Saturday 21 September 2019

Upgrade to AMD Ryzen 9 3900X Workstation Computer. (And Amazon Music HD/UHD opens to lossless and "hi-res".)

Being post-RMAF 2019 with all that talk in the last week or two with "high end" audio stuff, I figure I'll switch gears a little and take a look at computer tech instead.

One of the least enjoyable things I do every few years is to update the machines I use for work and the various ones I have here at home. While I don't enjoy the basic IT stuff and all the software installations that typically come with new computer builds, it's a good way to get updated on the machines out there, practice reasonable parts selection, and appreciate the price of the technology. Back in 2017, I updated my workstation here at home to an AMD Ryzen 7 1700 CPU which still is a very impressive processor for general use. But time marches on and my office workstation is really itching for an upgrade which hasn't happened since around 2014 and feels even slower than my laptop.

The idea then is to transition the Ryzen 7 over to the office and let's build an even more powerful workstation here at home where I do most of my writings and media encoding. Furthermore, let's try out one of the newer generation M.2 SSD drives that promise even higher transfer speed...

So, the result of a bit of online shopping:

AMD Ryzen 9 3900X 12-core CPU (MSRP US$500, current scarcity higher price)
MSI X570-A Pro motherboard  (~US$150)
EVGA Supernova 750 G3 80+ Gold 750W power supply  (~US$130)
Corsair LPX 32GB (2x16GB) 3200MHz DDR4 RAM (~US$160)
ADATA XPG SX8200 Pro 1TB M.2 Solid State Drive (~US$150)
Corsair Carbide 100R Silent mid-tower case (~US$70)

Throw in an inexpensive video card - maybe something like an nVidia GTX 1050 for workstation purposes, or if you do some gaming, one of the newer nVidia GTX RTX 2060 cards would really create quite a formidable machine.

We can see in the computing world the competition going on here between Intel and AMD! We can see the companies vying to be the best in speed, price, and overall feature set. AMD this year has really rocked the market with its latest 3000-series 7nm Zen 2 architecture of which the Ryzen 9 3900X, 12-core, 24-threads, 3.8GHz base, 4.6GHz max boost, TDP 105W sits as the top mainstream CPU (for now). While the power-user computer geeks might want their own CPU heatsink, fan, maybe water cooling solution for overclocking, I'm very happy that this CPU comes with the stock chunk of metal and fan in the box to keep the budget lower :-).

While this CPU can be used with previous Socket AM4 chipsets, the latest is the AM4 X570 which is the first mainstream solution to feature PCIe 4.0 which basically means it offers twice the speed of the previous PCIe 3.0 specification; we're looking now at about 2GB/s transfer speed per single PCIe 4.0 "lane".

Here she is, the Ryzen 9 3900X:

After doing it a few times, it's not hard to put all the components together on the motherboard, all ready to stick in the case with the power supply - done in a few hours...

Notice that the X570 motherboards commonly have small cooling fans these days. I assume this is a reflection of the chipset (PCIe 4.0) having greater power demands. By default, the fan control doesn't spin the chipset fan until the temperature goes quite high (like around 50°C). Therefore, it's silent most of the time.

For audio output, this MSI motherboard features the Realtek ALC1220 Codec. The chip apparently has a claimed 120dB SNR and one of the motherboards that featured this chipset back in 2017 got a measurements at 24/48 of ~108dB dynamic range and 0.0034% THD using RightMark in loopback. I'll make sure to measure this with my RME ADI-2 Pro FS ADC when I have time and report on the results on this CPU/MoBo combination. (Will be interesting comparing with the results of the ASRock Z77 Extreme 4's ALC898 as previously published.)

Apart from the CPU offering a speed upgrade, what should also be noticeable is the NVMe M.2 SSD which interfaces directly to the PCIe interface making it much faster than through typical SATA interfaces. Typically, a SATA III drive can give us ~600MB/s. With an NVMe M.2 SSD, we're looking at up to 3500MB/s!

