WEBVTT

00:00:00.240 --> 00:00:07.359
apple just announced what many of us have been anticipating for years now

00:00:04.640 --> 00:00:14.000
with equal measures dread and excitement they're ditching Intel yup the longest

00:00:10.719 --> 00:00:16.400
lived CPU partnership in mac history is

00:00:14.000 --> 00:00:21.359
done and apple has confirmed not only the development of their own ARM-based

00:00:18.560 --> 00:00:26.880
chips in-house but also their intention to complete this migration away from

00:00:23.600 --> 00:00:27.840
Intel systems over the next two years

00:00:26.880 --> 00:00:32.800
max and potentially even pcs may never be

00:00:30.720 --> 00:00:38.160
the same but me you can always count on me and my

00:00:35.120 --> 00:00:41.120
segways to sponsors like elgato elgato's

00:00:38.160 --> 00:00:46.079
new wave 1 and wave 3 usb broadcast microphones are here use their wave link

00:00:43.760 --> 00:00:49.280
app to control your wave mic up to seven other audio sources and create

00:00:47.600 --> 00:00:52.520
independent mixes check them out at the link below

00:00:59.680 --> 00:01:06.560
since the development of the a4 soc began in 2008 apple has been

00:01:04.159 --> 00:01:10.479
consistently designing their own silicon in-house thanks to ARM's generous

00:01:08.799 --> 00:01:15.280
licensing model that allows third parties to implement their instruction

00:01:12.400 --> 00:01:20.240
set in standard ways but without limitation in terms of core layouts

00:01:17.759 --> 00:01:26.000
this flexibility gave apple the ability to control the entire design of the

00:01:22.880 --> 00:01:28.320
iphone and since the iphone 4s they've

00:01:26.000 --> 00:01:31.680
consistently defended their fastest smartphone crown

00:01:29.920 --> 00:01:36.720
apple's chip design prowess has brought them to the point now where the current

00:01:33.439 --> 00:01:38.000
iphone 11 pro a phone can offer up

00:01:36.720 --> 00:01:42.960
single core performance that's competitive with their current macbooks

00:01:40.240 --> 00:01:47.520
in synthetic benchmarks and that's to say nothing of their other in-house

00:01:44.880 --> 00:01:50.479
silicon meanwhile it's no secret that Intel hasn't been performing up to

00:01:49.119 --> 00:01:55.360
expectations for years apple's efforts to build

00:01:53.040 --> 00:02:00.000
slimmer and lighter macs with longer battery life have been frustrated by the

00:01:57.920 --> 00:02:03.040
enormous heat that they've had to deal with

00:02:01.040 --> 00:02:08.399
or have outright refused to deal with now Intel's upcoming 10 nanometer tiger

00:02:05.439 --> 00:02:12.480
lake micro architecture with z graphics has been making headlines for its

00:02:09.920 --> 00:02:17.840
promising performance improvements but here's the thing if apple's promising to

00:02:15.120 --> 00:02:22.720
launch a laptop with their own silicon this calendar year they clearly got sick

00:02:20.319 --> 00:02:27.040
of waiting anywhere from 12 to 36 months ago and have been well past the point of

00:02:24.959 --> 00:02:31.680
no return for quite some time furthermore tiger lake's thermal

00:02:28.640 --> 00:02:33.760
performance is yet to be seen so

00:02:31.680 --> 00:02:38.480
with 12 years of experience designing cpus and having achieved a claimed 100x

00:02:37.040 --> 00:02:44.000
performance improvement since they started why buy cpus that might turn out

00:02:41.440 --> 00:02:47.519
to be flops from Intel or from AMD for that matter

00:02:44.959 --> 00:02:51.280
if you want it done right do it yourself and i don't even blame apple for seeing

00:02:49.200 --> 00:02:56.160
it that way i mean i love Android as much as anyone but look at how much

00:02:53.840 --> 00:03:00.640
runtime apple squeezes out of a comparatively small battery next to

00:02:58.480 --> 00:03:05.200
their Android competitors and this is all while offering a snappy user

00:03:02.560 --> 00:03:11.280
experience within many cases half the system memory it's this control and

00:03:08.000 --> 00:03:14.480
consistent execution that makes apple so

00:03:11.280 --> 00:03:16.800
darn profitable here's the thing there's

00:03:14.480 --> 00:03:22.959
always a trade-off between your upfront r d costs versus your ongoing costs

00:03:20.319 --> 00:03:27.599
you can allow someone else like google or qualcomm to handle your operating

