Solid Snake-Oil Storage: This SSD Is Aimed at Audiophiles

[keith_correa]

So if I write a file from RAM to the "noisy" SSD, read the file back from SSD, compare what I read, back to the original source in RAM, the delta would be "noise"?

 

[aarmath]

There is so much misinformation here - it is crazy. This item is pure snake oil. The SSD controller is an arm Cortex CPU and it runs sophisticated code in its firmware to make sure data integrity is maintained. It even detects and isolates any bad blocks due to wear. There is built in error correction. Otherwise what you write and what you read off an SSD will be very different which is never the case.

Timing issues occur in real-time transmission only when clock jitter can cause unrecoverable errors where there is no error correction such as spdif and to some extent USB. That is absolutely not the case here where error correction is in full force and the data pipe is nowhere close to saturation.

Here data is first read into the SSD cache, error corrected and then sent to the memory location pointed to by the host CPU. The total amount of audio data read is miniscule compared to hordes of data read for say something like a video game. If there are no issues there, for hell we can be sure there would be no issues with audio files.

Also anyone who uses SSDs extensively knows that the realtek controller on this SSD is a piece of crap. One is way better off with a WD SN850 or a Samsung 980 Pro.

Thank you for giving voice to my thoughts succinctly & to the point. & Also reaffirms my faith that there some good & well-informed folks as well on Hifivision.

 

[reignofchaos]

Thank you for giving voice to my thoughts succinctly & to the point. & Also reaffirms my faith that there some good & well-informed folks as well on Hifivision.

If only they spent more on actually improving the firmware of that controller, it would be far better off than adding bells and whistles. Every realtek product is affected either by firmware or driver issues of some sort

This controller is used mostly by bargain basement SSDs.

 

[keith_correa]

Sony "invented" the concept first!!!

 

[square_wave]

Just for clarity sake, while talking about noise and gremlins introduced into analogue chain, I was not talking about SSDs in particular. It was a general statement. There have been many instances where I was dealing with the same data and same dac but differences were heard depending upon the methods / design involved before it was fed into the dac. I dont know enough about SSDs.

 

[raghupb]

Let me wear my EE hat here and clarify a few things.

Noise: It affects the signal's amplitude from nominal
Jitter: It affects the signal's timing from nominal

Let's talk about noise for a bit (many an HFV thread is noise )
First, electrical/electronic noise is always there. How much depends on how well the circuit is designed and the power supply.
Let's assume that we have a perfect PS (for the sake of it).

In the digital domain, when a component is operating (an SSD or a processor) it is switching it's outputs (from 0-to-1 and vice versa).
This switching causes what is known as ground bounce. Meaning, the "ground" which is supposed to be at "absolute" zero is seeing some variation.
Thus the noise floor of any electronic component is not truly "zero". It is some uV (peak).
Typically, a good circuit design will have enough decoupling between I/O power supply and "system" ground.
The ripples that switching causes is mitigated to a certain extant.
Does this possible "increase" in noise floor matter in audio? Bingo!! Here comes the debate.
My take, if the component or next one is capable of working with a "given" SNR, NO it does not matter.

Say a streamer/player/processor "injected" switching noise into the noise floor. What happens?
The SNR decreases a wee bit. Why? The noise floor just got bumped up.
Remember, ground is always common between components (theoretically, only additions to noise floor happen).
Is the SNR within acceptable limits of the component that's processing it (say a 16-bit DAC)?
If yes, then you won't hear a thing different. If no, you may hear some things missing or buried under the noise floor.
This if you have "bat ears" or a very good pair of headphones (still unlikely).

Specifically, about SSD. Is it really required for audio? No!!
Will it "inject" noise due to switching? Yes.
Will this "injected" noise affect SQ? No!!
What does the SSD do then? Burn a hole in your wallet.

Cheers,
Raghu

 

[IndianEars]

Regarding the SONY SD Sound card, I think its only fair that some background be provided for this. The What Hi Fi Site (2015) says:

The SR-64HXA micro SDXC memory card is due on sale in Japan next month for 18,500 yen, which translates as around £100. That makes it around five times more expensive than the average memory card.
Why would you spend the extra? Sony claims the card helps to "reduce electrical noise generated when the file [your digital music] is read". So if you store music on a memory card, this super-SD should help suppress electrical noise, which can affect sound quality. That's Sony's theory, at least...
The Sony Japan website suggests while the capacity and speed of cards can be the same, the parts and materials can be different - and, so Sony says, that can make a difference to the sound. It is 'coincidental" that noise comes up in Sony's reasoning ?

