1
00:00:02,360 --> 00:00:09,559
I've always had a problem comparing Apple silicon to PC parts because gaming

2
00:00:07,040 --> 00:00:15,160
wasn't ready I've always wanted to know how Apple silicon compares to

3
00:00:11,559 --> 00:00:16,359
off-the-shelf PC gpus and CPUs and

4
00:00:15,160 --> 00:00:20,840
that's not something we've been able to test because Macs and PCs now use

5
00:00:18,840 --> 00:00:28,160
completely different architectures and there were no games that ran natively on

6
00:00:24,320 --> 00:00:30,039
both but that's changed check this out I

7
00:00:28,160 --> 00:00:35,200
am playing Boulders Gate 3 one of the hottest video games right now on a Mac

8
00:00:33,320 --> 00:00:39,600
and it's running natively on Apple silicon there's now a growing list of

9
00:00:37,399 --> 00:00:44,399
games that support both Apple's Graphics hardware and Metal Graphics apis without

10
00:00:42,200 --> 00:00:48,320
any sort of translation layer so with the help of LMG Labs we're going to

11
00:00:45,960 --> 00:00:53,600
finally compare Apple chips with PC graphics cards and see what gpus from

12
00:00:50,280 --> 00:00:56,359
NVIDIA and AMD compete with Apple's M

13
00:00:53,600 --> 00:01:00,000
series though not in this video that's part two because in part one we need to

14
00:00:57,920 --> 00:01:03,879
build the test benches that the gpus will be on and those test benches have

15
00:01:02,039 --> 00:01:12,320
CPUs so we're going to figure out what CPUs are the closest to Apple silicon

16
00:01:07,880 --> 00:01:12,320
and I need to ask Labs if they'll help

17
00:01:12,600 --> 00:01:18,720
me what did I think oh my god well it's going to be interesting it's going to be

18
00:01:16,720 --> 00:01:23,119
a learning experience we haven't tested Max really all that much so this is

19
00:01:21,079 --> 00:01:29,520
going to be the first time Labs really tackles Max Nicholas hey this is

20
00:01:27,240 --> 00:01:33,720
Nicholas Harris he's LTT lab software developer part of his job has been

21
00:01:31,680 --> 00:01:38,960
developing and automating the tests for PCs and all their parts here the first

22
00:01:37,159 --> 00:01:42,799
step is figuring out the tests and so Nicholas worked with test technician

23
00:01:40,479 --> 00:01:48,320
John Duran to figure out how to measure these two completely different computers

24
00:01:45,680 --> 00:01:53,119
our primary goal was to find tests that can natively work on either system to

25
00:01:50,840 --> 00:01:57,119
avoid the Rosetta layer because that's that's another variable that we want to

26
00:01:54,640 --> 00:02:01,600
isolate for but we also by isolating for the CPU that also limits us because we

27
00:01:59,119 --> 00:02:05,920
do have tests that test the whole system but we're not looking to test the memory

28
00:02:03,680 --> 00:02:11,440
and the SSD and and the graphics card yet but our current test suite for mark

29
00:02:09,160 --> 00:02:15,480
bench is very Windows focused so there wasn't really anything we could just

30
00:02:12,920 --> 00:02:19,440
reuse from that there is a test framework out there called fonics which

31
00:02:17,080 --> 00:02:24,920
has been there out there for a long time so we tried to find some that did

32
00:02:22,680 --> 00:02:30,800
compression stuff that did maybe some encoding things that were just pure

33
00:02:28,319 --> 00:02:35,120
computational they came up with seven tests and we'll get to them and their

34
00:02:32,840 --> 00:02:39,760
results but first there's a problem which I'll tell you about after award

35
00:02:36,640 --> 00:02:41,640
from this video sponsor Squarespace need

36
00:02:39,760 --> 00:02:45,560
to build a web page well Squarespace is your One-Stop shop they have a huge

37
00:02:44,000 --> 00:02:49,200
variety of templates for any blog portfolio or business and if you're a