Good to see that terabyte SSDs are now very reasonably priced:

ADATA XPG M.2 1TB SSD with heatsink metal sheet to stick on top over the chips.
And voilĂ ... All the bits and pieces in my full tower case on first boot-up with fancy undulating colored AMD CPU fan:

The previous Ryzen 1700 motherboard and CPU, with an older non-modular Seasonic power supply now transplanted to the new Corsair case which I'll bring into the office. As a low priced computer case, the Corsair Carbide 100R Silent is easy to work with and looks quite clean when all is said and done:

Corsair Carbide 100R Silent Edition: Notice the padded side panel for sound/vibration reduction.

So how fast is the Ryzen 9 3900X compared to my previous Ryzen 7 1700?

That's the CPU-Z benchmark comparing the Ryzen 9 3900X with the 16-core Threadripper 1950X. As you can see, this processor gets us to 95% of the speed of the 16-core CPU with ~30% increase in single-core performance. Compared to the Ryzen 1700's score of 3905 multi-threaded, the new CPU more than doubles that score, and >45% faster single threaded performance. All of this at stock speed of course.

Remember that although synthetic benchmarks like this are useful, real-life use will feel different. In fact, for most tasks like MS Office or surfing the web, one is unlikely to significantly notice these differences between the 8 and 12-core machines given how fast computers have become these days.

And as for the NVMe M.2 drive performance, there's a very substantial increase compared to the Western Digital Gold 6TB 7200rpm hard drive and Samsung 850 EVO SATA-III 500GB SSD:

As expected, massive increase in the sequential read and write speed with M.2. Significant reductions in speed across the board once you start doing lots of little 4KB random access but the M.2 being still significantly faster. Hard drive performance these days compared to solid state devices really looks unimpressive! In real life, we can certainly feel the difference in speed whether it's a faster boot time, instantaneous opening of programs like Firefox or Chrome, and much faster starting of large applications like Photoshop.

Power use and undervolting...

Unlike a few years ago wanting to get a bit more performance from the Ryzen 7 1700 by overclocking, I'm thinking that for the most part, this 12-core 3900X CPU provides more speed than I really need in day-to-day work. It's actually nice to feel that I can actually slow the thing down a little, save some energy, reduce fan noise, and know that there's a bit more power under the hood if needed in the future... And this is even without doing any kind of overclocking.

The interesting thing about the Zen 2 is that the CPU does a very good job with remaining stable even with lower voltages. For example, I can undervolt it by -0.15V in BIOS (typically the CPU uses 1.2-1.25V, quite variable) and the machine still runs stably:

I can also drop the CPU SmartFan settings since power use / heat production is lessened:

Notice the control points for fan speed % and temperature on the right. 95% fan speed by 80°C. I've also changed the fan "step up time" at 0.2s and "step down time" at 0.7s to reduce the "hunting" when the fan changes speed.
We can achieve a quieter machine with less noise variability (without as much annoying up and down frequency fan noise until the CPU is presented with sustained higher loads).

Using my trusty Kill-A-Watt meter, at stock voltage and speed, the computer sucked up 85W when idle and this dropped a little to 80W with undervolting. What was more noticeable was that when we ran the CPU-Z "Stress CPU" task, power usage with stock setting was 214W which dropped substantially to only 176W undervolted - this is a reduction of 17%.

The price we pay is of course with lower processing speed. The CPU manages the lower power "budget" that's available by throttling performance:

So, from about 95% the speed of the 16-core Threadripper 1950X at stock, we're down to 90%, about a 5% speed reduction for the multi-threaded benchmark. The drop in single-threaded performance is a little higher at ~8% reduction presumably as the CPU does not have as much leeway for fancy clock turboing while undervolted.

