PSA: Don't update to iOS 12.4 if you use apps in AB output slot or IAA apps that use the mic.
https://forum.audiob.us/discussion/34030/urgent-psa-hold-off-updating-to-ios-12-4-if-you-use-apps-in-output-slot-in-audiobus

new iPad Mini performance test

Received my new iPad Mini 5.

Here first quick "real world" test... in case anybody is interested in testing some particular plugin - let me know, if i have it, i try it. Of course i can try also other DAW (i have BM3, Cubasis2, Auria Pro, Gadget). Just let me know :)

Ok..

My first test is six instances of ModelD inside Nanostudio 2. Buffer size was set to high (which is 512 samples), sample rate is 48khz (*)

CPU meter shows steady 79% with random jumps to 78 or 80 (most of time it's 79). Playback totally smooth, no dropouts, even when moved to background. UI totally responsive, smooth ...

Quite nice amount of power, considering the fact, that my old Air 1 can't handle even two instances :)))

«1345

Comments

  • Show off ! hahahaha B)

  • edited April 5

    dendy:
    hell yeah:)
    pls try more fx, samplers, other hosts
    i'm mostly curious about fab plugs but cpu hogs like
    Aparillo and Quanta...:)

    thanks for your time and skills!!!!
    ps:
    pls also offer general impressions of the mini, i'm looking to buy an Air 3...

  • edited April 5

    Interesting.
    @dendy Do you also own Gadget and Lexington?
    I'm curious how many instances of Lexington repeatedly playing chords with 5-note polyphony are possible. At 2048 samples buffer size (Settings > Gadget > Audio Latency 2048).
    Thanks (if possible)!

  • @dendy That's great news and fits with the sort of performance I see with my XR. The iFixit teardown revealed that they're remarkably similar from a component perspective.

    The only recommendation I'd have from the XR is don't attempt to push it too far with the sample buffer. Having said that, mines pretty much permanently set at 256 and I only raise that on rare occasions.

    Can't wait to take delivery of my Mini 5!

  • @waka_x 18 instances of Aparillo playing default patch in same conditions (same buffer size, CPU load around 78% steady). It can handle also 19 instances, then cpu load is around 88% but there are ocassional dropouts and when i put it on background audio playback stopped (obviously cpu overload) - with 18 instances smooth, no dropouts(even on background)

    Regarding other daws - it's +/- same.. tried same plugins in BM3 and Cubasis (512 buffer in BM3, "medium" buffer in Cubasis and with basically same result - just other daws hav a bit laggy UI when CPU load is around 80% - nothing major, just you feel that you are starting pushing limits of device - this is probably because NS2 uses for UI directly OpenGL, not Apple UI libraries)

  • edited April 5

    @rs2000

    at buffer 512 it handles 9 instances of Lexington - default patch just tweaked to 8 voices unison - smooth even on background. 10 instances - until you are inside app, no dropouts, if you put app on background there are occasional dropouts when you are switching between apps or between home screens ..

    interesting is, that when i increased latency to 2048, result is almost same. When i add 10th instance, occasional dropouts starts ..

  • dendy:
    18 Apparilos you say?!:)
    i'm stunned.
    my iPad Air 1
    barely takes 2...

  • edited April 5

    @dendy said:
    @rs2000

    at buffer 512 it handles 9 instances of Lexington - default patch just tweaked to 8 voices unison - smooth even on background. 10 instances - until you are inside app, no dropouts, if you put app on background there are occasional dropouts when you are switching between apps or between home screens ..

    interesting is, that when i increased latency to 2048, result is almost same. When i add 10th instance, occasional dropouts starts ..

    Thanks!
    Just did the same test on the iPad 2018: 8x unison, default patch with longer release time
    8 instances work perfectly,
    9 instances and I get heavy dropouts.

    Update: To safely avoid stuttering while switching between apps, I have to reduce down to 6 instances of Lexington.

  • @dendy said:

    interesting is, that when i increased latency to 2048, result is almost same. When i add 10th instance, occasional dropouts starts ..

    You can get diminishing returns with super larger buffers unless the app in question is programmed to take advantage of those large buffers. It's exactly the same with desktop plugins.

  • edited April 5

    @jonmoore said:

    @dendy said:

    interesting is, that when i increased latency to 2048, result is almost same. When i add 10th instance, occasional dropouts starts ..

