WEBVTT

00:00:00.400 --> 00:00:06.720
if we look back Intel's high-end desktop

00:00:03.600 --> 00:00:09.440
or hedt lineup has for the most part

00:00:06.720 --> 00:00:13.679
been pretty clearly segmented from their mainstream lineup it's enjoyed

00:00:11.440 --> 00:00:18.720
processors with higher core counts and larger caches and the motherboards have

00:00:16.160 --> 00:00:24.640
had more RAM slots taking advantage of the bandwidth and capacity benefits of

00:00:21.520 --> 00:00:27.199
hedt's beefier memory controllers

00:00:24.640 --> 00:00:31.199
and of course on each edt workstation users have been able to count on being

00:00:29.119 --> 00:00:35.680
able to install a greater number of high bandwidth pci express devices without

00:00:33.760 --> 00:00:38.960
running into bottlenecks but

00:00:36.800 --> 00:00:44.640
i would actually make the argument that in the current environment

00:00:40.879 --> 00:00:47.920
Intel has actually damaged if not mostly

00:00:44.640 --> 00:00:52.320
destroyed the value proposition of their

00:00:47.920 --> 00:00:54.960
own entire high-end product stack

00:00:52.320 --> 00:00:58.960
so what happened exactly i will tell you

00:00:56.719 --> 00:01:03.680
after this message from our sponsor thursday boots thursday boots is a

00:01:01.359 --> 00:01:07.920
bootstrapped startup haha that handcrafts boots using high quality

00:01:05.840 --> 00:01:10.640
materials and sells them at honest prices check them out at

00:01:09.119 --> 00:01:15.040
thursdayboots.com forward slash LTT

00:01:13.439 --> 00:01:21.360
contrary to what you might think you can wear them on casual fridays

00:01:26.240 --> 00:01:33.119
Intel had a really good thing going in the absence of any competition they were

00:01:30.479 --> 00:01:38.159
able to stick with quad-core processors for consumers for 10

00:01:35.439 --> 00:01:42.400
years making the argument that well mainstream workloads like gaming ah they

00:01:40.560 --> 00:01:46.479
don't need more cores and anyone who does need more course they've probably

00:01:44.079 --> 00:01:50.159
got real work to do and they can justify ponying up for hedt

00:01:48.799 --> 00:01:56.640
but then AMD happened getting eight true cores

00:01:53.600 --> 00:01:58.799
with ryzen at the beginning of 2017 was

00:01:56.640 --> 00:02:03.399
a shock for the industry and i don't blame you if you've forgotten that eight

00:02:00.880 --> 00:02:09.759
core rising started at 329 us dollars with performance that

00:02:06.640 --> 00:02:11.920
compared favorably to Intel's hedt

00:02:09.759 --> 00:02:17.840
processors where eight cores at that time was going to cost you over a grand

00:02:14.879 --> 00:02:22.000
now Intel responded to that threat and they met AMD's high-end threadripper

00:02:19.599 --> 00:02:26.560
lineup head-on by dramatically increasing the core counts of its atdt

00:02:24.480 --> 00:02:32.800
lineup from 10 in the previous generation all the way to 18 with their

00:02:29.840 --> 00:02:37.360
7000 series core i9 cpus but that wasn't enough

00:02:34.800 --> 00:02:41.599
while mainstream core i7 was still the clear leader in single threaded

00:02:39.200 --> 00:02:47.280
workloads for anyone who did anything other than just gaming on their machine

00:02:44.560 --> 00:02:50.959
consumer ryzen had a huge price advantage thanks in part to its

00:02:49.200 --> 00:02:55.840
affordable motherboards so Intel finally had to bump their

00:02:53.599 --> 00:03:01.760
consumer chips as well first came the six core core i7 8700k at

00:02:59.200 --> 00:03:07.599
the end of 2017. now it didn't quite bridge the gap in budget workstation

