1
00:00:00,080 --> 00:00:06,000
you know what your last pc upgrade complete waste of money

2
00:00:04,720 --> 00:00:09,519
why because it's not about nanometers it's

3
00:00:07,919 --> 00:00:12,719
not about gigahertz it's bottlenecks the bottlenecks are

4
00:00:11,679 --> 00:00:18,640
gonna eat them all they don't want you to know about this because the truth

5
00:00:14,920 --> 00:00:20,160
is every pc has a bottleneck

6
00:00:18,640 --> 00:00:25,359
more than one i mean you can upgrade and upgrade and

7
00:00:22,720 --> 00:00:27,920
upgrade and get nothing haven't you been reading between the

8
00:00:26,720 --> 00:00:35,040
lines the come on sheeple just look look

9
00:00:31,679 --> 00:00:36,880
look at our sponsor see sonic has got

10
00:00:35,040 --> 00:00:40,559
you covered no matter what kind of system you're building and what kind of

11
00:00:38,719 --> 00:00:45,120
power supply you need they've got up to 80 plus titanium efficiency and

12
00:00:43,120 --> 00:00:50,160
warranties that go for as long as 12 years go check them out at the link in

13
00:00:47,120 --> 00:00:50,160
the video description

14
00:00:57,520 --> 00:01:03,760
a lot of you out there are probably perfectly happy with your pc you built

15
00:01:01,440 --> 00:01:08,479
it and now you're gaming working or streaming away with minimal hiccups but

16
00:01:06,400 --> 00:01:11,680
what about those of you who crave unlimited

17
00:01:11,360 --> 00:01:19,360
well power an enthusiast it can be absolutely devastating to finish a rig only to

18
00:01:16,640 --> 00:01:24,560
realize that one key part is holding back the performance of all of the

19
00:01:21,680 --> 00:01:28,080
others if only you'd spent a little bit more

20
00:01:25,600 --> 00:01:31,840
or alternatively if only you'd saved a bunch of money on the rest of it or is

21
00:01:30,000 --> 00:01:36,479
it really that straightforward in our most thorough investigation into system

22
00:01:33,840 --> 00:01:41,759
bottlenecks to date we took three games three cpus and four graphics cards and

23
00:01:39,840 --> 00:01:46,720
put them on our test bench first up one of our favorite games for

24
00:01:44,000 --> 00:01:51,439
showcasing smooth scaling across both CPU and GPU shadow of the tomb raider at

25
00:01:50,079 --> 00:01:55,600
1080p well as it turns out when your low-end

26
00:01:53,680 --> 00:02:00,640
consumer graphics card can only get you about 17 FPS at high settings you're

27
00:01:58,560 --> 00:02:07,280
really not going to get a big boost from upgrading your CPU in our case the core

28
00:02:03,520 --> 00:02:12,319
i5 9600k actually performed worse than

29
00:02:07,280 --> 00:02:14,879
the core i3 8100 and then the 9900k

30
00:02:12,319 --> 00:02:20,560
hardly made any difference at all now upgrading to a 1050 ti doubles our

31
00:02:17,840 --> 00:02:27,599
vram from 2 gigs to 4 and gives us a faster GPU and yet we still don't see

32
00:02:24,239 --> 00:02:29,120
any major improvement from a better CPU

33
00:02:27,599 --> 00:02:34,480
as you can see from the next graph however the second we swap for something

34
00:02:31,520 --> 00:02:38,879
beefier like an rtx 2060 with six gigs of vram and more than double the cuda

35
00:02:36,400 --> 00:02:43,680
cores we're off to the races now there isn't a huge gap between our six core i5

36
00:02:41,519 --> 00:02:48,400
and our eight core i9 but the performance gains from moving to a CPU

37
00:02:45,519 --> 00:02:55,360
that's both significantly faster and has an extra two cores and two threads is a

38
00:02:51,200 --> 00:02:57,040
staggering 18 gain in terms of FPS

39
00:02:55,360 --> 00:03:03,200
what's even more interesting is that when we upgrade to an rtx 3070 we see