One of the best ways to test for memory and CPU stability is with IntelBurnTest. Here are the results using 20GB RAM comparing stock speed with the -0.15V undervolt:

There is some GFlops variability as I was using the computer at times during the tests. Basically what we see is that at stock speed we get ~146GFlops, and undervolted down to ~140GFlops, about a -4% drop. When I look at power utilization, at stock speed, IntelBurnTest uses 236W peak while with undervolting, this goes down to 210W. In this instance, we see a drop in power use of 10% without as much speed penalty.

Considering I'm not feeling speed-constrained for now, the 4-8% loss in speed is IMO a good tradeoff by gaining a quieter experience with 10-17% reduction in power with all cores in use.

BTW: The overclocked Ryzen 7 1700 scored ~76 GFlops a few years back. Remember that this version of Linpack was optimized for Intel processors and they tend to score higher than AMD machines.

In summary...
The new series of AMD Ryzen 3000 "Zen 2" 7nm CPUs is impressive at the MSRP price. As I look around, the price for the 3900X is currently a bit higher due to parts shortage.

This 12-core CPU is fast and I suspect should last me quite a number of years - I think :-). Of course this is contingent on no "killer apps" coming down the pipeline that requires even more processing power!

Where I have very noticeably seen an improvement in speed has been multi-core 4K video encoding such as using Handbrake for 4K HEVC encoding in software. Around 1.75-2x the speed of my Ryzen 7 1700. Another doubling of speed and it'll be realtime software encoding for what I do. Also, the M.2 SSD substantially increased data copying/reading speed, and program start-up speed; I'll certainly be using these for future machine builds!

Let's consider other CPUs in this performance class. The current Ryzen Threadripper 2950X (16-core, 32-threads) processor is priced ~US$650. This is about 30% more expensive than the Ryzen 9 3900X's MSRP. Plus the TR4 motherboards (currently least expensive I can find on Amazon is the Gigabyte X399 AORUS Pro) starts around US$280, also significantly more expensive than the PCIe 4.0 X570 motherboards. Performance-wise, the TR 2950X is about 10% faster than the first generation 1950X which means that at stock speed, the Ryzen 9 3900X is at most 16% slower for multithreaded performance, yet still ~15% faster single-core performance, and uses significantly less power (TR 2950X TDP of 180W vs. 3900X of 105W).

As for the Intel offerings, the closest processor in price (also around US$500) is the Intel i9-9900K 8-cores, 16-threads. Notice that Intel does not include a stock heatsink/fan with their CPU, so add a few more dollars for that. I seriously entertained getting this processor since it does also incorporate a GPU (Iris Plus Graphics 650, excellent for hardware accelerated QuickSync HEVC encoding/decoding), plus the single-core performance on Intel chips remain top notch at ~5% faster than the AMD. This i9-9900K is especially good for Intel-optimized intensive gaming. However, for workstation multi-threaded performance which will be much more important for me going forward, the 3900X ultimately outperforms by something like 40% when all the cores are utilized!

GigaFLOPS per dollar as a workstation, the Ryzen 9 3900X is excellent value if you need/want this kind of processing speed.


In other news, as you've probably heard, Amazon Music is going into "HD" (a misnomer, HD is just lossless 16/44.1, see Dr. AIX's comment) and "UHD" (>16/44.1 lossless "hi-res") streaming. This is a big deal for the audio streaming world.

The interesting thing is that compared to others, they're undercutting the monthly price to around US$15 for a standard subscription. If you're an Amazon Prime subscriber, then the cost is down further to US$13. For those who are satisfied with the status quo, their standard lossy Amazon Music Unlimited MP3 320kbps tier of US$10/month remains intact.

We've known for awhile now that internet bandwidth can rather comfortably support lossless music streaming (just look at the fact that we've had 1080P and 4K movies for years now). Half-measures like the hybrid lossy MQA codec had always been a "solution in need of a problem". Clearly, Amazon's move is going to shake things up for all streaming services including the big names like Spotify and Apple. But I believe much more so the little guys like Qobuz and Tidal.