38
00:02:47,560 --> 00:02:53,959
business their marketing tools make it easier than ever to track and build your

39
00:02:51,080 --> 00:02:58,159
internet presence plus with 24/7 support you can leave the worrying to someone

40
00:02:55,640 --> 00:03:02,879
else in fact we use it here at LMG get started today at squarespace.com /mc

41
00:03:00,319 --> 00:03:07,360
address and save 10% off your first purchase okay so I have to confess

42
00:03:04,720 --> 00:03:12,159
something I had this idea fantasy really that Labs would test a bunch of Macs and

43
00:03:09,760 --> 00:03:15,920
then compare it to a matrix of CPU data that would show us what desktop and

44
00:03:13,640 --> 00:03:21,440
laptop CPUs perfectly match their M series counterparts but we have to use

45
00:03:18,760 --> 00:03:24,760
new tests so we're starting from scratch while we would have loved to investigate

46
00:03:22,680 --> 00:03:29,040
every CPU it is wholly unreasonable to get Labs to test them all for instance

47
00:03:26,599 --> 00:03:33,920
here in logistics there are about 150 different CP available to test and that

48
00:03:31,760 --> 00:03:38,519
doesn't even include the laptop CPUs and all the shapes and sizes they're cooled

49
00:03:35,640 --> 00:03:43,239
in either so we have to make trade-offs and this is about graphics cards and so

50
00:03:41,040 --> 00:03:48,159
that's why we're only including desktop components if you're after gaming

51
00:03:45,920 --> 00:03:52,159
performance uh the question then becomes like okay cool we have three contenders

52
00:03:49,879 --> 00:03:57,319
right we have apple Intel and AMD Intel changes their socket all the time but

53
00:03:55,040 --> 00:04:02,000
AMD you can go back three generations on am4 on the same platform and so it's why

54
00:04:00,480 --> 00:04:05,640
we're going to be sticking with am4 chips in our conclusion though we did

55
00:04:04,079 --> 00:04:09,480
run the tests on a few Intel chips earlier in this

56
00:04:07,280 --> 00:04:14,000
project all right so let's go through the tests and the results the first we

57
00:04:11,760 --> 00:04:17,000
did is of course cine bench it's widely used in the tech media space and they

58
00:04:15,560 --> 00:04:22,479
just came out with an update for it though we did our 23 it includes both a

59
00:04:19,759 --> 00:04:27,000
multi-core and single core score we chose cine bench because it kind of

60
00:04:24,320 --> 00:04:31,800
chooses itself as it's the prolific go-to processor Benchmark and it's

61
00:04:30,000 --> 00:04:36,240
really good that it supports apple natively as well as as Windows looking

62
00:04:34,520 --> 00:04:40,360
at the single core results you can see how the newest chips rise to the top but

63
00:04:38,360 --> 00:04:45,840
once all the cores get involved you can see just how the 24 and the M2 Ultra

64
00:04:43,120 --> 00:04:49,639
push it to the top sticking with single core tests lab did a flack encode test

65
00:04:48,360 --> 00:04:54,440
where they encoded a bunch of copies of a 9-in nail song from wave to Flack we

66
00:04:52,800 --> 00:04:58,720
actually struggled to find single core tests because most most tests are all

67
00:04:56,560 --> 00:05:03,720
about loading the CPU and trying to you know how fast you compute this thing in

68
00:05:01,560 --> 00:05:08,080
this test and another you'll see apple silicon is so far ahead of the other

69
00:05:05,639 --> 00:05:11,919
chips and they're all grouped together that's why we're going to be weighing

70
00:05:09,080 --> 00:05:17,440
this test less when we figure out our matches the last single core tests are

71
00:05:14,479 --> 00:05:21,520
the XZ and lz4 compression test with both compressing an iunu image we

72
00:05:20,039 --> 00:05:26,400
actually tried like four different compression algorithms or compression

