- Do you know what your last PC upgrade, complete waste of money, why? 'Cause it's not about nanometers,
it's not about Giga Hertz, it's bottlenecks, the bottlenecks
are going to eat them all. They don't want you to know
about this because the truth is every PC has a bottleneck, more than one. I mean, you could upgrade and upgrade and upgrade and get nothing. Haven't you been reading
between the lines? The cron sheeple, just look, look! Look at our sponsor! Seasonic has got you covered
no matter what kind of system you're building and what kind
of power supply you need. They've got up to 80
plus titanium efficiency and warranties that go
for as long as 12 years. Go check them out at the link
in the video description. (Electronic Dance Music) A lot of you out there are probably perfectly happy with your PC. You built it and now
you're gaming, working or streaming away with minimal hiccups. But what about those of you
who crave unlimited power? Well, for an enthusiast, it
can be absolutely devastating to finish a rig only to
realize that one key part is holding back the performance
of all of the others. If only you'd spent a little
bit more or alternatively if only you'd saved a bunch
of money on the rest of it, or is it really that straightforward? In our most thorough investigation
into system bottlenecks to date, we took three games, three CPU's and four graphics cards and
put them on our test bench. First up one of our favorite
games for showcasing smooth scaling across both CPU and GPU shadowed the Tomb Raider at 1080p. Well as it turns out, when
you're low end consumer graphics card can only get you about 17
FPS at high settings, you're really not going to get a big
boost from upgrading your CPU. In our case, the core i5-9,600k actually performed worse than the core i3-8,100 and then the 9,900k hardly
made any difference at all. Now upgrading to a 1050
Ti doubles our view RAM from two gigs to four and
gives us a faster GPU, and yet we still don't see any major
improvement from a better CPU. As you can see from the next
graph however, the second we swapped for something
beefier like an RTX 2060 with six gigs of I-RAM and more than double the CUDA cores,
we're off to the races. Now, there isn't a huge gap
between our six core i5 and our eight core i9 but the performance
gains for moving to a CPU that's both significantly faster
and has an extra two cores, and two threads is a staggering
18% gain in terms of FPS. What's even more interesting, is that when we upgrade to an RTX 3070, we see almost zero gain on
our poor little quad core i3. So maybe it actually doesn't
matter if you can't get your hands on a 3070, because
for a lot of you out there, there won't be much of
an improvement if any. As you can see though, with
our higher end processors, the gap does widen as the core i9 starts to leave the i5 in the dust. Now this is a perfect
illustration of why reviewers configure their test
benches the way they do. We've been criticized many
times over the years for using let's say an extreme edition processor in a review of a mid-range graphics card. And it usually goes a
little something like WTF are you guys doing? Nobody buys a $300 GPU and a $1000 CPU. And that's a fair point, sure
but the thing is reviews live online indefinitely and a
faster CPU will eventually come. We might even review them,
get subscribed by the way. And the thing is by
alleviating these bottlenecks proactively using the best
hardware we can get our hands on, we get a much clearer picture
of the true capabilities of that graphics card even if
it might take a little while for a mid-range gaming
tower to realize them. One of the other things you've
probably noticed is that reviewers tend to benchmark
games at different resolutions, depending on what's being tested. CPU reviews tend to focus on
1080p numbers while GPU reviews can include performance data
for 1440p, 4k and beyond. Let's look at why. On the same setup at 4k,
yeah look at this graph. When both the 1030 and
1050 Ti are asked to render such a high fidelity scene, they end up so bogged down that any
CPU can keep up with them. Even the 2060 hardly cares
which CPU it's paired with. And as for the 3070, while it
certainly gains a few frames from higher end processing
power, really it isn't gaining anywhere close
to what it was at 1080p. So at the end of the day at
either resolution you have to ask yourself, is it really
worth spending almost double the money on your CPU to
eke out just a few more FPS? Is a bottleneck really all that bad? Well it can be. Some programs inherently load certain components differently, take CSGO. CSGO is a game so it needs a
good graphics card or whatever. But the thing is, it's an
older game that was tuned for professional gameers to
run it at very high frame rates on previous generation graphics cards. So as you can see from our benchmarks, if you were to read on the
internet that CSGO is so CPU bottle-necked and then run
out and buy a shiny new CPU, you might end up sorely disappointed. That is until you also
upgrade your graphics card. This quad core 3.6 gigahertz