00:03:25.360 --> 00:03:30.159
system and your CPU design that keeps your engineering staff requirements

00:03:29.280 --> 00:03:37.840
lower but you might need to make up for some sloppiness with a big fat battery and

00:03:34.720 --> 00:03:41.120
some extra RAM or if you're apple you

00:03:37.840 --> 00:03:44.080
spend up front optimizing that hardware

00:03:41.120 --> 00:03:48.560
perfectly so that every unit you ship is as inexpensive as possible

00:03:46.400 --> 00:03:53.760
this model works out great as long as you can sell enough that your units

00:03:51.040 --> 00:03:57.760
times cost savings is greater than that upfront r d investment

00:03:55.680 --> 00:04:02.239
we see the same thing in game consoles sony for example built a custom rad

00:04:00.080 --> 00:04:07.680
tools crack and hardware decompression chip for the ps5 that was a complicated

00:04:05.120 --> 00:04:14.720
undertaking but without it sony would need a bigger faster SSD or a much

00:04:11.280 --> 00:04:15.519
beefier CPU in every console unit they

00:04:14.720 --> 00:04:22.160
ship so that initial investment could turn out to be absolutely game-changing

00:04:20.239 --> 00:04:27.360
pardon the pun but while using apple silicon in a laptop

00:04:24.479 --> 00:04:31.360
might not be painful the transition could be

00:04:28.560 --> 00:04:35.840
going from power pc to Intel was a lot smoother than from motorola to powerpc

00:04:34.160 --> 00:04:40.160
but it was still fraught with compatibility and performance

00:04:37.759 --> 00:04:45.600
implications early on while universal binary eventually became standard until

00:04:42.720 --> 00:04:49.919
powerpc died out the rosetta translation layer's first incarnation didn't run

00:04:47.520 --> 00:04:55.040
especially well and apple prevented their pro suites like final cut and

00:04:52.560 --> 00:05:00.880
logic from running under it as a result this time around apple says rosetta can

00:04:57.759 --> 00:05:02.639
translate an application at install time

00:05:00.880 --> 00:05:07.280
rather than at run time potentially making things much faster overall only

00:05:04.960 --> 00:05:11.919
time can tell how that works out though as with the transition to Intel

00:05:09.120 --> 00:05:18.160
eventually all applications will be ARM only and that has some both frightening

00:05:14.960 --> 00:05:19.039
and exciting implications for consumers

00:05:18.160 --> 00:05:24.639
first ipad os apps can basically

00:05:21.360 --> 00:05:26.960
just work on a mac apple arcade and the

00:05:24.639 --> 00:05:32.960
new maps and messages apps are working examples but the possibilities go much

00:05:29.440 --> 00:05:36.240
much deeper because iOS and ipad os are

00:05:32.960 --> 00:05:38.400
based on macOS it's relatively easy for

00:05:36.240 --> 00:05:43.919
apple to unify the entire ecosystem now around their custom ARM

00:05:41.120 --> 00:05:49.759
architecture effectively erasing the distinction between what's a mac and

00:05:46.639 --> 00:05:51.840
what's an idevice for developers the

00:05:49.759 --> 00:05:56.320
ipad is even getting relative mouse input for gaming and desktop grade

00:05:54.000 --> 00:06:00.160
spotlight search now however this convergence isn't without danger

00:05:58.560 --> 00:06:05.280
there's nothing official but it seems that the app store is being emphasized

00:06:02.400 --> 00:06:09.840
far more right now than alternative ways of installing applications and that's a

00:06:07.520 --> 00:06:15.039
big problem for services with external monetization schemes now to be clear the

00:06:12.400 --> 00:06:20.639
app store offers a valuable service and apple should make money from it but the

00:06:17.919 --> 00:06:24.960
thing is if i already have something in place why should i as a developer be

00:06:23.440 --> 00:06:28.800
forced to use it least of all on a desktop operating

00:06:27.039 --> 00:06:32.720
system where there isn't really much precedent for this that's where things

00:06:30.479 --> 00:06:38.479
get frightening because it's not clear who apple is putting first right now i

00:06:35.360 --> 00:06:40.800
mean promising custom hardware solutions

00:06:38.479 --> 00:06:44.880
tailor-made to what their customers need might actually be worth some

00:06:42.240 --> 00:06:49.199
monopolistic behavior to some people but there's one small problem does apple

00:06:47.039 --> 00:06:53.520
even know what their customers want to do with their computers the last few