73
00:05:23,880 --> 00:05:29,720
tests but not all of them worked sometimes they worked on one but not the

74
00:05:27,840 --> 00:05:35,120
other even though they're advertised for platform so we did find that XZ and lz4

75
00:05:33,039 --> 00:05:39,400
we were able to compile for for both natively lz4 single core shakes have

76
00:05:37,600 --> 00:05:44,120
slightly differently from cinam bench with most of the ryzen 5000 is closer to

77
00:05:41,639 --> 00:05:50,560
the M3 generation but it appears that with the XZ compression test ryzen has a

78
00:05:46,440 --> 00:05:52,560
bit more strength than it does in cine

79
00:05:50,560 --> 00:05:56,840
bench what was it like to do all the testing Illuminating testing is pretty

80
00:05:55,039 --> 00:06:00,639
straightforward once you identify the tests and you come up with your test

81
00:05:58,080 --> 00:06:04,960
Suite the execution it's just while you do your testing and as now as results

82
00:06:02,960 --> 00:06:09,039
come in it's like putting your your puzzle together right as the pieces

83
00:06:06,960 --> 00:06:13,080
slowly fit in more you you get more of the picture however the difference is

84
00:06:10,639 --> 00:06:16,319
you're completing a puzzle that doesn't you don't know what the end picture is

85
00:06:14,520 --> 00:06:20,520
so it's interesting that way to kind of see see the story reveal itself to you

86
00:06:19,160 --> 00:06:25,800
all right so how about the puzzle pieces that tested multiple CPU cores blender

87
00:06:23,720 --> 00:06:30,160
is a popular 3D modeling program and in it Labs rendered the barbershop scene we

88
00:06:28,280 --> 00:06:33,720
might need to find a new scene as it can render fairly quickly it's a popular

89
00:06:31,800 --> 00:06:38,280
scene used to to Benchmark rendering performance in blender in blender the

90
00:06:36,080 --> 00:06:44,199
mid-range amds provide a transition between the M2 and M3s the 5600 xng are

91
00:06:41,680 --> 00:06:48,599
surprisingly weaker here another render type test that I'd never heard of is Cay

92
00:06:46,880 --> 00:06:54,960
it's a simple Ray Tracer that outputs this '90s looking image Cay gives us

93
00:06:51,800 --> 00:06:59,960
another uh render type of test but

94
00:06:54,960 --> 00:07:02,759
mostly we chose it cuz it works on both

95
00:06:59,960 --> 00:07:08,960
so depth we got for some of these but it's a very simple efficient load to uh

96
00:07:06,120 --> 00:07:14,280
multicore AMD is relatively weaker in this test with the 3600 sitting between

97
00:07:11,360 --> 00:07:18,520
the m1's and their different cooling the next test Labs did is lib raw which

98
00:07:16,280 --> 00:07:23,080
tests how well CPUs handle raw photographs lib raw is also nice that it

99
00:07:20,919 --> 00:07:27,759
has a built-in post-processing Benchmark which we run 30 times on the test image

100
00:07:25,919 --> 00:07:32,240
that comes with fronix and then it spits out like a a megapixels per second Libra

101
00:07:30,919 --> 00:07:37,919
is the other test we're going to have to weigh less because of Apple silicon's apparent Supremacy it does feel like the

102
00:07:35,879 --> 00:07:43,479
Macs are especially tailored to calculate audio and visual codecs lastly

103
00:07:40,840 --> 00:07:47,440
if you're into numbers there's Prim siiv we chose it because y cruncher doesn't

104
00:07:45,120 --> 00:07:51,960
work natively on Mac because we do favor y cruncher it's a very popular

105
00:07:48,759 --> 00:07:55,560
benchmarking one but uh we found Prim

106
00:07:51,960 --> 00:07:57,440
Prim prime prime CV Prime C it

107
00:07:55,560 --> 00:08:01,840
calculates prime numbers up to a certain length so we can considered that it was

