SDCard performance discussion

#1

this topic is linked to the sdcard benchmarks topic, so we can keep the discussions separate from the numbers.

also useful for general discussion on how to improve SD Card performance, or related thoughts.

SDCard Benchmarks
Playback problem with more than 4 .wav files
#2

I think it has to do with the Class of the SSD
https://www.sdcard.org/developers/overview/speed_class/
sdcards get worse after lots of reads/writes and data can become more fragmented which influences speed (especially when running out of space).

Class 2, 4 or 6 is probably not fast enough.
not sure about class 10 .
Just get a UHS 3 (or 1) , and it should be fine.

#3

Card speed class rating is not consistent with small buffer access performance.

#4

agreed though my cards are all new smile

'not fast enough' for what?

rather depends on your plans...
I only want to use for patch storage, so class 4, are fast enough... as I'm not interested in samples... even i was to use samples they would be very short, and I would probably load into RAM (of flash)

also as @johannes stated,

so if your only dealing with one or two samples you should also be fine.

finally, perhaps we can 'tweak' the performance by using different buffer sizes etc?

#5

EDIT: moved from another topic

what buffer size does the play fn stereo use? perhaps this should be configurable?

it seems, that using larger buffers would increase throughput - no?
also does SD card performance improve if you have sustained throughput?

Id have thought we don't really have enough RAM for anything other than a small number of short samples.

stereo, id have thought wrapping stereo functionality within an AXS would be reasonable enough... whats the difference between stereo tables, and 2 mono tables?

@AdrianWillaert what card is it your using?

#6

@johannes
something a bit odd happened here...
I decided to test my 4 axoloti, all with identical cards (bought at the same time etc)
Brand : Kingston Technology
Type : Micro SDHC
Size : 4GB
Class : 4

3 reported: (axo1,axo 2, axo 4)
Axoloti says: write BW : 380 kB/s
Axoloti says: read BW : 766 kB/s

but the forth reported : (axo3 )
Axoloti says: write BW : 233 kB/s
Axoloti says: read BW : 1724 kB/s

I ran the tests multiple times, formatted the card multiple times, and still always the same... very consistent times.

so decided to open up, and swap the cards between axo 3 an axo 4
380 kB/s / 766 kB/s
i.e. the 'faster one' was now consistent, and axo 3 was consistent
again formatted cards, same results

swapped cards back i.e. so axo 3 had the 'fast card' again.
but now was back to the same 380 kB/s / 766 kB/s
formatted card same result...

as i say cant be formatting, as i reformatted card... doesn't appear to be card, as its 'slowed down' now, and doesnt appear to be board (axo3) as that is now consistent.
it doesn't appear to be card/software/board consistency, as I've run the test many times, and each time got the same numbers.

the only thing i can think, is that its board initialisation, as when I changed cards I powered down the axoloti boards.
thoughts? or is this just normal 'sd card quirks?'

EDIT: ok, something fishy going on...
decided to test another card:

Brand: Transcend
Type : Micro SDHC
Size : 8GB
Class : 10
Axoloti says: write BW : 225 kB/s
Axoloti says: read BW : 1456 kB/s

in my prototype board... same firmware etc.. note its a class 10.. but notice how the performance is kind of similar to what I got with the axo3 for a while. (again reformatting/powering down/multiple runs all same numbers) .
odd no?

#7

For some people (like me) this website could be useful before buying a SD-card.
It explains all the codes and also lists the fastest cards around (though not sure how new this page is, it seems quite up-to-date though).

http://www.bhphotovideo.com/explora/photography/buying-guide/fastest-memory-cards-money-can-buy

#8

does anyone have any advice on best practices for storing audio files on the SD?

i like loading longer audio files on the axoloti and then processing them inside the Axoloti, but currently longer WAVs (5min +) take forever to load. not complaining (because i like this workflow anyways), just an observation

also, how can i get a log of my read/write speeds? might be useful as i run 1.0.8 on Windows 7 and OS X and wouldnt mind comparing speeds

#9

@workerparasite303

Well, largest tablesize in version 1.06 is 2097152 samples long.

At 48000hz, which is Axolotis sample rate, there is 43,691seconds of table to load sample into. Everything more than 43,691 seconds will cut off, truncated. You wont br able to play anything abovr 43.681 secs.

I think Axoloti version 1.08 has got twice that tablesize, so around 1 and a half minut is possible in newest TEST version. Everything more that one and half minut will be cut off.

So loading 5 minutes doesnt really do you any good, it will only make loading time very long. Maybe split the file up in smaller portions that fits tablesize of Axoloti?

#10

correct, i was totally wrong and attempting to load a 44.1/16bit stereo WAV at 5, 6, and 7 minutes did not work. going to split them up as you suggested and, for now, just process using the audio in

#11

Maybe we can look into the table object code and make a new version that load even bigger files. Dont have my boards atm , but when I get it back i'll look into it. Would love to be able to load 5 minute files :smile: