WEBVTT

00:00:00.000 --> 00:00:06.960
hyperthreading is a feature that's been on high-end Intel CPUs for a long time

00:00:04.920 --> 00:00:12.519
but has often been misunderstood by the community almost as a whole today we're

00:00:09.920 --> 00:00:17.080
going to check out how actually it works on a Windows level how applications are

00:00:15.080 --> 00:00:24.839
interacting with it and what happens when you put load on actual

00:00:27.400 --> 00:00:34.000
threads the master case 5 by Cooler Master gives you the freedom to truly

00:00:31.519 --> 00:00:37.239
make your midtower PC case your own with a variety of modular parts and

00:00:35.760 --> 00:00:40.760
accessories check out the link in the video description to learn more many

00:00:39.480 --> 00:00:47.600
people are vaguely aware that hyperthreading improves performance of certain applications by making one CPU

00:00:44.879 --> 00:00:51.879
core behave as two so it can split the workload up more effectively but how

00:00:49.680 --> 00:00:55.600
does it work exactly many people seem to assume that there's a difference between

00:00:53.960 --> 00:01:00.079
real cores and hyperthreaded cores I mean it makes sense right once you get

00:00:57.520 --> 00:01:03.640
into your system Windows is going to see eight if you have eight hyperthreaded

00:01:02.199 --> 00:01:07.560
cores and then some of them will probably be weaker than others is that

00:01:05.479 --> 00:01:12.400
really how it works I mean Intel's own specs can be a bit confusing core i7s

00:01:09.960 --> 00:01:16.360
usually have four and then in Brackets eight listed as their number of cores

00:01:14.520 --> 00:01:21.040
does that mean that there are four subordinate cores or something are are

00:01:18.759 --> 00:01:24.000
are are the four extra cores weaker to start answering things instead of just

00:01:22.520 --> 00:01:28.400
repeatedly asking questions for the entire video open up toce manager on any

00:01:26.560 --> 00:01:33.520
system that has an i7 in it or a hyperthreaded CPU you'll see that all of

00:01:31.320 --> 00:01:38.720
the cores in the current case of this system the

00:01:34.640 --> 00:01:41.159
4790k it displays eight CPUs all of

00:01:38.720 --> 00:01:46.000
which show as the same thing that's because there is eight logical cores in

00:01:43.799 --> 00:01:51.719
the system when you read Intel spec sheet the four is for physical cores the

00:01:49.399 --> 00:01:56.079
eight is for what those physical cores result in which is eight logical cores

00:01:54.240 --> 00:01:59.240
that doesn't mean you have 12 cores that doesn't mean you have eight cores

00:01:57.840 --> 00:02:03.439
physically that means you have four cores physically physically and eight

00:02:01.000 --> 00:02:08.520
cores logically different things they don't add together very different things

00:02:06.200 --> 00:02:12.680
so how did these cores actually behave to find out we use unraid to set up two

00:02:10.440 --> 00:02:17.280
Windows Virtual machines we signed each VM four logical cores out of eight total

00:02:15.800 --> 00:02:22.360
then we ran some benchmarks to see if Windows actually cares whe which logical

00:02:20.080 --> 00:02:25.720
core it's using and to see if there's any preference between them if one of

00:02:24.000 --> 00:02:30.080
them is actually better than the other one some people on Forum seem to be

00:02:28.120 --> 00:02:34.800
under the impression that half of them like 0 1 2 and three in terms of cores

00:02:32.239 --> 00:02:39.680
are real cores and half of them four five six and seven are just the the

00:02:37.440 --> 00:02:46.000
extra hyperthreads so we're going to see how that works results showed virtually

00:02:42.480 --> 00:02:49.080
no difference cin bench got between 62

00:02:46.000 --> 00:02:51.280
and 625 per VM when we ran them

00:02:49.080 --> 00:02:57.560
separately when we ran them both at the same time they both got exactly

00:02:55.040 --> 00:03:01.720
383 so it looks like there's a bit of a performance hit even though the VM

00:02:59.319 --> 00:03:06.159
completely Ely separates its logical cores but at the end of the day the

00:03:03.959 --> 00:03:12.120
performance was the same similar story on City skylines which is a fairly CPU

00:03:08.680 --> 00:03:13.920
bound game both got 48 FPS when running

00:03:12.120 --> 00:03:19.640
separately with the performance dropping to 35 and 31 we ran them at the same

