2. digital
  3. 192922-142926

BS meter is pegged!

I was reading about a music streamer from a latest Stereophile review and what was posted in the review had my BS meters pegged. I'm from the high tech industry with friends that work at Intel labs and friends that work for ARM computers and they haven't heard of some of these things that were posted. Maybe we can get clarification on these items so they don't sound so far fetched and the specifics posted in the review tainted the reviewers judgement IMO.

1) The review states this piece uses "a cpu that's highly prioritized for audio playback only ensuring highly optimized sound quality". I asked around if somebody is making a specific CPU for audio playback only. You know the Intel/AMD fabs that make cpu's make millions of them at a time, not 10-1000 custom cpus. Even when you look at the ARM cpus, none of them are built specifically for audio. There are millions of servers in the world that do database work for example that no cpu maker is building a specific cpu for database only applications. If there is a small company that are creating this kind of cpu, what kind of OS will run on it? This piece runs Roon so it has to be a somewhat generic cpu with a generic Linux OS running on it.

2) the review states: this unit "it plays live with no other processes running in parallel. as far as we know, unlike any other streamer on the market, this streamers cpu plays directly and live from the kernel without any processing or lag." Meter is pegged now. NO OS will run only 1 process at a time without hundreds of other system processes running in parallel or in the background. Using Unix/Linux, the OS is always in a flux state moving data around in its caches, in and out of memory, doing read a head, swapping, paging, etc... And these system processes are a good thing to keep the system stable and running efficiently. 

3) this piece uses "new and faster enhanced memory". Meter is pegged again. During the last 2 decades using Linux servers and over 2 decades before that using Sun and IBM UNIX servers, I have never had the option of buying enhanced memory. I made a couple of calls and asked if they had any enhanced memory that they could sell me and they had no clue what I was talking about. Everybody can get fast memory but "enhanced"?

4) "the whole device plays 1 song directly from RAM". All linux OSs do this, you cannot go from any cache or ssd/hdd directly out of the computer, the data has to be read into ram 1st.If the system is paging, this data might be deleted from RAM and then have to reread into RAM before sending to a dac. I used many large PCIE cache cards to hold large amounts of data (used it as a database cache) but that cached data had to be moved from this fast cache to ram before sending out to the dac.

Most of the time, audio reviewers get psyched up when they hear new acronyms or a magical cpu or enhanced memory that taints their judgement. For example, this reviewer at the end stated "never before have I reviewed a stand-alone streamer/server so accomplished in the hardware department". 

Maybe somebody could clarify some of this up for me/us in the audiophile community.

p05129
secretguy
1,751 posts
 

BS in high end audio?!?!?! I'm shocked! SHOCKED, I tell ya!

oberoniaomnia
91 posts
 

There are other instances of objective BS being spouted by audio reviewers. My "favorite" example is Michael Fremer in The Absolute Sound last year, who re-parroted in a TT review with respect to speed accuracy (wow and flutter) that 1 arc second is audible. LOL! An angular distance metric vs speed change over time (dv/dt). And once you start calculating potential audible pitch or frequency changes and required accelerations over distance of 1 arc second, the whole idea collapses as well. 
Upshot, I cancelled my subscription to TAS again. When entertainment turns into annoyance, it ain't fun anymore.

erik_squires
16,330 posts
 

I’m not about to defend Fremer or John Atkinson in general for that matter, but I don’t think this is BS so much as mostly inaccurate translation from geek to consumer.

As I would translate it back to geek-speak, I’d believe (maybe inaccurately) they are using something like a real-time OS/kernel which are real things and used in areas where a specific task (like reading a heartbeat, or car sensor) takes priority over user interface actions.  The features of a real-time OS is that the response times of certain actions are bounded in time and unaffected by other work done by the CPU, something the standard Unix/Linux/Windows OS can't do.

It sounds like they've reduced the processing as well to the bare minimum required while playing a song, a good thing. 

As for enhanced memory, that could be anything like using the latest generation, or fastest possible for the CPU.

Lastly, yes, music data must be read from disk to memory before it is sent back out but it’s not necessarily done 1 entire song at a time. This is closer to the old Parasound CD player which would read the entire song into RAM before decoding. Streaming of data, whether a very large text file, a DB or song is not uncommon.

Overall I’d say they are just describing the care that goes into feeding the DAC. Of course it’s all quite glorious sounding.

p05129 OP
512 posts
 

Erik-you have been able to assign apps to kernels for decades, this is not new. Also, how do you know if Roon or audirvana don’t do this same thing? It could but the developer needs to be sure the kernel version will work.

I disagree with you on limiting processes because the article stated only 1 process is going on while doing its thing and any Linux person would know that’s physically not possible. Over 15 years ago when I was using audirvana, we went thru the he process of limiting any OS tasks that might interfere with audio.

Also, the reviewer stated “fast enhanced” memory. We all put in the fastest memory for the cpu that’s nothing “enhanced”. Even using “ecc” memory isn’t enhanced memory.

Both the OS and the app can process read ahead so multiple tracks are in memory, not new. Over 30 years ago when designing we based apps, you would do read ahead and build pages before the user clicks next for faster processing. Also, audirvana had hog mode in the late 2000’s.

My point in posting this was to point out how reviewers are gullible to synonyms or technology that manufacturers and grow out that taint their reviews. New people to audio look up to these reviewers for help and when the reviewer is clueless about what the manufacturer states, that’s on the reviewer. The reviewer should have either looked into some of these claims to verify them and get clarification from the manufacturer and then if the manufacturer claims these claims are true and if are in deed false, then it’s on both of them.

Maybe the reviewer can clarify these concerns or the manufacturer could explain some of these issues.

dean_palmer's avatar
dean_palmer
85 posts
 

I also work in high tech manufacturing, and my current position is Global IT Director - Applications for one of the top-tier electronics contract manufacturers on the planet. All that wording perks up my BS meter too, but I can see where the translation from the actual tech-speak to something useful for marketing and explaining to the masses could be the issue. I'm not even good at it, so I will not even try as others have already done that :-) 

More to discover