    You can get diminishing returns with super larger buffers unless the app in question is programmed to take advantage of those large buffers. It's exactly the same with desktop plugins.

    Hey, indeed, when reducing the buffer to 512 I can play 7 instances without stuttering while switching apps :+1:

  • for me, 15+ years of experience, best buffer size is 512 (or 1024)
    more or less equals trouble all the time...

  • @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

  • So interesting! I use a 128 buffer which explains why the new mini did not work out nearly as well for me. I’m a live player more often than a sequencer and anything more than 256 creates an unworkable lag for me for live playing. I’m really glad that Apple kept the mini alive even though it didn’t work out for me personally.

  • @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

  • @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    Thanks, looks like the search is finally over :smile:
    Need to decide on the size...

  • @dendy said:
    at buffer 512 it handles 9 instances of Lexington - default patch just tweaked to 8 voices unison - smooth even on background. 10 instances - until you are inside app, no dropouts, if you put app on background there are occasional dropouts when you are switching between apps or between home screens ..

    interesting is, that when i increased latency to 2048, result is almost same. When i add 10th instance, occasional dropouts starts ..

    Thanks. Also curious to know how the cpu heat/battery drain is when playing for 10 min or so.
    Good gadget to test with is Lisbon.

  • I noticed Crudebyte updated iSymphonic and Oriental Strings today to fix several issues with the new mini and iPad Air 3 models. They don’t state what the issues were ? Wondering if any other dev having issues with these models ?

    The performance sounds good though.

  • edited April 5

    Does it work properly at 128 buffers with moderate projects? For example, I can use my iPhone 7 Plus at 64 buffers in AUM with 3 Loopy slots routed individually to AUM, each one with an AU effect, and a mic lane with 5 AU effects and an Enso instance without any audio crackles except while switching between apps, so it’s usable live with pedalboard perfectly. Effects are Perforator, Replicant 2, DDMF 6144/NYC, VoiceRack FX, Blamsoft reverb, Dubstation2 (or Bluemangoo stereo delay, so DS2 workaround is not involved).

  • I think I’m going to trade my 9.7 pro for the Air 3, which I believe has the same performance as the new mini (?). I’ve been nervous about the scare stories and huge prices of the new pros. I think this performance report has made my mind up. Thanks @dendy :)

  • @Janosax yeah it works properly also with lower buffers.. i didn't experie ced single one issue.. of course with smaller buffer you can run a little by less plugins but that is expected .. check for example my AUM test with Zeeons on 128 buffer...

  • @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    @dendy you are fucking joking right? Never will this work without cracklings dropouts at 128 with NINE Zeeon synth running this patch in unison. I have the pro Version and did the same setup. I fed the ARP with the arpeggio of stranger things at 1024 buffer. I was at approx. 85%. Changed the buffer to 512 and got cracklings. Changed it down to 128 and just played one note and I got cracklings-galore. So if you are right... then the new iPad mini is more capable of making music than the pro version of 2018? Is apple fucking kidding me? What is their problem?

  • @david_2017 said:

    @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    @dendy you are fucking joking right? Never will this work without cracklings dropouts at 128 with NINE Zeeon synth running this patch in unison. I have the pro Version and did the same setup. I fed the ARP with the arpeggio of stranger things at 1024 buffer. I was at approx. 85%. Changed the buffer to 512 and got cracklings. Changed it down to 128 and just played one note and I got cracklings-galore. So if you are right... then the new iPad mini is more capable of making music than the pro version of 2018? Is apple fucking kidding me? What is their problem?

    Is it possible that the smaller screen needs less juice, allowing for more resources for audio processing?

  • @Keenan said:

    @david_2017 said:

    @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    @dendy you are fucking joking right? Never will this work without cracklings dropouts at 128 with NINE Zeeon synth running this patch in unison. I have the pro Version and did the same setup. I fed the ARP with the arpeggio of stranger things at 1024 buffer. I was at approx. 85%. Changed the buffer to 512 and got cracklings. Changed it down to 128 and just played one note and I got cracklings-galore. So if you are right... then the new iPad mini is more capable of making music than the pro version of 2018? Is apple fucking kidding me? What is their problem?

    Is it possible that the smaller screen needs less juice, allowing for more resources for audio processing?