40
00:03:00,080 --> 00:03:04,560
almost zero gain on our poor little quad

41
00:03:03,200 --> 00:03:08,400
core i3 so maybe it actually doesn't matter if

42
00:03:06,640 --> 00:03:13,519
you can't get your hands on a 3070 because for a lot of you out there there

43
00:03:10,800 --> 00:03:17,920
won't be much of an improvement if any as you can see though with our higher

44
00:03:15,200 --> 00:03:21,519
end processors the gap does widen as the core i9 starts to leave the i5 in the

45
00:03:20,319 --> 00:03:25,360
dust now this is a perfect illustration of

46
00:03:23,760 --> 00:03:29,440
why reviewers configure their test benches the way they do we've been

47
00:03:27,440 --> 00:03:33,920
criticized many times over the years for using let's say an extreme edition

48
00:03:31,920 --> 00:03:38,000
processor in a review of a mid-range graphics card and it usually goes a

49
00:03:35,760 --> 00:03:43,519
little something like wtf are you guys doing nobody buys a 300 GPU and a

50
00:03:41,360 --> 00:03:50,000
thousand dollar CPU and that's a fair point sure but the thing is reviews live

51
00:03:46,480 --> 00:03:51,840
online indefinitely and a faster CPU

52
00:03:50,000 --> 00:03:55,200
will eventually come we might even review them get subscribed

53
00:03:54,000 --> 00:03:59,920
by the way and the thing is by alleviating these

54
00:03:57,519 --> 00:04:03,920
bottlenecks proactively using the best hardware we can get our hands on we get

55
00:04:02,080 --> 00:04:08,799
a much clearer picture of the true capabilities of that graphics card even

56
00:04:07,040 --> 00:04:12,319
if it might take a little while for a mid-range gaming tower to realize them

57
00:04:10,959 --> 00:04:16,799
one of the other things you've probably noticed is that reviewers tend to

58
00:04:14,000 --> 00:04:22,160
benchmark games at different resolutions depending on what's being tested CPU

59
00:04:19,519 --> 00:04:28,160
reviews tend to focus on 1080p numbers while GPU reviews can include

60
00:04:23,919 --> 00:04:30,000
performance data for 1440p 4k and beyond

61
00:04:28,160 --> 00:04:33,440
let's look at why on the same setup at 4k

62
00:04:32,560 --> 00:04:38,800
yeah look at this graph

63
00:04:35,680 --> 00:04:41,440
when both the 1030 and 1050 ti are asked

64
00:04:38,800 --> 00:04:46,240
to render such a high fidelity scene they end up so bogged down that any CPU

65
00:04:44,720 --> 00:04:51,919
can keep up with them even the 2060 hardly cares which CPU

66
00:04:49,040 --> 00:04:56,639
it's paired with and as for the 3070 while it certainly gains a few frames

67
00:04:54,160 --> 00:05:01,120
from higher end processing power really it isn't gaining anywhere close to what

68
00:04:58,800 --> 00:05:05,919
it was at 1080p so at the end of the day at either

69
00:05:03,520 --> 00:05:10,560
resolution you have to ask yourself is it really worth spending almost double

70
00:05:08,240 --> 00:05:14,400
the money on your CPU to eke out just a few more FPS is the bottleneck really

71
00:05:12,960 --> 00:05:19,680
all that bad well it can be some programs inherently

72
00:05:17,199 --> 00:05:25,360
load certain components differently take cs go cs go is a game

73
00:05:22,639 --> 00:05:30,240
cs goes game so it needs a good graphics card or

74
00:05:27,600 --> 00:05:34,320
whatever but the thing is it's an older game that was tuned for professional

75
00:05:32,000 --> 00:05:39,039
gamers to run it at very high frame rates on previous generation graphics

76
00:05:37,199 --> 00:05:44,000
cards so as you can see from our benchmarks if you were to read on the

