MUSINGS: Myths, clichés and Hi-Fi+'s Taiko Audio SGM Extreme computer review. A hypothetical fast, fanless audio computer build. And "channels/bit-depth/samplerate" labeling convention - a suggestion.

"Nowadays people know the price of everything and the value of nothing."

-- Oscar Wilde, The Picture of Dorian Gray, 1890

Once awhile, we in the audio world run into devices that are so ludicrous that it becomes somewhat entertaining reading or watching reviews about them. Like Super Bowl ads, the advertising can be fascinating in themselves regardless of whether you care about or even know how to play American football. The Dutch Taiko Audio SGM Extreme is such a product which I find fascinating - not because the technology is all that mind-blowing (server CPUs, Optane storage, large SSDs, big fanless case) - but in seeing what they're trying to do with the "spin" around the product to try to justify an asking price! (Previously, I've referenced the device in a related discussion as well.)

I thought it was interesting reading this recent Hi-Fi+ review by Alan Sircom and examining the claims. Just because a review might be written in a "subjective" fashion doesn't mean readers can't or should not be critical about the beliefs expressed. To a certain extent, I believe it is the responsibility of "rational audiophiles" to push back against the nonsensical, even mythical, ramblings of certain subjective, supposedly honest, viewpoints. I agree with T.S. Gnu, typically we cannot expect magazine writers to be bluntly honest in today's environment.

For those unaware, The Absolute Sound (North America) and Hi-Fi+ (UK) are glossy, subjective-only, sister publications that IMO read like the unabashed advertising arm of the "high end" industry. Over the years, TAS & Hi-Fi+ have advocated snake-oil of all sorts and TAS even published pseudo-science articles purporting to be empirical research (look up the articles from Charles Zeilig and Jay Clawson around 2011 - I've mentioned this back in the day, and here's an example of what they do/say).

While writers can choose different directions on how they want to express their views when it comes to the Taiko audio computer, the myth/"hook" in this Sircom review centers around this belief:

While streamed audio is extremely good, there’s always a nagging sense of something a little more that can be extracted from a spinning disc.

The "nagging sense" that streamed computer audio (which many of us have been doing for probably 2 decades already) cannot routinely match a physical "spinning disc" CD player is a fascinating meme which has been floated around for years now among certain audiophiles. From there, Sircom then praises this computer to high heavens; apparently unlike any other computer he has used, finally "The SGM Extreme changes the game", he says.

Is there any evidence that a spinning CD player, extracting 16/44.1 digital content, is better than a computer which has the ability to easily stream high-resolution digital (PCM and DSD) to modern DACs? Beyond the mere anecdotes of certain folks like this, as far as I can tell, there is simply no such evidence available, nor is it even plausible these days with high speed computer storage, increasingly efficient yet fast processors, and asynchronous interfaces like USB.

A good primer on Abstract Art...

As you can see, Sircom seems to have difficulty describing what it is he's ostensibly hearing from the Taiko computer that's such a game changer (emphasis mine):

It’s a bit of an abstract property but the SGM Extreme just sounds ‘right’ whatever you play. The order and structure of the music are as you expect they should be, rather than ‘almost’ how they should be. This is something more than accurate tonality, wide dynamic range, precision, vocal articulation, detail, rhythm or any of those regular characteristics. In part because you can get pretty much all of those things to a very high degree from the most basic level streaming. It’s just that the SGM Extreme simply raises the bar on every aspect of music by several notches.

So we have a device that "raises the bar on every aspect", and "by several notches" even, but these are "abstract" qualities, and apparently beyond the "very high degree" already achievable elsewhere. I have no qualms with the concept of something being abstract, but there are 2 issues when applied to this computer.

First, I find the use of the word "abstract" as in "existing in thought or as an idea but not having a physical or concrete existence" inappropriate when talking about difference in sound quality. It's a machine that feeds data to a DAC from its storage. Either physically/concretely it actually impacts the sound, or it doesn't! It should not be about an abstract "idea" that this computer "sounds better", but in truth whether a difference actually exists with the sound reproduction and whether the difference is judged to be better.

Secondly, there is certainly beauty in the impressionistic uncertainties of art in which the abstract invites the perceiver to project feelings and thoughts into the artwork. But this computer is not "art", so the belief that this box imparts some "abstract" quality does not make it any more valuable. In fact, in 5 years when those Xeon processors, SSD storage, and Windows 10 become old news, good luck with finding significant value in this box. Unlike real art which can be a store of value over time, you probably will have trouble even giving this computer away in 10 years as technology marches forward.