manage i3 manages very solid frame rates with our
2060, but then look at this stepping up to an i5 straight
up doubles that performance. That's right, double! Don't you love it when your CPU does that? Show it some love by
rocking a processor t-shirt or CPU pillow from lttstore.com. Finally, step up to a 3070, and spending even more money starts to
look like an okay idea. Our 9,900 lead gets even
longer when it's paired with a powerhouse of a graphics card
like this one, gaining over 50 FPS while our core i5
lags behind with an increase that could be described as
benchmark margin of error. This illustrates why the
suite of applications used to evaluate a
product is so important. We've taken flack, again
many times for including games that no one plays, instead of just testing the most popular titles on Twitch. But the reason we do it that
way is to demonstrate how different types of games
might run on a new product. The thing is we don't need
the exact FPS of every single E-sports game to know that if
it gets 250 FPS in Dota two, it's probably going to run
League of Legends okay. We also want to know how it holds up in a more visuals focused game, right? Of course, while our testing
focused on gaming today, bottlenecks occur in all
types of computing processes and operating system for example, can limit your file transfer speeds. You take the same
computer throw Linux on it and all of a sudden, boom,
you're copying files faster. Your onboard network port might
not fully utilize your ISP's new two gigabit service or
your CPU might be sitting there with absolutely nothing to
do while you're opening up a program, because your hard
drive is struggling to deliver it any data that it needs to work. Thermals can also be a silent bottleneck. Taking a system that normally runs fine and then throttling it
down under heavy workloads. And then there's scientific computing. That's an area where unlike
gaming, you might need hundreds of gigabytes or even
terabytes of system memory in order to keep an extreme
multi-core processor like an AMD Thread Ripper
or Epic fed with data. Speaking of lots of cores,
do you need more for gaming? Well, the answer is yes,
but only to a point. While modern games have been
moving in that direction since the last generation of
home consoles included eight core CPU's there's just a limit
to how much can be done in parallel in a game because the
work of one core is so often dependent on work that another
core hasn't yet completed. So while it's notorious for
performing best on a 16 core Ryzenine 5950 X our Ashes of
the Singularity: Escalation tests didn't actually see as
much of an improvement in terms of FPS, as you might think after we moved from a respectably
powerful quad core. So in many cases, the
performance gains that you see on a chart where the higher
core count one outperforms the lower core count one really
come down to the manufacturer, tuning the turbos of the
individual cores so that those higher end ones
turbo up to higher speeds. On AMD at least this is
something you might be able to overcome on your own with
basic software like Onesmus' Clock Tuner for Ryzen,
so there you have it. The answer to the age old
question of what do I upgrade first CPU or graphics card is, it depends. That's because if you think
about it every system has a bottleneck because if it
didn't have a bottleneck, the performance would just
keep going to the moon. And the crazy part is those
bottlenecks shift around depending on what you're doing and how the rest of the
system is configured. That's where the title came from. And that's one of the main
reasons it's so important to look at multiple reviews when you're
considering spending money on a gaming PC, 'cause no
one can test every workload with every combination of
hardware, while also considering every upgrade path that may or
may not be available to you. But one thing you probably won't need to upgrade is your power supply. Seasonic power supplies are
built with the future in mind. They've got everything in
their lineup up to 1300 watt 80 plus platinum power supplies. They've got units that have
12 year freaking warranties. So clearly there's a great
degree of confidence in the product that it's going to
last you like three to four upgrade cycles of the rest of their gear. And they've also got a bunch
of unique products, too. If you want a quiet build you can go for one of
their fanless options. If you need something for an
all white build like the one we just put together, they've
got you covered there too. And the bottom line is man,
Seasonic just makes good stuff. So go check them out at the
link in the video description. Thanks again, Seasonic
for sponsoring this video and guys let us know what other kinds of bottlenecks would you like
us to explore in the future? Leave a comment and then maybe
take a look at some of the best 240 Hertz gaming monitors
that we've been looking at. So you can make sure you're not missing out on all those FPSs
is your system's making after you upgrade the
rest of your hardware.