77
00:05:41,199 --> 00:05:47,840
internet that cs go is so CPU bottlenecked and then run out and buy a

78
00:05:46,160 --> 00:05:52,160
shiny new CPU you might end up sorely disappointed

79
00:05:50,400 --> 00:05:56,880
that is until you also upgrade your graphics card this quad-core 3.6

80
00:05:55,120 --> 00:06:01,440
gigahertz i3 manages very solid frame rates with our

81
00:05:58,880 --> 00:06:07,280
2060 but then look at this stepping up to an i5 straight up doubles

82
00:06:04,560 --> 00:06:10,720
that performance that's right double don't you love it when your CPU does

83
00:06:08,720 --> 00:06:16,160
that show it some love by rocking a processor t-shirt or CPU pillow from

84
00:06:12,479 --> 00:06:18,319
ltdstore.com finally step up to a 30-70

85
00:06:16,160 --> 00:06:23,600
and spending even more money starts to look like an okay idea our 9900s lead

86
00:06:21,600 --> 00:06:28,319
gets even longer when it's paired with a powerhouse of a graphics card like this

87
00:06:25,600 --> 00:06:32,880
one gaining over 50 FPS while our core i5 lags behind with an increase that

88
00:06:30,800 --> 00:06:37,840
could be described as benchmark margin of error this illustrates why the suite

89
00:06:35,840 --> 00:06:43,520
of applications used to evaluate a product is so important we've taken flak

90
00:06:40,880 --> 00:06:47,840
again many times for including games that no one plays instead of just

91
00:06:46,240 --> 00:06:53,199
testing the most popular titles on twitch but the reason we do it that way

92
00:06:50,560 --> 00:06:58,960
is to demonstrate how different types of games might run on a new product the

93
00:06:55,840 --> 00:07:01,039
thing is we don't need the exact FPS of

94
00:06:58,960 --> 00:07:05,680
every single esports game to know that if it gets 250 FPS in dota 2 it's

95
00:07:04,080 --> 00:07:11,440
probably gonna run league of legends okay we also want to know how it holds

96
00:07:08,240 --> 00:07:13,520
up in a more visuals focused game right

97
00:07:11,440 --> 00:07:17,840
of course while our testing focused on gaming today bottlenecks occur in all

98
00:07:15,599 --> 00:07:21,759
types of computing processes an operating system for example can limit

99
00:07:19,840 --> 00:07:25,599
your file transfer speeds you take the same computer throw Linux on it and all

100
00:07:23,919 --> 00:07:30,080
of a sudden boom you're copying files faster your onboard network port might

101
00:07:27,680 --> 00:07:35,440
not fully utilize your isp's new 2 gigabit service or your CPU might be

102
00:07:32,720 --> 00:07:39,599
sitting there with absolutely nothing to do while you're opening up a program

103
00:07:37,599 --> 00:07:43,840
because your hard drive is struggling to deliver it any data that it needs to

104
00:07:41,599 --> 00:07:48,560
work thermals can also be a silent bottleneck taking a system that normally

105
00:07:45,759 --> 00:07:52,800
runs fine and then throttling it down under heavy workloads and then there's

106
00:07:50,479 --> 00:07:57,599
scientific computing that's an area where unlike gaming you might need

107
00:07:55,039 --> 00:08:02,240
hundreds of gigabytes or even terabytes of system memory in order to keep an

108
00:07:59,599 --> 00:08:07,919
extreme multi-core processor like an AMD threadripper or epic fed with data

109
00:08:05,680 --> 00:08:11,199
speaking of lots of course do you need more for gaming

110
00:08:09,199 --> 00:08:16,319
well the answer is yes but only to a point while modern games

111
00:08:14,400 --> 00:08:21,039
have been moving in that direction since the last generation of home consoles

112
00:08:18,000 --> 00:08:22,639
included eight core cpus there's just a

113
00:08:21,039 --> 00:08:28,240
limit to how much can be done in parallel in a game because the work of

