00:00wow it's not using any of the efficiency
00:07them that's crazy in the last two mac
00:10review videos I basically discovered
00:12that not all dolls can fully utilize all
00:14the cores in an apple silicon chip and
00:17as a result a newer chip doesn't
00:18necessarily give you better performance
00:21depending on which doll you use over
00:23here I have Apple's newest M3 Pro
00:27MacBook Pro and this time by popular
00:29request I'm going to test seven dos on
00:31it to see how well they utilize the
00:34cores in the chip then I'll compare to
00:36my 10 core M2 Pro Mac and my Tor M1 Pro
00:39Mac to show the performance differences
00:41across seven dolls my goal is to provide
00:43you with accurate information so you can
00:45make an informed decision on which Mac
00:48best suit your needs and so in the last
00:50two weeks I tested and retested all
00:53seven dolls on all three Macs multiple
00:56times to ensure that all the results
00:58were repeatable and not just oneof
01:00occurrences the M3 macro Pro I bought
01:02has the 11 core M3 Pro 18 gig of RAM and
01:05512 gig of SSD out of the 11 cores in
01:08the chip five of them are performance
01:10cores and six of them efficiency it's
01:12quite interesting that Apple has been
01:14replacing more and more performance
01:15cores with efficiency ones for the pro
01:18chips since the M2 line with the T core
01:21M1 Pro you get eight performance cores
01:24and two efficiency cores with the Tor M2
01:27Pro you get only six performance but
01:30four efficiency cores and now with the
01:3211 core M3 Pro you get six efficiency
01:35cores but only five performance cores
01:38this change in performance core to
01:40efficiency core ratio does make a
01:42noticeable difference in Real Life audio
01:45processing scenarios and that brings us
01:47to the testing to assess a doll's
01:49multi-core utilization I first put a
01:51mono guitar di on a track then I put an
01:54apple silicon native plugin archetype
01:56noly on the track set it to high
01:58oversampling so it would be more taxing
02:00on a CPU and then I duplicated the track
02:03with the plugin as many times as
02:05possible until I heard crackling in the
02:07audio or saw a CPU overload pop up when
02:10I try to play all the tracks uh back at
02:12the same time I would then delete one
02:14track hit play to see if the dog could
02:16play back smoothly if so I would note
02:19down the track count if not I would keep
02:21deleting tracks until the dog could play
02:23back smoothly and then note down the
02:25track count so these track counts
02:27represent the maximum CPU power can get
02:30with a specific doll on a specific chip
02:33with this specific plugin to take
02:34thermos into account I did run each test
02:37for about half an hour in a room that's
02:39about 23° C or 73° Fen before I wrote
02:44down the results all dolls were also set
02:46to 1024 buffer size which is suitable
02:49for mixing a mastering and the doll
02:51projects were all in 44.1 khz and set to
02:5484bit summing when the option was
02:57available now I want to explain my
02:58testing methodology a bit here so you
03:01understand why I did the test this way
03:02in this video and in the last two Mac
03:05videos and also why the test results are
03:07helpful in real life in the last two
03:10years I've been testing uh Apple silicon
03:12Max for audio processing I've learned
03:14that testing doll performance on these
03:17chips really comes Downs to multicorp
03:19utilization now obviously no one will
03:22duplicate the same track so many times
03:24in a real mixing session but you
03:26probably have many tracks and those
03:28tracks will probably have all kinds of
03:30plugins on them in a typical mixing
03:32session like that you are basically
03:34relying on the multi-core performance
03:35your doll can get out of the Chip And
03:38that's what I'm simulating here with my
03:40test setup it doesn't really matter that
03:42I'm duplicating and amping many times
03:44what matters here is that we can see how
03:46each doll uses the cores in the CPU and
03:49how much of a difference the M1 Pro M2
03:52Pro and M3 Pro chips have compared to
03:55each other let's start with Reaper and
03:57here's a screenshot of the audio
03:59configuration I used for the test I was
04:01able to play back 91 tracks with the Tor
04:04version of both the M1 Pro and M2 Pro
04:08before the audio started to crackle with
04:10the 11 core M3 Pro chip Reaper could
04:12handle 105 tracks which is 14 tracks
04:15more if we look at a CPU monitor here
04:17you can see that Reaper is able to
04:19utilize all the cores both performance
04:22and efficiency to the absolute limit
04:24with M3 Pro having one extra core Reaper
04:27was able to give you more performance as
04:29one would expect because of this I would
04:31consider Reaper to be fully optimized
04:33for all these chips and this means that
04:36if your main doll is Reaper you can
04:38expect more performance if you get a
04:40chip with more cores I want to remind
04:42you that the exact track down here means
04:44nothing because with a different plugin
04:46you could have way more or way less
04:48tracks the numbers are here to show the
04:51differences on how well these seven
04:53dolls utilize each CPU if you enjoy the
04:57video so far let me know by leaving a
04:59like and or leaving a comment uh it
05:02would really help with the channel
05:03moving on to qbase 13 and these are the
05:06audio configurations I had for the test
