WEBVTT

00:00:00.160 --> 00:00:04.680
when i first saw the pentium g4560

00:00:03.919 --> 00:00:08.960
at 64. earlier this year i thought

00:00:08.160 --> 00:00:15.440
gee the core i7 6850 is almost exactly 10

00:00:13.840 --> 00:00:20.320
times the price we should see if it's at all worth it

00:00:18.640 --> 00:00:24.960
but we kind of missed the boat on that one

00:00:22.080 --> 00:00:28.000
because the core i9 is here now so i mean

00:00:26.000 --> 00:00:31.840
who would even care about this comparison anymore anyway

00:00:32.079 --> 00:00:40.239
wait we'll just use its replacement the core i9 7820x

00:00:37.760 --> 00:00:54.000
and we'll just agree to disagree about how many times 64 goes into 600

00:00:54.000 --> 00:01:01.760
synergy allows you to share your mouse and keyboard between multiple computers

00:00:59.440 --> 00:01:05.920
at once check it out now at the link in the video description

00:01:03.680 --> 00:01:14.400
let's get this out of the way right now we are not expecting the pentium g4560

00:01:09.840 --> 00:01:15.760
to go toe-to-toe with the core i9 7820x

00:01:14.400 --> 00:01:19.280
that's stupid i mean for one the core i9 has turbo

00:01:18.400 --> 00:01:24.400
boost to say nothing of the six

00:01:21.840 --> 00:01:28.159
extra cores that it's packing under the hood

00:01:25.360 --> 00:01:32.159
well okay Linus then what is the point of this video

00:01:29.680 --> 00:01:36.640
i'm glad you asked first since one is roughly

00:01:34.320 --> 00:01:41.600
ten times the cost of the other we thought this would give us a good feel

00:01:38.560 --> 00:01:44.720
for how pricing scales with performance

00:01:41.600 --> 00:01:45.840
within Intel's product stack because

00:01:44.720 --> 00:01:51.520
i think obviously we're not going to be getting 10 times the performance with the core

00:01:49.759 --> 00:01:56.640
i9 right or will be

00:01:53.280 --> 00:01:58.880
no we've been through that that's stupid

00:01:56.640 --> 00:02:03.439
reason number two is that this will give us some interesting insight into

00:02:00.960 --> 00:02:08.440
bottlenecks and where they show up within the system for example will the

00:02:06.399 --> 00:02:14.239
rebalanced cache and avx-512 support of the core i9 help it

00:02:11.680 --> 00:02:17.520
will it hinder it will it depend on the workload

00:02:15.520 --> 00:02:20.959
how much will the extra cores help with real world tasks does it even really

00:02:19.680 --> 00:02:27.040
matter now i know i know

00:02:24.080 --> 00:02:32.000
raw performance isn't the whole story the core i9 has more than double the pci

00:02:29.760 --> 00:02:36.800
express lanes quad channel memory support blah blah blah

00:02:34.160 --> 00:02:41.040
and those features don't come cheap so it makes sense that you'd be paying a

00:02:38.560 --> 00:02:46.400
premium for both the CPU and the accompanying motherboard and because the

00:02:43.920 --> 00:02:50.239
core i9 demands a better cooling solution

00:02:47.360 --> 00:02:54.080
surprise you get to buy one of those too they even require more thermal compound

00:02:52.160 --> 00:02:58.239
thanks to their larger physical size i mean

00:02:55.360 --> 00:03:01.920
the pentium comes with a cooler and compound that's more than adequate for

00:03:00.080 --> 00:03:07.280
it right out of the box so then with all these differences how

00:03:04.400 --> 00:03:11.599
can we possibly compare the two easy we'll just dump a bunch of

00:03:09.280 --> 00:03:15.440
benchmarks into excel like we always do and wait really

00:03:13.440 --> 00:03:19.440
is that all i'm paying these guys for oh no there's more here actually oh cool

00:03:17.519 --> 00:03:25.040
okay so while we'll be running the two chips on otherwise identical hardware we

00:03:22.000 --> 00:03:28.400
are also going to run a second set of

00:03:25.040 --> 00:03:31.120
tests on the 7820x with it configured

00:03:28.400 --> 00:03:36.080
with only two cores enabled and clocked like a g4560

00:03:33.120 --> 00:03:40.560
so in the past this method of cutting down higher-end chips has generated

00:03:38.480 --> 00:03:45.760
pretty accurate simulations of lower-end ones but today the new cache setup and

00:03:43.280 --> 00:03:50.799
support for avx-512 might make things a little more interesting

00:03:47.840 --> 00:03:56.000
our gaming results well okay these ones aren't going to shock anybody our

00:03:52.959 --> 00:03:59.760
simulated 4560 does slightly better than

00:03:56.000 --> 00:04:02.879
the real thing and the stock 7820x blows

00:03:59.760 --> 00:04:04.319
the doors off of both in every test

00:04:02.879 --> 00:04:09.519
but while that difference is measurable our

00:04:06.560 --> 00:04:13.519
arch nemesis doctor point of diminishing returns

00:04:10.799 --> 00:04:19.040
makes a bold appearance here interestingly our simulated chip