OK Gentlemen....

All those who cannot hear differences in the sound from 2 Bit Perfect Playout softwares... (eg J River & Roon), stand to the Left, as the "Informed Group"

All those who can hear differences in the sound from 2 Bit Perfect Playout softwares... (eg J River & Roon), stand to the Right, as the "Enlightened Group"


Despite "The Audiophile SSD" being the current poster boy for Audiophile extravagance, I am not so sure I would write it off in such an off handed manner..... I think it deserves to be heard, before passing judgement.

There are several tech innovations that deserve a listen.

An LPS feeding an entire PC with about a dozen SMPSs on the motherboard, feeding individual sections, is not just OK, but desirable.... But if an SSD takes juice directly from the LPS, it is snake oil ?

Femto clocks are great everywhere, but on the SSD, its Phoo Phoo ?

SSD Controllers are considered the key for SSD performance, but when a custom Controller is implemented in this SSD, it attracts ridicule ?

Running the SSD for peak performance and endurance .... (pSLC) mode ... is acknowledged as being good but unaffordable for PC work, yet when implemented for SOTA Computer Audio its bullshit ?

Lets think this over guys, and not join the heard in ridicule ....

 

[reignofchaos]

Despite "The Audiophile SSD" being the current poster boy for Audiophile extravagance, I am not so sure I would write it off in such an off handed manner..... I think it deserves to be heard, before passing judgement.

There are several tech innovations that deserve a listen.

An LPS feeding an entire PC with about a dozen SMPSs on the motherboard, feeding individual sections, is not just OK, but desirable.... But if an SSD takes juice directly from the LPS, it is snake oil ?

Femto clocks are great everywhere, but on the SSD, its Phoo Phoo ?

SSD Controllers are considered the key for SSD performance, but when a custom Controller is implemented in this SSD, it attracts ridicule ?

Running the SSD for peak performance and endurance .... (pSLC) mode ... is acknowledged as being good but unaffordable for PC work, yet when implemented for SOTA Computer Audio its bullshit ?

Lets think this over guys, and not join the heard in ridicule ....

Lets get to the snake oil in these comments:

There are no SMPS stages on the motherboard - not sure who told you this. What you have in the power stages are voltage regulators (also called with fancy names (VRM, DrMOS and what not) that take 12v and bring it down to the 1v range for the CPU. If the CPU is running fine and not out of spec, there is nothing more that needs to be done. As for SSD - it takes 12v and 5v from the power rails and probably has its own voltage regulator.

As for clocks - it is debatable as it is in the digital domain - yes clock drift might cause trouble but then there are so many clocks in a typical PC, changing one doesn't make sense.

This SSD doesn't have a custom controller - it is a bog standard realtek with bog standard realtek firmware. This is its achilles heel to be honest.

If you want to run an SSD in SLC mode, better put SLC nand in the first place instead of TLC nand. This is a jugaad - and FYI all TLC drives have a significantly big SLC cache (~8-60GB depending on capacity) and onboard DRAM of around 1GB per TB of storage. Running in SLC mode does no help to the endurance as the cells are still TLC. Also endurance is taken care of in the firmware if the firmware is written correctly - it is called wear levelling so that all cells get written into instead of bombarding only certain cells.

Unfortunately the firmware of all realtek controller based SSDs is semi-garbage. If they really wanted to do something good, they should have started with a Phison or Silicon Motion controller as those are more mature and more performant.

This thing *is snake oil*. No two ways about it.

 

[IndianEars]

When I said SMPS, it was used loosely.

The Voltage Regulators that you refer to are SWITCHING Voltage regulators, and the key issue here is the generation of switching noise by these SWITCHING Voltage Regulators. Actually they are SWITCHING DC to DC converters... Far more Noisy than Linear Voltage regulators, or the DC available from an audiophile grade Low Noise LPS.

The Audiophile Grade SSD has the option to feed it "Clean" 5 VDC via a low noise LPS, rather than getting this jiuce from a switching regulator or a PC Multi output voltage SMPS.


Clearly you do not like Realtek .... But Realtek has a Great history and rich experience in PC Audio. I suspect there must be good reason the Realtek controller has been chosen for an audiophile SSD, rather than use the stock controller..... You may or may not like his choice of Realtek, but clearly he has made a conscious choice, based on factors he has deemed important for an audiophile SSD. Again, I would not call a decision I disagree with, as snake oil !