05:08qbas 13 could handle 89 tracks with M1
05:11Pro and M2 Pro and with M3 Pro it could
05:14handle 114 tracks which is 15 tracks
05:17more looking at the CPU monitor here you
05:19can see that the story is the same as
05:21Reaper qbas 13 was able to fully utilize
05:24all the cores though not as to the Limit
05:27as Reaper did but still the results are
05:29very similar I suspect it takes a bit
05:32more resources to run qbase 13 simply
05:34because it has so many more features
05:36than Reaper and that might be why the
05:39track counts are just a little bit less
05:41than Reaper but you can see that um qbas
05:4413 is still fully optimized for Apple
05:46silicon chips next up is Protools
05:48ultimate and these are the audio
05:50configurations on M1 Pro Protools could
05:53handle 86 tracks but on M2 Pro only 71
05:56on the M3 Pro however it could only
05:59handle 64 tracks which is 22 less tracks
06:02compared to Pro Tools running on M1 Pro
06:05I'm going to duplicate this track one
06:09time and this is what
06:14happens can't even play it so why does
06:17the third generation chip with more
06:19course perform worse than the first
06:21generation chip well the answer is in
06:23the CPU monitor you can see that on all
06:26three chips Pro Tools only used the uh
06:29performance course fully and the
06:30efficiency course remained idle despite
06:33M3 having one extra core in the than the
06:36M1 Pro if you look that only the
06:38performance scores M1 Pro actually has
06:41three more performance cores than the M3
06:43Pro and this is why pro tools on the 10
06:46core M1 Pro outperforms it being on the
06:4911 core M3 Pro and also why Pro Tools on
06:53M3 Pro sits in the middle since the T
06:55core M2 pro has one more performance
06:58core than the M3 Pro but two less than
07:02the M1 Pro it feels like I'm doing
07:03something wrong here but I double and
07:05triple checked to ensure that any
07:07settings that could get Pro Tools to use
07:09the CPU better I had already enabled
07:11them and this is still the best I can
07:13get so I can only conclude that Protools
07:16simply does not make full use of the
07:18efficiency course let's now take a look
07:20at Logic Pro X Apple's own doll and here
07:23are the audio configurations many of you
07:25in my last videos told me to set the
07:27processing threads to the max in of
07:29automatic I already said in those videos
07:31or the comments that I already did that
07:34and once again I'm saying that here that
07:36I did that okay so on the M1 Pro logic
07:40could handle 79 tracks but on the M2 Pro
07:43only 67 tracks on the M3 Pro again it
07:46could handle even less tracks than the
07:48first and second generation chips at
07:50only 64 tracks looking at the CBU
07:53monitor on three computers you can see
07:55that logic only fully utilized the
07:57performance course moving on to ablon
07:59live 11 I know that live 12 is out in
08:02beta but I don't have access to it so 11
08:05is what I tested here's the audio
08:07configurations I had for the test on M1
08:09pro life 11 could handle 72 tracks but
08:12on the M2 Pro only 62 tracks and on an
08:16M3 Pro even less at 60 tracks some of
08:19you pointed out that Ableton has a web
08:21page saying that for Apple silicon chips
08:23you should use smaller buffer sizes for
08:25lower CPU usage and larger buffer sizes
08:28for higher CPU usage however in my
08:30experience live 11 behaves exactly the
08:33same as other dolls when it comes to
08:35buffer size latency and CPU usage which
08:38is that if you set the buffer size to
08:41larger you'll be able to do more audio
08:43processing at the expense of higher
08:46latency if you set the buffer size to
08:49smaller you would get lower latency at
08:51the expense of not being able to do as
08:54much audio processing so right now I
08:56have the buffer size at 1024 which is a
08:59pretty large buffer it will be able to
09:06tracks now if I change it to a small
09:12128 and I'm going to try to play the
09:19thing yeah it's even worse I'm more
09:22inclined to believe that Ableton did not
09:24make a mistake in the article I think
09:26the wording is just a bit confusing I'm
09:29honestly not sure maybe I did something
09:31completely wrong so please correct me in
09:33the comments if I did taking a look at
09:35CPU monitors you can see that once again
09:37only the performance cores were being
09:39fully used by live 11 on all three Macs
09:43next up is the new doll FL Studio I
09:45think some of you wrote to image line
09:47about FL Studio not being included in my
09:49previous Mac testing videos and the
09:52company was kind enough to send me a
09:54license to test it they set absolutely
09:56no strings uh attached they did not not
09:59even ask me to include it in this video
10:01but of course I read your comments and
10:03included it of my own valtion but the
10:06fact that I got a license for free did
10:07not impact my testing in any way I'm I'm
10:10just here reporting hard data here are
10:13the audio configurations I had for the
10:15test on both M1 Pro and M2 Pro FL Studio
10:18could only handle 61 tracks while on the
10:21M3 Pro it could handle 70 tracks this is
10:25an interesting case because if you look
10:27at the CPU monitors all the cores were
10:30being used almost to the max
10:32theoretically FL Studio should be able
10:34to give us performance levels similar to
10:37uh those from Reaper and qbas 13 but in
10:41reality it's far from it now I did find