For years, I have said that the value of "hi-res" should not be high for all kinds of reasons including the fact that the majority of "hi-res" is nothing more than old analogue-sourced material or modern dynamically compressed digital. For example, in this article back in 2015, I thought that compared to a regular CD of say $10, a good high-resolution album might find reasonable value at $13. Well, it looks like Amazon's price structure got it right. Regular "CD-quality" MP3 320kbps at $10/m and lossless plus hi-res at $13/m certainly fits with what I thought those years ago. Tidal and Qobuz subscriptions, both currently at US$20/m for lossless 16/44.1 (with or without MQA) are in jeopardy once Amazon competes in that country.

For consumers, this is a big win since price is the bottom line for services like this where there is ultimately no ownership of the music. It's even in line with the Netflix HD/UHD price of US$13-$16. So long as Amazon's streaming software/app is good enough (hopefully output bitperfect to ASIO/WASAPI devices in Windows soon), the service is stable (no reason why not as AWS is already the infrastructure for many cloud, including music services), and the sound quality reflects the advertised claims, this will be a big win for Amazon Music.

From a business perspective, I think this is a smart move. Amazon is currently the 3rd largest subscription service at ~32M subscribers (behind Apple with ~56M and Spotify with ~100M paid subscribers) but growing fast and this maneuver suggests they want to really shake things up at the top.

Despite all the potential benefits to consumers, I do appreciate the nuances of having small players like Qobuz (200,000 subscribers) and Tidal (~4M subscribers if numbers are to be believed due to previous questionable claims) around to serve audiophiles with perhaps deeper catalogues in certain genres. The other thing I think we need to be mindful of is that the artists get a fair share of the proceeds; Amazon is not known to be generous.

Only a few months ago I wondered when lossless streaming will become a "standard". It looks like the time is soon.

I hope Amazon gets the HD/UHD tier up and running for us in Canada soon. And Roon, please adapt and make sure to incorporate Amazon HD/UHD streaming into your software - this will be very important for audiophiles and obviously for Roon's success in the years ahead - I suspect especially as the Tidal users decline. I just hope Amazon is also playing ball to open up their API to allow as many 3rd party devices and apps access as possible.

I think the consumer's winning as these changes happen.

Hope you're all enjoying the music!


  1. Re Amazon HD Music:

    It's worth checking on your own devices exactly what the Amazon app plays compared to the claims. I have two Android phones capable of "Hi Res"; a Sony XZ1 which supports up to 24-bit 384 kHz and a LG V20 which supports up to 24-bit 192 kHz. In each case the Amazon app somehow determines that the device is only capable of 16/24-bit 44.1 or 48 kHz, so while the app boasts "HD" playback and Amazon's own promotion of the Android app claims "lossless HD quality" this is not in fact the case, even with phones fully capable of delivering. Instead you get a transcoded stream.

    It's also worth noting that Amazon's apps don't support gapless playback. Inserting a click or a gap into seamless audio sounds anything but "lossless" to people who enjoy live music recordings, opera, classical, numerous classic rock albums etc. It's perfectly possible, likely eve, that people might be paying for HD lossless but actually getting a lossy stream that is inferior to a straight flac/wav/wavpack rip from CD.

    1. Thanks for the note julian67,
      Looks like there's still some growing pains to be had for the Amazon streaming system. Gapless would certainly be important for live albums and many genres.

      I suppose downsampled/dithered FLAC should still be superior to MP3. Would certainly be interesting if anyone has a chance to test out the quality!

    2. "Instead you get a transcoded stream..."

      It wouldn't be transcoded, but rather downsampled and bit reduced. And, unless a recording was specifically made to take advantage of the increased dynamic range of 24 bit, you're unlikely to notice a difference. Most home listening environments aren't really set up to take advantage of it either, as the the increased DR is lost in the noise floor.