00:06:51.199 --> 00:06:57.600
years have been spotty at best when it comes to listening to both customer and

00:06:55.600 --> 00:07:03.120
developer feedback i mean who asked for the touch bar and then how did it take

00:07:00.160 --> 00:07:06.479
an unacceptable three years for the company with the world's eighth highest

00:07:04.960 --> 00:07:11.199
r d spending to get rid of a fundamentally broken keyboard design to

00:07:08.960 --> 00:07:15.039
be fair apple has demonstrated a willingness to change lately user

00:07:13.280 --> 00:07:19.120
upgradable storage on the mac pro for example that's a great step forward

00:07:17.360 --> 00:07:23.840
there's just a lot of questions still to be answered in how this whole thing is

00:07:21.199 --> 00:07:28.080
gonna play out like how did the ipad pro with magic keyboard and macbook differ

00:07:26.160 --> 00:07:32.479
now what if Intel gets their [ __ ] together will pcs completely overtake

00:07:30.960 --> 00:07:37.440
them like what happened during late game power pc or what if the opposite happens

00:07:35.520 --> 00:07:42.400
the computer industry has a long and proud tradition of doing something kind

00:07:39.919 --> 00:07:46.960
of half ass at first waiting for apple to do it properly and then finding a

00:07:44.319 --> 00:07:50.560
hundred slightly worse ways to copy that will Windows for ARM become more widely

00:07:48.800 --> 00:07:54.800
available what's going to happen to boot camp only time can answer these

00:07:52.560 --> 00:08:00.160
questions but in the meantime i do have some predictions for how things are

00:07:56.400 --> 00:08:02.400
going to play out one Intel has tried to

00:08:00.160 --> 00:08:07.759
make low idle power consumption and always on computing a thing for years

00:08:05.440 --> 00:08:13.440
with initiatives like smart connect but it's never actually worked very well

00:08:10.160 --> 00:08:16.000
meanwhile apple has this absolutely on

00:08:13.440 --> 00:08:19.280
lockdown with iOS you can leave your ipad untouched for literally weeks at a

00:08:18.479 --> 00:08:25.039
time pick it up and all your files notifications and settings are right

00:08:23.440 --> 00:08:30.160
there waiting for you and you probably still have 20 battery to boot this along

00:08:28.160 --> 00:08:34.800
with the instant on experience that we take for granted on phones is coming to

00:08:32.399 --> 00:08:38.560
mac which means that apple has a major selling point on their hands even for

00:08:36.640 --> 00:08:41.760
people like me who do basically everything in a browser anyway

00:08:40.719 --> 00:08:48.160
next this is going to completely fix the thermal throttling that we have

00:08:45.440 --> 00:08:54.160
criticized apple for mercilessly over the last few years now putting a TDP

00:08:51.120 --> 00:08:56.480
rating on apple's soc is a little bit

00:08:54.160 --> 00:09:02.560
challenging but a non-text performance data indicates that the a12z's vortex

00:08:59.279 --> 00:09:04.880
cores in the ipad pro can draw about

00:09:02.560 --> 00:09:10.000
four watts in a single threaded workload if we were to extrapolate that out to

00:09:06.720 --> 00:09:12.160
all four vortex cores which is unlikely

00:09:10.000 --> 00:09:17.760
to be possible on a tablet but very possible on a laptop with better cooling

00:09:14.240 --> 00:09:20.880
and power delivery that's about 16 watts

00:09:17.760 --> 00:09:22.959
fully unchained which curiously happens

00:09:20.880 --> 00:09:28.160
to be almost exactly the cooling capacity of the 2020 macbook air with

00:09:26.080 --> 00:09:33.040
its weird fanless heatsink which is why my conspiracy theory might actually not

00:09:30.800 --> 00:09:37.839
be that far off they could stick an ARM processor into that same cooling and

00:09:35.200 --> 00:09:42.560
chassis design and it'll run faster cooler and probably with better battery

00:09:40.480 --> 00:09:46.560
life thanks to its ability to use low power cores to handle background tasks

00:09:45.120 --> 00:09:51.200
now i actually doubt that they will change much else on the first go because

00:09:49.200 --> 00:09:55.680
the marketing practically writes itself you know same mac as before

00:09:53.839 --> 00:09:58.959
more apple than ever magic

00:09:57.200 --> 00:10:02.480
probably a little catchier than that but i think you guys catch my drift

00:10:00.640 --> 00:10:07.040
prediction number three if they want this to be a success on the desktop