I think the word Mr. Sircom is actually searching for is not "abstract", but simply "vague". The idea/belief that this box making a positive difference is more likely than not just his own mental projections. I can see how he doesn't use the word "vague" though because no honest person would also claim the difference "raises the bar... several notches" in the same article. An example I think of how reviewers play with ideas, introducing positive biases in the service of the advertising when in reality they have little to say.

So after all these decades, finally (!!!) it's claimed, in the 2020's we have a very expensive computer with USB output that can surpass the best spinning disc player since the birth of the CD. Can any self-respecting, knowledgeable audiophile accept this claim? As you can see, there is no deeper discussion to be had in the article about why the device is better (not even claims of better jitter!). All we have are talking points about stuff like "carefully selected 4GB DIMM memory modules", "linear power supply with 700,000µF of reservoir capacitance", "OS stored on PCIe Intel Optane non-volatile storage", "panzerholz wood" in the CNC milled chassis, "custom USB board" and "custom drivers" running in Windows 10 Enterprise. Just because a machine uses expensive parts, or the company customized a Windows USB driver (who knows with how many lines of code altered), in what way does any of this result in better sound quality?

In order to find some vague rationale for this stuff, we have to go to the Taiko website on the SGM Extreme to be treated with technical gobbledygook with claims like the dual-CPU somehow nullifying the effect of the Roon GUI interface on sound, SSD sounds worse than PCIe, low-current-draw RAM modules making a difference ("big impact" they say), stuff on OS minimization, stuff on vibration control, and the metal holes on the chassis acting as "waveguides"... Needless to say, none of these factors are discussed with beneficial effects demonstrated in any way, and the CEO/founder's blog just shows pretty pictures without any depth of knowledge nor words of wisdom (I suppose they like leaving it to others with bold imaginations)!

Guys like Sircom seem to love dragging out fancy car analogies trying to dissuade readers from thinking that this is "just" a computer, but rather, more of a "Rolls-Royce Phantom VIII". The reality is of course much more pedestrian. Does a gold plated toilet handle flush any better than a nice firm brass lever?

With this TAS/Hi-Fi+ crew, one never expects any objective results to show the "several notches" improvement (never mind even showing a single notch), and their opinions are typically predictable based on price tag and physical characteristics (appearance, weight, etc.). You, dear audiophile readers, just have to take the claims on faith.

A review like this, IMO, is effectively a very poor infomercial. For me, it's hard not to cringe at the plethora of clichés used in the article: the concept of "un-digital" sound (I guess compared to LP playback), the idea that this device is somehow better able to make the sound more "live", digital audio sounding "grey" (whatever that means), that this device "resets" expectations, providing a "musical revelation". All this simply from a Windows 10 based, warm-running, expensive, fanless computer sending USB data to the DAC! Is this not an embarrassment of a review for a technological product? Honestly, how does Mr. Sircom and a magazine like this expect to find any respect among technologically astute hobbyists and the generations well-versed with modern electronics, Internet hype, and fake news?

[Seriously folks, what else can you expect from TAS and Hi-Fi+ when at the top, the CEO of the publishing company, Lee Scoggins, is clearly a name-dropping Industry salesman who apparently loves Synergistic Research nonsense. He even believes that AudioQuest Diamond Ethernet cables (not even HDMI!) improves the picture quality of his Samsung TV playback (1:42:29). "I don't know why that is, but it works!", he says. He comes across as a guy who would say and do anything to make a sale. A man who has been banned at places like AudiophileStyle for basically being a shill for MQA for example. Audiophiles, this is the level of intellect and maturity managing these publications.]

As entertaining as magazine articles are at times, it's potentially just as entertaining seeing and listening to the personalities selling these products. For example, Emile Bok, the smiley founder of Taiko Audio is fascinating to watch as is the uncritical, overly effusive interviewer (here's Part II). Beautiful example, I think, of a man selling modern snake oil and his shill. Remember the audiophile psychology that permeates through all this, and how the Industry tries to benefit from the market created of fetishisms, fantasies and apparently milking quite a bit of disposable income out there.

Unfortunately, these days I don't have much time to participate on audio forums. It was fun recently jumping into the AudiophileStyle "Subjectivist Despair" thread (starting on Page 2) chatting with the denizens. In the thread, there was a discussion of setting up a blind test with the Taiko SGM Extreme versus a Raspberry Pi streamer (maybe something like this). I think this would be a nice experiment although highly unlikely to ever happen for all kinds of reasons. I don't think most rational audiophiles would care to purchase such an expensive computer for a bit of testing (imagine the shipping these days and who knows about restocking fee for Taiko's 99lb monster on a return), plus extreme subjectivists don't want to believe in blind testing results anyways unless I suppose it supports their faith; what's the point other than likely proving the obvious?