You also have your views against (pSLC) mode. The computer industry does have proponents for (pSLC) mode, ofcourse with its overhead costs of greatly reduced capacity. It would be unfair to dispose off pSLC as snake oil in one wide stroke.....


To clarify my stand, I am not saying that the Audiophile SSD is the greatest development since sliced bread.... Just that there may be serious effort that has gone into the development, and I feel it unfair to pass summary judgement... particularly as an audiophile who has an open mind on the subject and the effort put in by the developer.

 

[reignofchaos]

When I said SMPS, it was used loosely.

The Voltage Regulators that you refer to are SWITCHING Voltage regulators, and the key issue here is the generation of switching noise by these SWITCHING Voltage Regulators. Actually they are SWITCHING DC to DC converters... Far more Noisy than Linear Voltage regulators, or the DC available from an audiophile grade Low Noise LPS.

The Audiophile Grade SSD has the option to feed it "Clean" 5 VDC via a low noise LPS, rather than getting this jiuce from a switching regulator or a PC Multi output voltage SMPS.


Clearly you do not like Realtek .... But Realtek has a Great history and rich experience in PC Audio. I suspect there must be good reason the Realtek controller has been chosen for an audiophile SSD, rather than use the stock controller..... You may or may not like his choice of Realtek, but clearly he has made a conscious choice, based on factors he has deemed important for an audiophile SSD. Again, I would not call a decision I disagree with, as snake oil !



You also have your views against (pSLC) mode. The computer industry does have proponents for (pSLC) mode, ofcourse with its overhead costs of greatly reduced capacity. It would be unfair to dispose off pSLC as snake oil in one wide stroke.....


To clarify my stand, I am not saying that the Audiophile SSD is the greatest development since sliced bread.... Just that there may be serious effort that has gone into the development, and I feel it unfair to pass summary judgement... particularly as an audiophile who has an open mind on the subject and the effort put in by the developer.

That is a given - You can't build a voltage regulator for a modern CPU that consumes 250W with linear regulators - it's simply not possible physically. Only way around it is to get a low power CPU which can be powered by a 2-3 phase regulator. Also these operate in the MHz domain - the noise they generate will be in that range not audio frequency.

You can feed 5v and 12v DC directly from a linear psu to any SSD with a sata connector - actually I'd argue SATA is a better interface if you want to get there as it is completely isolated from the motherboard in the first place compared to NVMe and only connected via a SATA cable. You can still get real 2 bits/cell SATA drives (Samsung 850 pro) which will be more than enough instead of SLC emulation over TLC. The endurance of MLC is exponentially better than TLC. The transfer rates that SATA offers are more than sufficient for any kind of audio file.

Realtek has no history with PC audio - it makes garbage audio chips that sound like $hit and garbage network chips that drop connections randomly. Any internal sound cards worth their salt use CMedia chipsets. Realtek's NVMe controller is equally bad - loads of people complaining about sudden disconnections. Only thing realtek does is sell things for cheap - ultra cheap. Probably that was the only motivation here.

I'll keep an open mind to products that actually provide me some benefit not sell snake oil.

 

[venkatcr]

I want all our learned people to explain one thing to me. Let us say I take a 'cheap' ssd and store and read a spreadsheet containing data with up to the sixteenth decimal point. In all my years of using HDDs including SSDs, I have never had a spreadsheet coming up with a data that is in any way altered from what I put in. I still use spreadsheets, Word files, videos, and even highly dense images that stay the same year after year after year, for over 40 years.

I genuinely have not lost a single byte from any file. You can argue that I will not be able to see the differences in an image file. Agreed. But, if a storage is affecting an image file, it should also affect text or numbers in the same storage that I can easily discern.

My understanding is that disk storage does not or cannot identify the difference between file types. So whether it is an audio file, a video, text, image or just about anything else, it is just a block of data in digital from. Thus if a HDD is injecting noise into an audio file, how come it is not doing it on other files. Forget about what I store on my hard disks. Financial information, stock information, security information that are very precious are also stored in hard disks. How come they are not affected? In large storages, back ups are there. But everywhere, the system identifies whether it can read or not read from a block or a sector. If it faces issues, it just ignores that block and reads from a backup. It then marks that block as a bad sector and copies from the backup into a different sector. SSDs also have internal algorithm that verifies the health of each sector and blocks sectors that cannot be used. There is no way a storage or accompanying systems can get 'bad' data from a storage. In 99.99% of the cases, what you put in is what you get out.