00:03:04.080 --> 00:03:10.319
performance with ryzen but it did reduce

00:03:07.599 --> 00:03:15.200
AMD's lead somewhat and thanks to its superior single threaded performance it

00:03:12.640 --> 00:03:19.680
kept Intel on top for gaming and some other key workloads then fast forward

00:03:17.599 --> 00:03:24.319
another nine months and we got the core i9 9900k

00:03:21.680 --> 00:03:29.519
the first Intel branded a core consumer CPU and the first CPU ever on Intel's

00:03:27.360 --> 00:03:36.319
mainstream platform with core i9 branding along with the core i7 9700k

00:03:33.840 --> 00:03:40.879
also eight cores but without smt or hyper threading technology

00:03:38.080 --> 00:03:45.360
and then with those cpus we've got the real reason that Intel has been

00:03:42.879 --> 00:03:49.680
sandbagging consumer core accounts for so long

00:03:47.360 --> 00:03:55.040
because the hedt lineup has been traditionally based on Intel's

00:03:51.680 --> 00:03:57.280
workstation and or server platform

00:03:55.040 --> 00:04:02.720
where by the nature of these markets tech actually tends to move a little

00:03:58.799 --> 00:04:05.120
slower it has tended to lag behind their

00:04:02.720 --> 00:04:09.760
consumer processors architecturally sometimes as much as by two generations

00:04:07.840 --> 00:04:14.879
so compounding this performance disadvantage is the fact that hedt

00:04:12.480 --> 00:04:20.479
processors don't hit such high clock speeds due to power or thermal

00:04:17.280 --> 00:04:22.880
constraints and that they don't have an

00:04:20.479 --> 00:04:27.440
onboard graphics processor which over the last five years in particular has

00:04:25.040 --> 00:04:31.759
come to act as a co-processor for certain workloads on the consumer chips

00:04:30.400 --> 00:04:37.040
so we're at a very interesting crossroads right now

00:04:33.680 --> 00:04:39.360
think about it when we benchmark cpus

00:04:37.040 --> 00:04:44.080
for our reviews or whatever we tend to go out looking for workloads that help

00:04:42.000 --> 00:04:49.919
us demonstrate the potential difference in performance from one chip to another

00:04:47.280 --> 00:04:55.520
but in the real world how many workstation tasks like even workstation

00:04:52.639 --> 00:05:00.320
tasks do you actually perform in the course of a workday that require more

00:04:57.919 --> 00:05:06.080
than eight cores and of those how many of them can't be

00:05:03.199 --> 00:05:10.240
GPU accelerated in some way that is the ace up the sleeve of

00:05:08.479 --> 00:05:14.639
consumer chips and we'll be demonstrating that using the platforms

00:05:12.080 --> 00:05:20.080
you're looking at so with consumer chips you now have up to 16 threads enough to

00:05:17.520 --> 00:05:24.479
handle h.264 encoding without breaking a sweat and the same goes for light

00:05:22.320 --> 00:05:29.360
rendering for your 3d modeling and cad applications along with other

00:05:26.400 --> 00:05:34.800
traditionally CPU intensive tasks and this is especially true if you have a

00:05:31.759 --> 00:05:38.080
GPU that can be used to accelerate them

00:05:34.800 --> 00:05:41.520
so just look at how little our high-end

00:05:38.080 --> 00:05:43.840
desktop cpus affect spec view perfure

00:05:41.520 --> 00:05:48.880
it's basically just run to run variants in most scenarios in fact what's really

00:05:46.479 --> 00:05:55.360
interesting here is that our mainstream processors enjoy a significant advantage

00:05:52.880 --> 00:06:01.759
in applications like solidworks which is a traditionally workstation workload

00:05:58.639 --> 00:06:03.440
thanks to their much higher clock speeds

00:06:01.759 --> 00:06:07.199
this is again apparent in the case of adobe premiere where as we've tested

00:06:06.080 --> 00:06:12.800
before more cores does matter but only to a

00:06:10.400 --> 00:06:17.840
point so here we've reached that happy medium where the thread count the