108
00:08:00,159 --> 00:08:06,639
kind of a it's our standin for y cruncher and that it's something

109
00:08:03,319 --> 00:08:08,800
computational uh generating number over

110
00:08:06,639 --> 00:08:14,080
over a long period it's multicore as well we learned a lot by doing this and

111
00:08:11,520 --> 00:08:17,879
that is that this is hard for one picking am4 means that we've got an

112
00:08:15,680 --> 00:08:22,759
array of chips that don't quite fit with single core performance as that's where

113
00:08:19,680 --> 00:08:24,479
Apple silicon shines and these are old

114
00:08:22,759 --> 00:08:28,639
but then with am5 there aren't any low-end chips to compare with the

115
00:08:26,080 --> 00:08:32,399
lower-end Mac chips either single core is more important for gaming so we

116
00:08:30,080 --> 00:08:36,120
weighed higher but we weighted Flack and Libra less because they favor Apple

117
00:08:34,279 --> 00:08:41,800
silicon egregiously in a way that's not related to gaming all right our picks

118
00:08:39,519 --> 00:08:46,120
these choices for CPUs are still even a best guess because we don't know what's

119
00:08:43,640 --> 00:08:50,760
the bottleneck CPU wise once the gpus are installed and running games so these

120
00:08:48,600 --> 00:08:55,040
are not exact matches I really was in fantasy land thinking this was possible

121
00:08:53,480 --> 00:09:01,200
but we've learned a lot we're going to use the AMD 5800 x3d as a control and to

122
00:08:58,800 --> 00:09:05,560
match the M2 Ultra because its 3D cache really helps in gaming and the M2 Ultra

123
00:09:03,480 --> 00:09:09,880
screamed well ahead in every test we threw at it the M2 Pro and Max as well

124
00:09:08,040 --> 00:09:15,959
as the M3 Pro chips will be matched against the 5800 X the 5700x were

125
00:09:13,399 --> 00:09:21,240
pitting alongside the basic M3 and the basic M2 and M1 chips are matched

126
00:09:18,160 --> 00:09:24,560
against a 5600 G I'm feeling as well as

127
00:09:21,240 --> 00:09:26,959
I could the only way to feel better is

128
00:09:24,560 --> 00:09:31,079
just you're never done testing you could always test more right there's no like

129
00:09:29,240 --> 00:09:35,600
oh well I guess I'm done and then walk away with you know I solve you know

130
00:09:33,240 --> 00:09:40,120
everything and that I guess is the biggest lesson on this journey and it's

131
00:09:38,000 --> 00:09:43,880
that we're always learning but now that we have our CPUs figured out the next

132
00:09:41,760 --> 00:09:47,959
step will be to test the gpus have NVIDIA and AMD met their match we're

133
00:09:46,000 --> 00:09:52,600
going to have to see where they line up but things are looking good in their own

134
00:09:49,640 --> 00:09:57,040
little way personally I wasn't expecting the Mac the apples to be as strong as

135
00:09:55,600 --> 00:10:01,120
they they were I knew they were super efficient so this is my first like

136
00:09:58,640 --> 00:10:07,000
really EXP experience um now during this project I started daily driving the

137
00:10:04,079 --> 00:10:12,839
15-in MacBook Air and I loved it the fact that I could close it neglect it

138
00:10:09,200 --> 00:10:15,399
for 3 days and it still had power um I

139
00:10:12,839 --> 00:10:19,680
mean I ended up buying one what you bought a Mac from this project I mean I

140
00:10:17,839 --> 00:10:25,440
did immediately sticker bomb it how dare you but that's fine thanks for testing

141
00:10:22,720 --> 00:10:29,160
this Mac Address Labs uh if you want to check out another video we did check out

142
00:10:26,959 --> 00:10:33,240
the iPad tier list video and I'm curious in the comments below who of you are

143
00:10:30,800 --> 00:10:36,240
like Nicholas and bought a Mac for gaming