I can understand 'noise' coming into an audio file before or after the storage, during transmission, or amplification. But if a storage device works well with some digital files, it must work well with ALL digital files. Am I not right in thinking that you are fooling yourself if you think audio files are stored differently and that a storage device can introduce noise into only an audio file?

I think the following two statements from Tom's Hardware should put all these arguments to rest. Italics are mine.

"Regardless of the tech surrounding the chips, at their most basic level, SSDs store data in binary 1s and 0s, meaning that you can only have 1s and 0s of data. Simply put, adding extra gadgets or power isolation to an SSD can't change those 1s or 0s into something better, and any rarely encountered errors are immediately corrected through various types of error-correcting code.

If anything, the SSD should be very fast simply due to the nature of pSLC, and have crazy overkill power conditioning. Besides that, it's hard to believe that it would provide a better audio experience. That's because SSDs are incapable of altering audio quality — that's all done by the audio processing unit on your computer."

'The audio processing unit on you computer' is what happens before and after the storage.

 

[amit11]

Here the issue is not at all related to data getting corrupted. So audio file is lying in hard disk or lying in ssd, it does not matter. Data will be perfectly read.

It is the connection (electrical connection) which gets disturbed (electrical noise profile) and it impacts the circuit of receiver.

And that is the reason many have experienced that playing song from hard disk (which has motor) sounds less enjoyable than ssd.

Similarly.. different ssd have different circuits inside them... and those can contribute to different kind of electrical disturbance.

 

[fLUX]

That is a given - You can't build a voltage regulator for a modern CPU that consumes 250W with linear regulators - it's simply not possible physically. Only way around it is to get a low power CPU which can be powered by a 2-3 phase regulator. Also these operate in the MHz domain - the noise they generate will be in that range not audio frequency.

You can feed 5v and 12v DC directly from a linear psu to any SSD with a sata connector - actually I'd argue SATA is a better interface if you want to get there as it is completely isolated from the motherboard in the first place compared to NVMe and only connected via a SATA cable. You can still get real 2 bits/cell SATA drives (Samsung 850 pro) which will be more than enough instead of SLC emulation over TLC. The endurance of MLC is exponentially better than TLC. The transfer rates that SATA offers are more than sufficient for any kind of audio file.

Realtek has no history with PC audio - it makes garbage audio chips that sound like $hit and garbage network chips that drop connections randomly. Any internal sound cards worth their salt use CMedia chipsets. Realtek's NVMe controller is equally bad - loads of people complaining about sudden disconnections. Only thing realtek does is sell things for cheap - ultra cheap. Probably that was the only motivation here.

I'll keep an open mind to products that actually provide me some benefit not sell snake oil.

Agree realtek sound chips did not sound good in the past, was very evident when such audiophile jargons were not too common for me and even with fairly entry level audiophile headphones while using Asus Xonar DX vs the onboard realtek chipsets. Have not used onboard chipsets since a long time now so cant comment on the current condition.

 

[IndianEars]

We are back to "Since its digital, its perfect"

Wonder why they have error correction on HDDs and SSDs and for Digital transmission too....

 

[venkatcr]

In answer to Amit11

"Here the issue is not at all related to data getting corrupted. So audio file is lying in hard disk or lying in ssd, it does not matter. Data will be perfectly read."

So we agree on that. Great.

It is the connection (electrical connection) which gets disturbed (electrical noise profile) and it impacts the circuit of receiver.

Now we are adding a completely new paradigm to the argument. In any case, your 'receiver' is outside the realm of this discussion.

And that is the reason many have experienced that playing song from hard disk (which has motor) sounds less enjoyable than ssd.
Similarly.. different ssd have different circuits inside them... and those can contribute to different kind of electrical disturbance.

And then you come back to the same point that I don't understand. You just said data will be 'read perfectly'. So how can data from one storage be less 'enjoyable' than another? Does that not sound illogical to you? Even assuming you are referring to read speed that might introduce a sync issue, this does not in any way reflect on the storage's capacity to store and deliver digital files (of any kind) perfectly. Moreover, audio files are so small that they can be loaded and stored in the memory to mitigate any read speed issues.