00:06:15.120 --> 00:06:23.520
superior per core performance and the integrated graphics of the core i9 9900k

00:06:21.280 --> 00:06:29.120
put it in a league of its own way out ahead of Intel's own hddt chips

00:06:26.880 --> 00:06:33.199
even though some of them have more cores i mean

00:06:30.000 --> 00:06:35.680
this is amazing when you recall again

00:06:33.199 --> 00:06:40.240
that just two years ago we were stuck with four cores on Intel's consumer

00:06:37.840 --> 00:06:45.919
platform and had to pay a huge premium to get six or eight let alone the 10

00:06:42.560 --> 00:06:49.120
core 6950x that actually cost more by

00:06:45.919 --> 00:06:51.280
itself than the entire mainstream test

00:06:49.120 --> 00:06:56.400
bench that we are running here because remember the difference in CPU price is

00:06:54.080 --> 00:07:01.199
just part of the story the price difference between the platforms

00:06:58.160 --> 00:07:03.440
themselves can also be significant so

00:07:01.199 --> 00:07:09.039
all it'll take now is for AMD to continue to press the advantage of their

00:07:06.080 --> 00:07:12.319
modular CPU design and push core counts even higher with zen 2

00:07:11.199 --> 00:07:16.800
and then assuming that Intel follows suit and we

00:07:14.560 --> 00:07:22.160
know by now that they will have to the likely result we think is going to

00:07:19.680 --> 00:07:28.400
be the contraction and eventual disappearance of the traditional hedt

00:07:25.599 --> 00:07:33.599
lineup from Intel like think about it for light workstation use honestly apple

00:07:31.520 --> 00:07:39.199
hit the nail on the head with the imac photographers haven't really needed

00:07:36.080 --> 00:07:40.240
powerful workstations for a very long

00:07:39.199 --> 00:07:45.919
time now NVIDIA production hedt has offered

00:07:43.440 --> 00:07:50.880
clear performance improvements even as recently as two to three years ago and

00:07:48.560 --> 00:07:55.680
has also leveraged the increased pci express bandwidth with expansion cards

00:07:53.280 --> 00:08:00.800
like red rocket accelerators but GPU compute has eroded the market

00:07:58.479 --> 00:08:04.400
for devices like that very significantly i mean how many expansion cards do you

00:08:02.639 --> 00:08:09.360
have in your system so we're not saying that chips like

00:08:06.800 --> 00:08:15.199
Threadripper and Intel's own high core count cpus won't continue to have a

00:08:12.000 --> 00:08:16.800
place in desktop workstations there are

00:08:15.199 --> 00:08:21.360
workloads for them we're just saying that the use cases for

00:08:18.800 --> 00:08:27.039
those chips are not very mainstream anymore and that hedt is the wrong

00:08:24.560 --> 00:08:34.320
product for those kinds of customers and the reason is ecc memory support

00:08:30.479 --> 00:08:36.320
ryzen supports ecc from the ground floor

00:08:34.320 --> 00:08:41.440
all the way up to threadripper2 which makes it perfect for an entry-level

00:08:38.320 --> 00:08:45.200
workstation that has a need for ecc

00:08:41.440 --> 00:08:48.399
by contrast Intel has desperately clung

00:08:45.200 --> 00:08:52.160
to the paradigm of removing ecc from its

00:08:48.399 --> 00:08:54.320
consumer and hedt processors to force

00:08:52.160 --> 00:09:00.560
anyone doing more mission critical work to spend still more on a xeon

00:08:57.440 --> 00:09:02.160
that obviously isn't going to last so

00:09:00.560 --> 00:09:05.839
the bottom line is this we were wrong when we did our review of

00:09:04.399 --> 00:09:11.440
the xeon w we said xeon w had no reason to exist

00:09:08.959 --> 00:09:17.519
with only ecc to differentiate it from hedt because the performance was the

