With all of the recent discussion about quantities of unidentified albums vs. performance, I thought this would help us all share and create insight - here we go…What is YOUR experience?
A) 0-100 unidentified albums; good performance
B) 0-100 unidentified albums; poor performance
C) 101-500 unidentified albums; good performance
D) 101-500 unidentified albums; poor performance
E) 501-1000 unidentified albums; good performance
F) 501-1000 unidentified albums; poor performance
G) 1001-5000 unidentified albums; good performance
Oh, and if anyone cares, I currently have 160 unidentified albums (out of 4,758) and have no issues. I started in 2022 with over 800, and never had problems back at that time, either.
An interesting poll but as it takes into account one factor only (and doesn’t define what good or poor performance looks like) it will be difficult to draw any conclusions from the votes.
However it has given me an idea to test this issue in isolation. If someone with some time and spare hardware on their hands could do the following:
On a known well performing Roon system with no known network issues:
Start with a fresh database
Load up a few dozen albums that are identifiable, wait for analysis to finish then verify what ‘good’ performance looks like.
Download 20,000+ files from the web that are unlikely to identifiable by Roon. I’m thinking of freeware sound effects (SFX) files as an example, as these are readily available and bulk downloadable in quantity as zip files. Getting hold of a large quantity wouldn’t be too difficult.
Add them to the Roon library and wait for the circle to stop spinning and any analysis to complete.
Repeat the performance test for the identifiable albums and verify if ‘good’ performance still exists.
Try playing a load of the unidentifiable files and assess performance.
If good performance still exists for both types of files, then unidentifiable albums in quantity and isolation make no difference to performance.
This isn’t quite ‘in isolation’ as such because the folder structure of the 20,000+ could have an impact?
Yes just because they are in Qobuz means nothing, as Qobuz is not an online metadata database and they have no metadata beyond artist album and track names. Roon can’t identify albums in your library from Qobuz they need to be in MusicBrainz to be identified and Allmusic/Wikipedia to get the extra metadata bios etc. albums in any streaming service bypass the identification process as the ids already been done.
That would not really be meaningful stress for roon according to my experience as these files usually are small, come in easy-to-handle formats such as MP3 and Flac and have a minimum of metadata and references to the existing library. So roon has to do just a minimum of crawling and computing related to these files.
What was slowing down my system and what I found to be a litmus test was rather classical anthologies and boxsets such as ´Bach´s Complete works´ or ´Karajan´s recordings on 330 CDs´ as well as lots of opera recordings and unidentifiable audiophile stuff in formats such as DSD128, DXD and alike.
Had a lot of such stuff combined adding up to 50,000 tracks and a significant amount of storage space (well north of 5TB). That brought my system to a standstill a year ago.
I still have most of these files in several migrated folders which I usually do not enable. If I do, things get slower immediately despite from me having a more powerful core now. Usually I keep it disabled and things are very smooth.
Would not expect roon to slow down because of some few 1,000s of unidentified tracks. It has to be a bit more and even in this case it is not clear if the sheer capability of local computing power and fast cloud connection would make up for the disadvantage.
James_I
(The truth is out there but not necessarily here)
8
Sorry to say my option isn’t in the poll: 825 unidentified albums out of 10400, intermittent poor performance.
That is what is so baffling about the performance issue. Sometimes a Roon page just won’t populate for a few minutes.
Fair point but the specific thread topic and issue under test is unidentified tracks. File formats would be another test entirely. One variable at a time.
I would love to see ROON stepping in here to help with the research. They are the ones to know what is going on, they should be interested in this kind of data.
Listen, I am just as curious about this issue as all of you since it was originally revealed. For the uninitiated, this is what we are all referring to:
That said, this poll can never be scientific; I thought that if enough people responded, we might be able to get some sense for whether or not the issue is real. There are certainly many cases of correlation not being causation, such as this one:
Frankly, if you look at the results as of this morning, where we only have 35 respondents, I believe it already reveals a lack of correlation. I suspect the folks in the know on the team at Roon are rightly rolling their eyes a bit at us right now, because @Joachim_Herbert 's original interaction wound up revealing that he had over 200,000 unidentified tracks causing his particular issue. I deliberately didn’t even include numbers that high in this poll. @Joachim_Herbert is an edge case for sure, and definitely a real one that needs to find a way to be addressed. But I suspect the issue is not affecting most of us as much as we would like to believe. There are other more mundane things going on that affect the performance issues that many people may have.
That might be an explanation why they are not too eager to address these problems fundamentally: They might simply know it is affecting a very limited number of users and whenever they investigate there is a ´bad performance smoking gun´ to be found (such as the 200,000 unidentified tracks). At least that was how I understood statements of roon staff here responding to performance reports.
It is a legitimate position but I would nevertheless wish we as users would know more about the technical background in order to act accordingly… How can we optimize our library? What’s the limits? How to understand which aspect is slowing down things the most? Which ISP to choose?
I think publishing a guideline or workaround hacks or simply adding some sort of ´overload feedback´ to the software would be helping people being affected as well as roon´s reputation of being a company who actually listens to their users.
Some good theories there for sure. All of these point to a benefit in the development team creating an ability for us to mark albums for “identification exclusion” for lack of a better term.
with no guidance good vs bad performance are subjective measures and could be argued to have as much to do with the mood of the scorer as to the speed of Roon - let alone the variables of library size and of server processing power/memory. As such it will be hard to draw any conclusions even with a very large sample size
Agreed. And it is all there. Unidentified files already carry a flag to indicate this. It woul be sufficient to have a setting that excludes these from identification and run a identification batch manually or once a month.
I tend to look at it the friendly way as I do have a very positive opinion on roon as an enthusiast´s company not driven by greed. Neither do I see any base for a theory that they would not provide the best service they can out of indifference or will to throttling performance or whatever attitude towards their customers. I think even talking about that is unfair.
Would rather compare it to car manufacturers´ attitude to people taking their street cars onto the racetrack. Yes, it is theoretically possible, we are impressed, 99.9% of our customers don’t do it, and in this case we deny all warranty. Pretty fair, maybe some additional communication beforehand would have be helpful. I mean, standard Nucleus also came with a warning that 100,000 tracks is the limit. Make it 50,000 with max. 5,000 unidentified and everyone is aware of the limitations.
I am pretty sure the absolute majority of cases of bad performance has to do with a disadvantageous combination of relevant factors, may it be library size, CPU power, internet connection responsiveness, library complexity (including unidentified) or whatever. A combination that differs from user to user and you can test 10,000 systems, they all work flawlessly, and number 10,001 gets sluggish for whatever reason. You might call it a design flaw but that is how the software works and the base for all others to have an experience other systems do not deliver (I have recently visited a friend using amazon music and I wished my roon back almost every 10 seconds).
Nevertheless I hope we all can help finding out which factors are slowing down roon making it possible for users whose system is affecting to improve things, or roon team to implement some routines. Appreciating the effort of starting a poll @DDPS!