    Interesting thought. Cannot give a valid opinion on that, but could be. Here is a small cap I made (its a pain):

  • @david_2017 said:

    @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    @dendy you are fucking joking right? Never will this work without cracklings dropouts at 128 with NINE Zeeon synth running this patch in unison. I have the pro Version and did the same setup. I fed the ARP with the arpeggio of stranger things at 1024 buffer. I was at approx. 85%. Changed the buffer to 512 and got cracklings. Changed it down to 128 and just played one note and I got cracklings-galore. So if you are right... then the new iPad mini is more capable of making music than the pro version of 2018? Is apple fucking kidding me? What is their problem?

    Yeah, Apple better get their shit together on their 2018 "Pro" iPads.

    @dendy Good stuff. So in real-world usage, the Mini 5 should be able to load ~4x as many AUs than the Mini 2.

    Any chance you could load Phasemaker (or KQ MiniSynth) in AUM using a 64 frame buffer? Does it crackle/stutter? iOS 12.2 didn't fix it on the 2018 Pros, so now I'm curious if it fixed it on the 2019 Mini/Air. This would be crucial for playing live performances with a single synth. Thanks again.

  • edited April 5

    Haha this thread is kind of blowing my mind. I don’t know what the hell I could have done using the exact same hardware and software to experience the exact opposite results ¯_(ツ)_/¯

    Edit: How is D1 performing for you?

  • @david_2017 said:

    @dendy said:

    @recccp said:
    @dendy
    Thanks!!!
    Sounds good so far:)
    Can you test AUM as well please.
    With something in midrange like Zeeon in poly mode.

    AUM - buffer 128, 9 instances of Zeon with patch 3 (Legend Bass) - just switched from mono to unison. Feeded by Rozetta Arp - 90% cpu load, smooth playback, no dropouts ..

    @dendy you are fucking joking right? Never will this work without cracklings dropouts at 128 with NINE Zeeon synth running this patch in unison. I have the pro Version and did the same setup. I fed the ARP with the arpeggio of stranger things at 1024 buffer. I was at approx. 85%. Changed the buffer to 512 and got cracklings. Changed it down to 128 and just played one note and I got cracklings-galore. So if you are right... then the new iPad mini is more capable of making music than the pro version of 2018? Is apple fucking kidding me? What is their problem?

    Which device do you have? (11, 12.9)

  • edited April 6

    @ocelot 14 dixie instances at buffer 64, cpu load 77-80%, clean sound, no dropouts... if i add one more instance, rare dropouts appears

    @marmakin uhm, just single instance of D1 runs at 70% of CPU at buffer size 128.. at 64 it jumps to 113% CPU... this must be some bug in D1 i really do not believe it needs more DSP power than Model D ... @analog_matt ?

    @recccp @david_2017 Aaah.. I'm verry sorry guys, i revisited my Zeeon test and obviously i messed up used buffer size. 9 instances are running smoothly on buffer size 512, at 128 i can run just 6 instances of that patch

    other tests i did are ok, i rechecked them all again just to be sure that zeeon test was only one where i messed up buffer size..

  • edited April 6

    Noticed one thing in AUM but this makes sense to me.

    In case i'm feeding synths with any sequencer/notes generator (lets say Rozetta Arp or Bassline), CPU load is steady at some value.

    But in case that i just route build in keyboard to all synhs and then i hit keys - cpu load jums up for split second to almost double value and then jumps back.. usually it doesn't make any dropout, but in theory it can ? For example at buffer 128 with six Zeeons it jums for a while from 88% to around 160%

    I guess this is because iOS gives high priority to touch events - but i would expect that it leaves audio thread untouched and handles touch event by different core.. hard to say what is happening, just note that all my tests inside AUM are done by feeding synths with arp/bassline, not by playing at onscreen keyboard...

  • One more important thing i realised now - device sample rate. Sample rate is affecting CPU load too... in beta version of NS2 i see current device sample rate (no idea if any other available host on iOS displays this, Matt added this for us for testing purposes)

    Now, if i have headphones connected, i shows 44khz - but if i disconnect headphones it jumps to 48khz. This is happening since iPhone 6S. I guess, devices without headphone output are running all the time at 48khz.

    At 48khz CPU load with same buffer, same plugins is slightly bigger - it not much, based on my observations it's arounf 10-15% more in average.

  • edited April 6

    Oh @dendy just real quick... did you do the tests with headphones? This changes everything
    (Because then the new pro would probably be good too)

    @Ben I have the 11 inch model, but should be nonetheless more powerful than any other device in the line up without a “pro” 😂 oh Apple 🍏

Sign In or Register to comment.