    3. In fact, it may well be transcoded. It could have originated from the record company as flac, wav, alac, pcm or even DSD, and been both transcoded, and resampled and bit depth reduced!

      Because the original file is unknown to the end user it's impossible to tell if it has been transcoded from one lossless format to another. It's equally impossible to know if it has *not* been resampled or bit depth reduced!

      As to audibility, I made no claims to that in terms of sample rate conversion, bit depth etc. as my point was to contrast what Amazon claims to offer with what the subscriber may actually get, with those being two identifiably different things.

      What *is* plainly an audible difference is the insertion of gaps into supposedly lossless streams.

      Anyway, I took up the free trial and will check for improvements with any subsequent app updates within that period, but Amazon has never seemed to care to offer gapless playback in its long existing music app and seems unlikely to start now.

    4. Absolutely, the insertion of gaps is a big one and I hope the users provide enough feedback (complain enough!) that Amazon will make this right. I think it's silly to go lossless and hi-res if they can't even make sure that gapless comes along to achieve a high quality experience.

      I assume gapless is not an issue with Tidal and Qobuz, right? It has been years since I tried out Tidal and unfortunately neither Qobuz or Amazon HD/UHD available here in Canada.

      I'm sure at some point there will be folks analyzing the quality of the stream from Amazon. Will be interesting to see which releases / masterings the music comes from and whether the stream is actually bit-perfect compared to the CD for example except for maybe ReplayGain-type volume equalization.

    5. "Sony XZ1 which supports up to 24-bit 384 kHz and a LG V20 which supports up to 24-bit 192 kHz"

      I'm curious what LG and Sony mean by "supports". The thing is--the audio DSPs used in mobile platforms operate at 48 kHz. Anything that passes through the ADSP has to be resampled to 48 kHz. And ADSP is used for pretty much every audio input and output on the phone. The only exception is the USB output *if it's accessed directly* via USB drivers as most of the "audiophile" audio players do. I don't know what Amazon player uses though.

    6. @Mikhail

      Not all Android devices use the standard Android audio framework for everything, hence the capability is not always limited to 48 kHz. It's easy enough to check what an Android device is actually doing with audio by enabling developer options and using adb (android debug bridge). In your adb shell you run the command, with audio playing: "adb shell dumpsys media.audio_flinger". Now you can identify and inspect the active stream. A normal Android phone/tablet will indeed run everything through the stock ADSP, but some devices can bypass it entirely for certain apps. This is by default the manufacturer's dedicated music app, but third party apps such as Neutron Music Player and USB Audio Player PRO can also achieve this, whether playback is handled only on the phone's hardware or when output to a USB DAC. In these cases you can ensure bit-perfect playback; no sample rate conversion or bit depth reduction.

      Even some older phones did not use stock Android audio specifications. My 2012 Samsung Galaxy Note II has a modified audio framework that uses 44.1 kHz instead of 48 kHz, though unlike modern devices it does only offer that one possibility and resamples everything else! Anyway it was great in 2012 when all my music collection was derived from CD :-)

    7. @julian67

      You are right, I didn't consider the possibility of the phone vendor adding a dedicated audio path for high resolution audio, thanks for pointing that out!

  2. If only audio components would be so meticulously and objectively reviewed and dissected like PC components, while also being aggressively priced based on real irrefutable performance benefits.

    However, as soon as "subjective" properties have entered the enthusiast hardware domain, we already witness the premium tags on fancy RGB fans and lights, water cooling systems, composite alloy "designer" heatsinks, tempered glass cases, various "gaming" accessorises, etc.

    Thankfully, while the latest cutting edge HEDT always had the premium price because it IS the (measurable) pinnacle of available performance, all of the above is for the most part clearly labelled simply as "bling" and "aesthetic", without dubious claims of performance increases.

    1. Yup, very true Turrican,
      Bling is entering the computer hardware space. I've always considered "high end" audio as the harbinger of what is to come for all the technological hobbies...