00:10:04.560 --> 00:10:11.760
either apple's ARM cores need to get competitive with x86 cores in single

00:10:09.360 --> 00:10:17.200
threaded tasks and i mean heavy desktop ones or they need to augment them now

00:10:14.640 --> 00:10:21.519
desktops are definitely coming and based on these performance claims i think

00:10:19.600 --> 00:10:27.120
apple is taking the latter path here giving themselves two years to finally

00:10:24.399 --> 00:10:32.399
push developers to both properly take advantage of many core CPU designs

00:10:30.240 --> 00:10:38.959
and to properly take advantage of heterogeneous compute which is using a

00:10:35.040 --> 00:10:41.200
combination of CPU and GPU or neural

00:10:38.959 --> 00:10:44.959
processor for that matter to do conventional CPU work

00:10:43.040 --> 00:10:49.360
faster when it comes to performance though there's more to consider

00:10:46.800 --> 00:10:54.240
ARM given that it's a risk architecture is able to do more work in fewer clock

00:10:51.680 --> 00:10:58.959
cycles and with less power than a sisk architecture like x86 can that's an

00:10:56.720 --> 00:11:03.600
advantage that apple used to enjoy with early power pc macs outperforming their

00:11:01.360 --> 00:11:06.959
Intel counterparts at much lower clock speeds but there are some drawbacks to

00:11:05.839 --> 00:11:11.200
this approach while risk is more streamlined with no

00:11:09.120 --> 00:11:15.920
microcode for example it also lacks flexibility meaning that if your

00:11:12.880 --> 00:11:18.959
application uses an x86 instruction for

00:11:15.920 --> 00:11:21.040
which there is no ARM equivalent you as

00:11:18.959 --> 00:11:26.160
a developer are looking at either having to work around it manually or simulate

00:11:24.000 --> 00:11:30.399
it with many more cycles than you would otherwise need that's part of why

00:11:28.000 --> 00:11:35.600
microsoft's Windows rt didn't support x86 emulation when it launched what

00:11:32.880 --> 00:11:39.920
makes this different is that compared to powerpc or even Windows rt

00:11:38.079 --> 00:11:44.320
many developers are by now well accustomed to working with ARM thanks to

00:11:41.839 --> 00:11:47.519
mobile devices although there are still unknowns

00:11:45.360 --> 00:11:52.880
will adobe for example be able to quickly implement every feature in the

00:11:49.519 --> 00:11:54.480
ARM version of creative cloud i mean

00:11:52.880 --> 00:12:00.000
given they can't even get every feature working in the regular anyway will apple

00:11:56.880 --> 00:12:00.000
try to force them into the app

00:12:00.560 --> 00:12:08.000
i mean i can't imagine adobe giving up 30 of their subscription fees come on

00:12:05.600 --> 00:12:12.480
the bottom line anyway is that nothing quite like this has happened before

00:12:10.399 --> 00:12:17.360
we're watching history being made today and whether apple succeeds or fails one

00:12:15.680 --> 00:12:22.800
thing is clear and that's that the computer industry will never be quite

00:12:19.519 --> 00:12:24.399
the same unlike my segway to our sponsor

00:12:22.800 --> 00:12:28.480
which is the same as it was in the beginning of this video even drop.com is

00:12:26.959 --> 00:12:33.600
bringing you the mass drop and sennheiser pc 37x gaming headset it's

00:12:31.519 --> 00:12:38.160
angled drivers and open back design results in very natural sound they're

00:12:36.160 --> 00:12:42.880
super comfy the drivers come from the same family as their hd 598 and hd 600

00:12:40.959 --> 00:12:46.720
series headphones meaning you got great stereo imaging and locational accuracy

00:12:45.279 --> 00:12:50.320
and they come with a noise canceling microphone they've sold over 40 000 of

00:12:48.959 --> 00:12:54.959
these things so if there was a problem with them you'd know about it because people be talking about it everywhere

00:12:53.519 --> 00:12:58.880
they include a 2-year manufacturer warranty so go check them out at the

00:12:56.880 --> 00:13:02.399
link in the video description if you guys enjoyed this video actually

00:13:00.959 --> 00:13:06.880
sorry Anthony i'm going to hijack this i'm going to send them over to the

00:13:04.480 --> 00:13:10.880
subzero water cooling on the 2020 macbook air video because that's where

00:13:08.560 --> 00:13:16.880
we found out that that laptop is designed around seemingly an a12z

00:13:14.320 --> 00:13:16.880
processor