I know some audiophiles claim that one needs to "try" and listen before passing judgment (and often claiming the need for extended listening due to "break in"). In reality, this is simply not a wise way to live life unless you have a dealer nearby to pop in for a visit for something like this. Imagine if one needed surgery and you came across an article claiming that "psychic surgery" produces amazing results (at a price of course). Must a rational individual entertain such a belief or even try it in order to "know" it's not a good idea before standing firm against such techniques? So it is the same that there is no need to try a computer like this if we know already how these devices work and why. Let's leave it to our would-be psychic surgeon equivalent, Emile Bok, to show us that this thing is worthwhile first. For the record, IMO, generally "Bits Are Bits" and there would be no difference in the audible sound quality for bit-identical playback with the Taiko computer or the Raspberry Pi to the same high quality, modern USB DAC (consistent with previous testing).

Personally, I think it would be more interesting comparing the multichannel HDMI output of the inexpensive MeLE Quieter2Q also running Windows 10/11 versus the Taiko SGM Extreme. Oops... Hang on... It looks like the Taiko is missing HDMI and only has VGA output for video - rather limited functionality for a computer these days at this price point!

I would recommend staying clear of the unsubstantiated, the unlikely, the "abstract", and the mystical audio claims. At £25,000 (US$32,500) for what Sircom reviewed, the value proposition here is indeed "Extreme", as in supremely poor value for the money that could otherwise be used in countless ways, or simply saved wisely.

[You might also be interested in this older related article on the "computer audio mythos".]

--------------------

For fun, let's price out a fast, fanless audio computer / server in 2022 which I think would work very nicely.

While I don't believe it's a good idea to have a high-power computer in the sound room and much prefer a "Type 2 Distributed Server-Streamer model" of computer audio, I wondered how much it would cost to put together a fanless fast machine using current-generation CPU and components.

I think this would work beautifully (approximate street price on Amazon as of April 21, 2022):

Intel i7-11700 CPU 8-cores, 16 threads with UHD Graphics 750 (US$325)
   - fast with CPUMark of ~20700
   - 2.5-4.4GHz, only 65W TDP 

ASUS Prime Z590M-Plus microATX motherboard (US$170)
   - plenty of USB ports including USB 3.2 Gen 2x2 Type C
   - M.2 for higher speed storage, HDMI output for multichannel audio

32GB DDR4 3600 DRAM (2x16GB Corsair Vengence LPX - US$150)
   - 32GB is overkill for any audio streamer/server, but inexpensive

WD_Black SN750 SE 1TB M.2 (US$100) storage for OS
   - 1TB for Windows 11 install and Roon would be plenty
   - Sure, get 1TB Optane-accelerated if you want; I don't see the point

Crucial MX500 4TB 3D NAND SSD x 2 for 8TB music storage (2 x US$400)
   - Crucial SSDs are based on Micron chips, should be reliable for years
   - plenty fast SSD, low latency, low power and 8TB is a great start

HDPLEX H5 passive cooling case + 400W HiFi DC-ATX+HDPLEX 300W linear PSU
   - US$1100; nice looking 18lb box, supports up to 95W TDP CPU, linear PSU

Conservatively, we can budget a total price of less than US$3,500 with taxes and shipping for a computer like this. Speed-wise, this i7 CPU would be about 2/3 that of the Taiko's dual-Xeon Silver 4210 but requires <50% the power (lower electrical noise). The i7 would have significantly faster single-threaded performance (>60% faster) with its newer 11th Gen Core architecture so it will feel quicker and perform overall better than those Xeons unless you're hammering the system with tasks like video encoding or rendering that can be efficiently multi-core optimized (not the case in audio, I presume Mr. Bok knows this). The DIY build should also run significantly cooler.

Gigabit motherboard ethernet is plenty fast enough for audio purposes and some audiophile companies are still peddling slow 100Mbps stuff (like this recent £1,100 Network Acoustics muon "filter"!). Feel free to upgrade with a 10GbE network card (~US$100) if you have the need for network speed; "sounds" awesome with no noise issues here at home (these days my network also has mixed 2.5GbE speeds). I see the Taiko has an SFP port for fiber networks - no biggie, just get a PCIe card if you need that. As SSD prices drop, try 8TB drives like the Samsung 879 QVO (~US$750 currently so not bad).

Going on the Taiko Audio site, if I want to order one of their SGM Extreme base units with 8TB storage, without any special USB card (I fail to see the need, and again no evidence it will make any difference), this is how much it costs:

They're asking €24,660 = US$26,700 before taxes and shipping. Wadda steal!

