't realized it was lacking until I heard the Sablon). At a fraction of the price of the Intona Ultimate, this new Sablon is a NO BRAINER.