00:09:13.440 --> 00:09:20.800
same but actually hedt has no reason to

00:09:17.519 --> 00:09:22.959
exist if it doesn't support ecc because

00:09:20.800 --> 00:09:28.720
it's getting eaten away at from both the bottom and the top by Intel's own 9000

00:09:26.160 --> 00:09:32.720
series consumer chips and AMD's threadripper so here's our new proposed

00:09:31.760 --> 00:09:37.600
lineup you continue to expand the consumer

00:09:34.880 --> 00:09:41.440
chips with more cores when possible but don't compromise single threaded

00:09:39.360 --> 00:09:46.000
performance that's still your key advantage in certain workloads like

00:09:43.360 --> 00:09:51.519
gaming and solidworks then once that's done you replace the high core count

00:09:48.240 --> 00:09:53.279
lineup wholesale with xeon w so that's

00:09:51.519 --> 00:09:57.360
the lower end single socket only workstation z online and while you're at

00:09:55.680 --> 00:10:00.800
that you get your head out of your butt when it comes to the pricing of those

00:09:58.800 --> 00:10:05.839
chips they should end up in line with the core i9s that they will replace

00:10:03.279 --> 00:10:10.800
that way you cut out one entire platform to support which makes life easier on

00:10:07.600 --> 00:10:13.360
marketing teams and board partners and

00:10:10.800 --> 00:10:17.440
consumers now i suspect Intel won't take my advice they do like making money

00:10:15.440 --> 00:10:21.120
after all but there may come a time when it's absolutely necessary just like

00:10:19.920 --> 00:10:27.120
they're eventually going to have to bring coffee lake refresh to their lga

00:10:23.360 --> 00:10:29.040
1151 xeons threatening even xeon w

00:10:27.120 --> 00:10:34.240
but that's a conversation for another day for now the bottom line is this let's

00:10:31.519 --> 00:10:38.880
give credit where credit is due AMD brought more course to the table and

00:10:36.320 --> 00:10:43.600
Intel responded in kind so there has never been a better time to build a

00:10:41.279 --> 00:10:48.880
value-oriented desktop machine that can do serious workstation work and if this

00:10:46.480 --> 00:10:52.640
is what the death of hedt as we know looks like then i am super okay with

00:10:51.600 --> 00:10:57.920
that tang is the mobile carrier that's focused on customer service and customer

00:10:55.920 --> 00:11:01.680
satisfaction when you call ting you don't speak to a robot you get put

00:10:59.839 --> 00:11:06.160
through directly to a person and it doesn't cost extra you pay only for what

00:11:04.399 --> 00:11:10.720
you use with the average ting bill coming in at just 23 a month per device

00:11:09.040 --> 00:11:15.279
and they've got lower mobile data rates than ever at just 10 a gig beyond the

00:11:13.200 --> 00:11:19.360
second gig and if you're stuck in a contract and you switch to ting they'll

00:11:17.120 --> 00:11:24.240
actually cover 25 of your cancellation fee up to 75 so head over to

00:11:21.920 --> 00:11:27.519
linus2018.ting.com we're going to have that linked below

00:11:25.680 --> 00:11:30.959
and try out their savings calculator you enter your last couple bills and how

00:11:29.040 --> 00:11:35.279
much you paid and find out how much you'd save on ting then when you sign up

00:11:33.360 --> 00:11:40.959
with our link you get 25 bucks in service credit or towards a new device

00:11:37.519 --> 00:11:43.600
sweet right linus2018.ting.com

00:11:40.959 --> 00:11:47.279
down there so thanks for watching guys if this video sucked you know what to do

00:11:45.440 --> 00:11:50.640
but if it was awesome get subscribed hit that like button or check out the link

00:11:48.800 --> 00:11:54.000
to where to buy the stuff we featured in the video description

00:11:52.640 --> 00:11:59.200
also down there is our merch store which has cool shirts like this one and our community forum which you should totally

00:11:56.720 --> 00:11:59.200
join