114
00:08:25,199 --> 00:08:30,879
one core is so often dependent on work

115
00:08:28,240 --> 00:08:33,839
that another core hasn't yet completed so

116
00:08:31,759 --> 00:08:39,120
while it's notorious for performing best on a 16 core ryzen 9 5950x

117
00:08:37,039 --> 00:08:44,159
our ashes of the singularity escalation tests didn't actually see as much of an

118
00:08:41,279 --> 00:08:49,279
improvement in terms of FPS as you might think after we moved from a respectably

119
00:08:46,560 --> 00:08:52,959
powerful quad core so in many cases the performance gains that you see on a

120
00:08:50,880 --> 00:08:57,519
chart where the higher core count one outperforms the lower core count one

121
00:08:55,279 --> 00:09:02,399
really come down to the manufacturer tuning the turbos of the individual

122
00:09:00,000 --> 00:09:06,320
cores so that those higher end ones turbo up to higher speeds

123
00:09:04,480 --> 00:09:10,399
on AMD at least this is something you might be able to overcome on your own

124
00:09:08,240 --> 00:09:15,360
with basic software like one us mrs clock tuner verizon so there you have it

125
00:09:13,040 --> 00:09:19,920
the answer to the age-old question of what do i upgrade first CPU or graphics

126
00:09:18,320 --> 00:09:25,440
card is it depends that's because if you think

127
00:09:21,760 --> 00:09:27,680
about it every system has a bottleneck

128
00:09:25,440 --> 00:09:31,200
because if it didn't have a bottleneck the performance would just keep going to

129
00:09:29,519 --> 00:09:35,040
the moon and the crazy part is those bottlenecks

130
00:09:33,200 --> 00:09:39,200
shift around depending on what you're doing and how the rest of the system is

131
00:09:37,120 --> 00:09:43,360
configured that's where the title came from and that's one of the main reasons

132
00:09:41,279 --> 00:09:47,519
it's so important to look at multiple reviews when you're considering spending

133
00:09:44,959 --> 00:09:52,160
money on a gaming pc because no one can test every workload with every

134
00:09:50,080 --> 00:09:57,200
combination of hardware while also considering every upgrade path that may

135
00:09:54,800 --> 00:10:02,000
or may not be available to you but one thing you probably won't need to upgrade

136
00:09:59,519 --> 00:10:06,560
is your power supply seasonic power supplies are built with the future in

137
00:10:04,399 --> 00:10:12,560
mind they've got everything in their lineup up to 1300 watt 80 plus platinum

138
00:10:10,080 --> 00:10:17,360
power supplies they've got units that have 12 year freaking warranties so

139
00:10:15,760 --> 00:10:20,480
clearly there's a great degree of confidence in the product that it's

140
00:10:18,880 --> 00:10:25,680
going to last you like three to four upgrade cycles of the rest

141
00:10:23,279 --> 00:10:29,600
of their gear and they've also got a bunch of unique products too if you want

142
00:10:27,680 --> 00:10:33,040
a quiet build you can go for one of their fanless options if you need

143
00:10:31,360 --> 00:10:36,720
something for an all-white build like one that we just put together they've

144
00:10:34,640 --> 00:10:42,720
got you covered there too and the bottom line is man c sonic just makes good

145
00:10:40,399 --> 00:10:46,399
stuff so go check them out at the link in the video description thanks again

146
00:10:44,480 --> 00:10:50,320
see sonic for sponsoring this video and guys let us know what other kinds of

147
00:10:48,720 --> 00:10:54,440
bottlenecks would you like us to explore in the future leave a comment and then

148
00:10:52,720 --> 00:10:58,240
maybe take a look at some of the best 240hz gaming monitors that we've been

149
00:10:56,720 --> 00:11:03,120
looking at so you can make sure you're not missing out on all those FPS's your

150
00:11:00,959 --> 00:11:06,079
system's making after you upgrade the rest of your hardware