10:44FL Studios workflow to be rather
10:46different than other dolls so just to
10:49make sure I didn't make a a huge mistake
10:51I'll quickly show you how I set up the
10:53test I first imported the Modo di file
10:56into a track then I assigned it to
10:58insert one in track mode in a mixer I
11:01added archetype nly as an insert then I
11:04right click on the track and use the
11:06clog option to duplicate the many copies
11:09of the track unless I did something
11:10wrong here I don't know why FL Studio is
11:12underperforming compared to Reaper and
11:14qas despite being able to use all the
11:17cores I thought maybe it was uh because
11:20of all the fancy animations in FL Studio
11:23so I tried setting animation to minimum
11:25and that didn't help so I really don't
11:28know but what we do know from the test
11:30result is that having more cores be they
11:33the performance type or efficiency type
11:36does give you a leg up in FL Studio the
11:39last doll I tested was Studio 16 and
11:41here are the audio configurations I had
11:43you may notice that I had Dropout
11:45production set to high instead of
11:47Maximum the reason is to keep the
11:49process block size the same as other
11:51dolls which is 1024 samples if I set
11:53Dropout production to maximum the
11:56process block size would become 20 48
11:59and that wouldn't make a fair comparison
12:02on M1 Pro Studio One could handle 70
12:04tracks but on M2 Pro only 66 tracks on
12:08M3 Pro is 65 tracks looking at the CBU
12:11monitors you can see that Studio One is
12:14similar to Pro Tools logic and live 11
12:16in that it doesn't fully utilize the
12:18efficiency cores so the chip with more
12:21performance cores M1 Pro in this case
12:29all right I'm going to duplicate it one
12:31more one more time here we go I'm going
12:41oh yeah definitely cannot handle that I
12:44did notice that something weird is
12:46happening with Studio 1 on the M1 Pro
12:48Mac though it's not using one of the
12:50performance cores for some reason I'm
12:52assuming it must be some kind of bug so
12:55where does that leave us well if your
12:57main doll is Reaper C Bas or FL Studio
13:00the number of cores a chip has will
13:02directly translate to how much
13:04performance you get it actually doesn't
13:06even matter the number of performance
13:08cores versus efficiency cores because
13:10even though the Tang core M2 Pro traded
13:13to Performance cores to be efficiency
13:15cores it offers the same performance as
13:18the Tang core M1 Pro this means that if
13:21the kind of work you do requires mostly
13:23audio processing like mixing and you do
13:25it in Reaper qbase or FL Studio a Mac
13:28with T core M1 Pro chip will give you
13:30the best value or if you think you need
13:33more cores and if I were to extrapolate
13:36from the testing data so far the 12 core
13:38M2 Pro should give you the same
13:40performance as the 12 core M3 Pro I'll
13:43emphasize that this is merely my
13:44extrapolation based on empirical data
13:47and I did not actually have the chance
13:49to test these 12 core chips personally
13:51if your main doll is Pro Tools logic
13:53prox ailon live 11 don't know about 12
13:56or Studio One it's important to keep in
13:58mind that your doll only fully utilizes
14:00the performance scores in an apple
14:02silicon chip at least as of this video's
14:04publishing date it would be a mistake to
14:06assume that the newest MacBook Pro with
14:09an 11 core M3 Pro would offer you more
14:12performance than a Mac with a 10 core M1
14:14Pro when choosing a Mac you should
14:16really pay attention to how many
14:17performance scores the chip has because
14:20those are what will be powering your
14:21doll here's some advice for you to
14:23consider I know we all want to buy the
14:26newest Macs with the newest ships I know
14:28some of you love that new matte black
14:30color but if you want to get the most of
14:32the money you spent or if you want to
14:35allocate more of your money for Ram or
14:37SSD the best Mac for you might be from
14:41the M1 line or the M2 line for example
14:43if you need a lot of RAM for your film
14:45scoring sound libraries it would make
14:47much more sense to get this M1 Max mcro
14:50Pro with 64 gig of RAM and 2 TB SSD from
14:53Apple's refurbished store instead of
14:56configuring 64 gig of RAM for a brand
14:59new M3 Max MacBook Pro is a difference
15:02$1,400 unless you really need that 16
15:05core M3 Max chip as well it would make
15:08no sense to spend that extra cash when
15:11all you need is a lot of ram my M1 and
15:13M2 Mac buying Guy video is more relevant
15:15than ever now that more and more Apple
15:17silicon Macs are hitting the secondhand
15:18Market you can see from this video that
15:20the M1 line can actually be even more
15:23powerful than the M3 line for audio
15:25production it's really a great time to
15:27upgrade your Mac or jump on the Apple
15:29slicking line honestly I have been using
15:31my M1 Max MacBook Pro since launch and
15:33I've yet to feel the need to upgrade it
15:36like at all I work with a lot of large
15:39mixing projects and most of the time I'm
15:41only using like 50% of this CPU it's
15:43Bonkers that's it for today if you find
15:45this video helpful at all please
15:47consider giving it a like it will really
15:49help out the channel don't forget to
15:51subscribe if you want more content like
15:53this and I will see you next time buying
15:56Guy video is more relevant than we