00:04:16.000 --> 00:04:22.479
actually fell slightly but measurably

00:04:19.040 --> 00:04:25.040
short of the real one at 4k in cs go and

00:04:22.479 --> 00:04:29.440
in 3dmark which makes us wonder where else we might see Intel's new caching

00:04:26.800 --> 00:04:34.240
mechanism perform worse bottom line here if you thought that

00:04:31.440 --> 00:04:38.960
spending 10 times was gonna net you 10 times the performance in gaming

00:04:36.400 --> 00:04:42.560
uh i believe as the gamers say get wrecked

00:04:40.400 --> 00:04:48.400
productivity though was really interesting in applications like y

00:04:44.960 --> 00:04:50.639
cruncher that support avx 512 we saw a

00:04:48.400 --> 00:04:54.080
massive difference with the simulated g4560

00:04:52.240 --> 00:04:59.199
while in more traditional tasks like cinebench CPU mark and real bench the

00:04:56.960 --> 00:05:04.080
cash related performance boost is less exciting with real benches and coding

00:05:01.360 --> 00:05:08.000
tests revealing another situation where it may actually cause a slight

00:05:05.840 --> 00:05:12.000
performance regression it's not over for the simulated chip

00:05:09.680 --> 00:05:18.400
though because our real world blender test also sees a huge performance boost

00:05:15.520 --> 00:05:24.720
over the real 4560. in productivity applications the lack of

00:05:20.720 --> 00:05:28.960
avx and course hurts our value chip big

00:05:24.720 --> 00:05:32.320
time the 7820x still doesn't give you 10

00:05:28.960 --> 00:05:33.120
times the performance for its price

00:05:32.320 --> 00:05:37.680
but given that many of these workloads are

00:05:35.759 --> 00:05:42.560
business oriented ones time is money and even a 2x performance

00:05:40.479 --> 00:05:47.280
improvement or a more modest one could allow it to pay for itself over time so

00:05:45.360 --> 00:05:50.800
your mileage may vary power consumption is probably not going

00:05:48.880 --> 00:05:54.800
to surprise anybody so we'll skim through that but what might is the

00:05:53.120 --> 00:05:58.880
performance scaling using some simple math helpfully

00:05:56.479 --> 00:06:04.479
provided by alex we crunched the numbers and calculated a very theoretical

00:06:01.600 --> 00:06:10.080
ballpark for how pricing would look if it scaled linearly with performance

00:06:07.759 --> 00:06:14.479
so in that world from the perspective of the core i9 7820x

00:06:12.400 --> 00:06:20.720
the performance averaged price of the pentium would actually be roughly five

00:06:17.360 --> 00:06:23.280
times as expensive as it is today

00:06:20.720 --> 00:06:28.960
then looking at it the other way around the price of the core i9 7820x

00:06:26.240 --> 00:06:32.800
would work out to about a quarter of its current msrp of

00:06:31.160 --> 00:06:37.600
600. so what did we learn today well first

00:06:35.120 --> 00:06:41.600
and foremost that CPU pricing at least for team blue is more complicated than

00:06:40.639 --> 00:06:46.319
just how does it benchmark on a test bench

00:06:43.840 --> 00:06:50.319
and uh then uh you know what price tag should we slap on it and more

00:06:48.240 --> 00:06:56.000
importantly perhaps we've also confirmed that skylake x's new cache

00:06:53.120 --> 00:07:01.680
generally helps more than it hinders though it can hinder situationally

00:06:59.840 --> 00:07:06.639
there are a hundred reasons for wanting multiple computers at your desk maybe

00:07:04.080 --> 00:07:10.560
you're a mac person primarily but you need a non-virtualized Windows

00:07:08.560 --> 00:07:16.800
environment for some reason synergy solves the problem of having to

00:07:13.199 --> 00:07:20.160
have two keyboards and two mice for two

00:07:16.800 --> 00:07:21.759
or more separate computers you can share

00:07:20.160 --> 00:07:27.120
one set of peripherals so you'll no longer get confused and the whole thing

00:07:24.319 --> 00:07:30.479
is seamless you just drag the mouse from one screen to another we're talking pc

00:07:29.680 --> 00:07:36.160
mac and even Linux support they offer a

00:07:33.520 --> 00:07:40.240
basic and pro option for synergy with a one-time payment and features include

00:07:38.319 --> 00:07:44.400
advanced stuff like clipboard sharing between computers dragging and dropping

00:07:42.400 --> 00:07:50.240
files between computers the ability to set up hotkeys and more so use our link

00:07:47.599 --> 00:07:53.360
in the video description to get 50 off synergy today

00:07:52.080 --> 00:07:57.759
so thanks for watching guys if this video sucked you know what to do but if it's awesome get subscribed hit that

00:07:56.240 --> 00:08:01.360
like button or check out the link to where they'll buy the stuff we featured

00:07:59.360 --> 00:08:04.400
in the video description also down there is our merch store with cool shirts like

00:08:02.720 --> 00:08:07.919
this one and our community forum which you should totally join so you can talk

00:08:06.000 --> 00:08:11.120
shop with other tech enthusiasts from around the world