Cheers

 

[OM_2K19]

No data is lost, whether HDD or SSD, irrespective of the storage medium, period. There are enough guard rails in a purely digital domain at both h/w, s/w levels to counter any bit of corruption during transmission. But in a mixed environment, it's not only the bits that we receive from the medium but the line noise that accompanies it also contributes a significant role. If these line noises are not taken care of, they have the potential to compromise the DAC's ability to reconstruct the Analog signal properly. The Audiophile Grade equipment tries to minimize this effect of various noise on the final Analog signal via different techniques. They will not re-invent how data is stored in the disk but try to reduce the line noise passed to the audio chain. But again, it's debatable how much we can hear the difference.

 

[venkatcr]

Wonder why they have error correction on HDDs and SSDs and for Digital transmission too....

Digital system were originally not built for audio, but for critical and non-critical data. You can easily and accurate identify any alteration in the data. Data were being used and transmitted over long distances. It is quite normal to assume that errors would occur in this process. At the same time, it was very easy to introduce error identification and correction algorithm into the data packets that could be verified by the receiver. Checksum is one of the simplest algorithm, though now there are much better and more robust algorithm.

The beauty is that, irrespective of the its original form, once converted to digital form, it is just data and is dealt with in the same way to ensure that packets arriver in perfect order. That is the advantage of digital storage and transmission. And this stays true for all types of information - audio, video, images, text, etc., etc.

Analog systems usually work on very small paths (read distances), and they have no way to identify or correct errors in storage, transmission, on in the ultimate playback. And this could simply be because the flow is in one direction only.

Cheers

 

[venkatcr]

No data is lost, whether HDD or SSD, irrespective of the storage medium, period. There are enough guard rails in a purely digital domain at both h/w, s/w levels to counter any bit of corruption during transmission. But in a mixed environment, it's not only the bits that we receive from the medium but the line noise that accompanies it also contributes a significant role. If these line noises are not taken care of, they have the potential to compromise the DAC's ability to reconstruct the Analog signal properly. The Audiophile Grade equipment tries to minimize this effect of various noise on the final Analog signal via different techniques. They will not re-invent how data is stored in the disk but try to reduce the line noise passed to the audio chain. But again, it's debatable how much we can hear the difference.

No data is lost, whether HDD or SSD, irrespective of the storage medium, period.

It is great we all are in agreement here. I must add that data is also not altered in any way.

But in a mixed environment, it's not only the bits that we receive from the medium but the line noise that accompanies it also contributes a significant role. If these line noises are not taken care of, they have the potential to compromise the DAC's ability to reconstruct the Analog signal properly.

As I said before, none of these have anything to do with storage or storage mechanism.

The Audiophile Grade equipment tries to minimize this effect of various noise on the final Analog signal via different techniques. They will not re-invent how data is stored in the disk but try to reduce the line noise passed to the audio chain. But again, it's debatable how much we can hear the difference.

Again, this has nothing to do with storage. Line noise happens only when the data is outside the storage. A good Samsung, Crucial or any other brand will and can store and deliver audio without any issues or errors. Adding a gold plated capacitor to the storage electronics or powering it with an external source, is, as stated by Tom's Hardware pure snake oil. As I said before, if it can store a 500 page document file without errors, why should it have issues with a 250MB audio file?

Cheers

 

[IndianEars]

Hence, it seems you are saying that Digital DOES Get corrupted, and 0s & 1s dont always remain the same ....

I agree.

That is why, a SOTA implementation tries to minimise if not eliminate errors, rather than correct them later.

The Audiophile SSD seems to be an effort in minimising errors (I'm not saying its successful or otherwise) ,

However, the "Digital Is Perfect ... Forever" brigade and some others just don't appreciate or accept that there are second order (and less obvious to the partially informed) effects that affect Digital.

 

[OM_2K19]

As I said before, if it can store a 500 page document file without errors, why should it have issues with a 250MB audio file?

You are missing the point; no issue with the 250MB of music; it's a bit-perfect. However, when those 0 & 1 arrived at the DAC, they also carried line noise with them, and this noise affected DAC's ability to reconstruct the original signal. Where is that noise generated? It could be generated at the drive's storage/USB/power controller. Is storage in question here? No, and as I mentioned before, they are not inventing how data is stored in the storage medium—but employing different techniques at various stages, probably at the storage/USB/power controller level, to reduce this line noise. Cheers.