I'll happily take my chances on the $3,500 computer as way more reasonable with useful features like HDMI 4K video (great as HTPC) and multichannel/Atmos audio output, modern 20Gbps USB-C, plus I think it's better to start with Windows 11 these days rather than Windows 10 Enterprise (on a relatively small 280GB Optane drive, probably one of these discontinued 900P Series PCIe cards).

BTW, I see that all they're doing for the music library storage is using the ASUS Hyper M.2 X16 PCIe x 4 NVMe M.2 V2 card (less than US$60, notice small fan). Filling that with some M.2 drives like maybe a couple of WD_BLACK 4TB SN750 (~US$700 each) for an 8TB build. So, apparently doing this transforms the sound (negatively IMO) to become "only matched by the very best vinyl and tape playback systems" - sure, Emile...

One more thing about the SGM Extreme, it looks like it's based on an ASUS WS C621E Sage motherboard (US$650) released in late 2017. Apparently, recent BIOS updates turned off the Intel Platform Trust Technology on the board. Might need to grab a dedicated TPM 2.0 module to install Windows 11 if needed one day with the latest BIOS; bit of a pain.

I guess it would be fun building something like the HDPLEX machine with the kids, and enjoying the satisfaction of getting it working. If anyone has built something like this, let me know in the comments! With the $20k saved, one could upgrade the DAC / amp / speakers / headphones / room treatments - all of which are way more likely than any computer (without DSP processing) to make a sonic difference.

Or put the money into the down payment for that Rolls-Royce. ;-)

--------------------

To end, I just wanted to discuss a minor quibble; a little suggestion around labeling conventions for albums of different channels, bit-depth, and sample rates.

With the flexibility of computer audio, I bet like many of you, the music library consists of all kinds of files including lossy, standard CD-resolution lossless, hi-res downloads, maybe some DSD, and even multichannel albums. Often, I will also have more than one copy of the same album in different resolutions. Here are the 3 main parameters for defining the nature of digital albums:

Channel Layout: Stereo (2.0) vs. Multichannel (3.0, 4.0, 5.1, 7.1...)

Bit-Depth: 16 vs. 24 vs. 32-bits (vs. 1-bit DSD)

Sample Rate: 44.1 / 48 / 88.2 / 96kHz... (vs. DSD64/128/256...)

Of the 3 parameters above, how would you prioritize them when it comes to potential sound quality?

A number of years ago in 2017, I spoke about how I labelled/tagged my music library. To achieve consistency is IMO the most important housekeeping task so that I can effectively search out and categorize a large library of thousands of albums. Depending on how one prioritizes the 3 parameters above, this will determine the convention we might want to label an album whether it's in the filename or how it's tagged.

The other day I added a Blu-Ray rip of The Doors' The Singles (2017, DR11 stereo, DR13 4.0/Quad) into Roon, this is what it looks like after the import:

As you can see, I found no value in keeping this rip 24-bit since the noise floor is high - dithered down to 16-bits. Discussed in the "post hi-res" article awhile back.

Notice that I always label as "channels bit-depth/sample rate" as in "4.0 16/48". This is the opposite of Roon's "48kHz 16bit Quad" ordering displayed.

And here's another with Gregorian's Pure Chants (2021, DR9 stereo, DR11 in 5.1):

I like the "Archimago convention" more than the "Roon convention" ordering for these parameters. ;-)

IMO, the most significant sound difference comes with the jump from stereo/"2.0" to multichannel layouts. If it's a multichannel mix, that's the first thing I want to know! Then the bit-depth IMO makes more difference than sample rate as a reflection of potentially lower noise when fed into hi-res DACs.

As for sample rate, even 44.1kHz already can provide better frequency response than human hearing, so I see this as the least likely to be important for audible effect. As discussed previously, if we look at album data in an audio editor, especially older analogue-sourced recordings, there are often ultrasonic anomalies present in high sample rate albums that would be better filtered out with downsampling (for example bringing 192kHz --> 48kHz). This will also in the process save storage space. I typically still target 16/48 as the sweet-spot for the majority of non-hi-res recordings especially analogue-sourced stuff; usually high-quality resampled and dithered transparently with iZotope RX.

I've been consistent with the "16/44" and "24/96" convention for years; it just always made more sense than some folks using "44/16". And whenever I have multichannel, I'll use "4.0 24/48" (or "Quad 24/48") as the convention, capturing better the essence of potential sound quality for the format. Yeah, I know this is a "nit pick" - this is what "perfectionistic" audiophiles do... ;-)

Needless to say, I would love if Roon has a setting to reorder those parameters in the user interface for those who might think they currently have it all totally backwards!

That's all for now, dear audiophiles. Until next time, I hope you're enjoying the music, staying rational, and honestly evaluating the value of some of the audio hardware up for sale out there!