      Once we reach a place where the utilitarian concerns have been achieved, there is little need for companies to truly innovate technically. They will then seek out subjective interests of the consumers. Provide ways of "personalization" that appeals to esthetics.

  3. Gapless... I subscribed to Amazon Music "Hi-Res" the first day it was available. SQ seems fine, but I have no way to measure it. But this discussion got me to check out gapless. I am a stickler for gapless playback. My favorite test track is Evgeny Kissin playing Beethoven's 32 Piano Variations Wo0 80 on Deutsche Grammophon. (It is also my favorite exhibition of pianistic virtuosity. WOW!) It consists of 32 tracks played in 12 minutes. Many of them last only 11 seconds. They are mostly played at a furious pace and any gaps totally ruin the flow if the music.

    Anyway, streamed from Amazon at an indicated "HD" resolution (CD quality) using the BluOS plugin on my Bluesound Node 2, it is perfectly gapless. But the Bluesound does an excellent job of buffering, so not necessarily a good test. So I then tried it on my Android phone using the Amazon Music app outputting directly to the headphone. That plays gapless too. Next I chose the cast icon in Amazon Music and picked my Chromecast Audio. Amazon admits to not dealing well with Chromecast and indeed there were large gaps between the tracks. I don't have the Windows or Mac apps so someone else will have to comment on gapless for those platforms.

    1. Gapless works great via the macOS app using the Kissin example you gave.

    2. I went back to the app and tried again: this time I got gapless playback (The Beatles - Abbey Road [remastered]). Unfortunately I don't remember which album did not play gaplessly the first time around but will keep looking and see if I can work out what went wrong.

    3. Good to hear gapless is working - hopefully for the most part!

  4. I am currently using Spotify for casual listening (still use CDs for focused listening). I’m intrigued that for an additional $3/month I may be able to stream CD quality music. My problem is I currently use a Chromecast Audio connected to one of the analog inputs of my 30-year-old Yamaha stereo receiver. So I'm looking for the best, economical way to be able to wirelessly stream Amazon HD to my aging receiver. It seems a little ridiculous to me that Google could come up with a device for $35 and nobody else seemingly comes close to that price. Bluos Node 2i is $500. Is it really worth 14 times the cost of my Chromecast? I'm skeptical. Maybe I'll finally look at upgrading my amp, but I've always liked the sound I get from it even 30 years later.

    1. For the price CCA can't be beat. But to be strict, nothing is gained going to hi-res digital source since it's D/A was measured at 91dB, vs. 96dB inherent to 16 bit CD red-book.


      Of course, that assumes your preferred music uses all 16 bits. Popular music, whether highly compressed modern masterings or classic recordings engineered for playback on LP's with dynamic range of 60-70 dB, is unlikely to exceed the CCA's capabilities.

      My lament with the CCA is host-device support. I can't cast to it with my Macbook, and I can't cast to it with my Fire tablet. I can and do use it with my and my wife's Android phones, but it's annoying that I have to switch devices when I switch playback source.

    2. Hey Allan,

      I took a quick look at the link you shared. I would have to spend more time looking at and fully understanding the measurements shown. And there really is no convenient way for me to know if the music I am listening to is highly compressed (is there?). My tastes are pretty eclectic, ranging from the Who to 21 Pilots and everything in between, Dean Martin, Ella Fitzgerald, Willie Nelson,Johnny Cash; pretty much everything except hard core jazz and classical. For my purposes the CCA works beautifully. My reasoning for keeping to the budget side of things is I would like to eventually replace my receiver with something that does it all (WiFi streaming, DAC, integrated amp, subwoofer EQ). The ELAC DS Series DS-A101-G looks very intriguing, but it seems limited in that it only supports Spotify Connect. Not knowing too much about streaming yet beyond the ease of casting with the CCA, I have more learning to do before I really consider an all-in-one solution. In the meantime I'm looking for the least expensive way that I could stream Amazon HD to my current receiver. Any and all suggestions are welcome.

    3. If you're trying to come up with a solution comparable price-wise to the CCA, about the only thing I'm aware of is a used Gen 1 Fire TV. They have a TOSLINK output which many 30 y.o. receivers support.

      Perhaps of interest:

      I run it through my Oppo 105 for DAC duties. Whether using the HDMI port or TOSLINK, the Oppo reports the signal input as "LPCM 2.0 48K." This is true both for Tidal "Hifi" and Amazon MP3 streaming. I haven't signed-up for Amazon HD yet, but if I were a betting man, I'd wager the Fire TV converts everything to 16bit 48k LPCM, which is CD quality.

      As far as Hi-Res streaming since you listen to everything except hard-core jazz and classical, you'll be fine with CD quality that the Fire TV provides for the reasons I previously mentioned.

      After the Fire TV, I'd say the next-best option if you are optimizing for price is the Raspberry Pi, though I doubt that's going to have Amazon support just yet (if ever). Our host has written an article or 4 ;) on them.

    4. Nice guest post! I did some more digging and DTS PlayFI is a preferred partner of Amazon Music. And Klipsch makes the Gate streamer that has PlayFi available and only costs $99. More research to do, but it may be just what I'm looking for. Or I could wait and see if Amazon HD becomes available on the CCA, but we all know Amazon and Google don't like to play well together.
      BTW, in my digging on this site I found where archimago tested the analog output of the CCA and concluded "is capable of 16/44 with ease", and that it's "achieving 17.5-bits of resolution". Again, I would have to look more closely at how and with what the measurements were done to know if I'm comparing apples to apples. Just interesting that measurements came out different.

    5. Yeah, I keep forgetting Archi tested the CCA long before ASR.

      ref1 -- distortion to dB calculator:

      ref2 -- dB to bit depth calculator:

      ASR used red-book. 16bits/44.1 and measured:
      0.002736% == 91.3 dB == 14.8 bits (ENOB)

      Archi used a far lesser ADC (relative to the AP555) and tested the CCA at 16/48, 24/48, & 24/96. His IMD+Noise measurements were (swept freq):
      0.0051%, 0.0029%, & 0.0028% ==
      85.4 dB, 90.8 dB, & 91.1 dB ==
      14 bits, 14.8 bits, & 14.8 bits

      I assume Archi used the dynamic range of 105.4 dBA for his "17.5 bits" statement.

      ASR measured Dynamic Range per AES17 as 98.4 dB.

      No idea if RightMark uses AES17 or something else for its dynamic range measurements. That could certainly account for their differences.

      Per ASR, "Dynamic Range is a test of playing (almost) nothing versus maximum signal."

      SINAD is a more detailed measurement of the "correctness" of everything in the analog signal that you (theoretically) are able to hear than the raw full-scale dynamic range.

      Now, having said all that 90 dB is very d@mn good in the real world. Over 90 dB is certainly good enough for me and the music I listen to. HTH.

  5. I tried a few things trying to reproduce the problem with gapless. This is what seems to be happening:

    I used to buy CDs from Amazon. They have a service "autorip" which automatically provides the buyer's Amazon Music account with an MP3 rip of the purchased CD. It's not something I required or asked for but there you are, it just happens. However, now when I add music to my Amazon Music HD library it turns out that even if I specifically add a CD or higher quality version of an album I previously purchased on CD I am stuck with the crappy mp3 version, gaps and all!

    If Amazon apps had some kind of bug tracker I would happily report a bug but they seem to value opacity and inscrutability over communication with customers, so I think I'll just stick to running my own Minimserver and BubbleUPnP set-up and save the subscription fee for beers or used SACDs and CDs.

  6. This comment has been removed by a blog administrator.

  7. This comment has been removed by a blog administrator.

  8. This comment has been removed by a blog administrator.