00:03:17.319 --> 00:03:24.799
time so there was the difference but not a huge one and due to City skylines not

00:03:21.840 --> 00:03:28.040
being our most consistent Benchmark it might come down to just slightly

00:03:26.200 --> 00:03:32.360
differences within the game we looked at it much more deeply with s zip running

00:03:30.360 --> 00:03:38.480
it simultaneously on both VMS gave nearly identical results between 11k and

00:03:35.519 --> 00:03:43.640
12K mips but when we ran it on just one it jumped to 17K to or 18K mips starting

00:03:42.040 --> 00:03:49.200
one first and then starting the other one a while later while it was running

00:03:46.200 --> 00:03:51.239
allowed us to see this balancing in real

00:03:49.200 --> 00:03:54.879
time it was interesting to watch as the processes were running the one you

00:03:52.959 --> 00:03:58.280
started first would be running faster at the beginning but over time they would

00:03:56.760 --> 00:04:01.959
start to equal out and it didn't actually take that long these results

00:04:00.079 --> 00:04:07.239
seem to make sense as hyperthreading works by sending threads to whatever the

00:04:04.599 --> 00:04:12.159
physical core actually has free to process instructions it doesn't split

00:04:09.959 --> 00:04:17.239
the physical cores into two things and leave the other half unutilized so if

00:04:15.000 --> 00:04:20.519
you have just one thing running it will perform better than if you have two

00:04:18.919 --> 00:04:25.800
running what's actually happening is that Windows can schedule two different

00:04:23.040 --> 00:04:29.720
threads on one physical core at the same time if the physical core is working on

00:04:27.840 --> 00:04:35.560
the first task but still has some resources free it can go ahead and use

00:04:32.479 --> 00:04:38.680
its remaining resources to process the

00:04:35.560 --> 00:04:41.360
second task so you won't always have a

00:04:38.680 --> 00:04:46.400
clean doubling or halfing of performance if you're running One Thing versus two

00:04:43.800 --> 00:04:50.960
but you will see substantial performance differences if you're leveraging the

00:04:48.759 --> 00:04:55.120
hyperthreading feature on a physical core but there is no real distinction to

00:04:52.919 --> 00:04:59.560
draw between The Logical cores on a hyperthreading CPU they're all the same

00:04:57.560 --> 00:05:04.120
thing both physically and how a hyper threading aware OS like Windows 10 sees

00:05:01.840 --> 00:05:07.960
them so no it's not the same thing as having eight physical cores like at all

00:05:06.360 --> 00:05:11.320
but it can be a good way to make sure your CPU Cycles don't go to waste

00:05:09.800 --> 00:05:16.280
hopefully that was interesting I'm sure there will be more questions basically

00:05:13.440 --> 00:05:21.319
the gist of it is you don't have 8 cores but you have a much more effective way

00:05:18.440 --> 00:05:26.120
to utilize the four that you do have today we're highlighting the k7xx

00:05:23.560 --> 00:05:29.919
limited edition ruby red headphones from Mast drop of course Mast drop still has

00:05:27.800 --> 00:05:33.479
a bunch of other cool products you can check out in the link in the video

00:05:31.400 --> 00:05:37.720
description also and hopefully you guys know about Mast drop by now but if not

00:05:35.759 --> 00:05:41.759
the concept is actually pretty simple the more people commit to purchase a

00:05:39.720 --> 00:05:46.280
certain product the lower the price of that product actually goes the product

00:05:44.440 --> 00:05:51.400
we're showcasing today is the same spec-wise as the k7xx headphones that

00:05:49.319 --> 00:05:56.120
lonus reviewed last year you can check that video out up here the only real

00:05:54.000 --> 00:06:00.240
difference is that this run uses red accents on the ear cups and headband

00:05:58.520 --> 00:06:04.000
remember that this is a limited drop so if you want a pair you're going to have

00:06:01.759 --> 00:06:08.120
to act uh pretty fast these headphones were configured by mass drop and

00:06:05.479 --> 00:06:12.599
manufactured by AKG they're open backed headphones with large cushiony ear cups

00:06:10.400 --> 00:06:17.240
they feature a flat wire voice coil ver motion two-layer diaphragm genuine

00:06:14.919 --> 00:06:21.720
leather headband memory foam ear pads with vure covering a 2-year warranty

00:06:19.560 --> 00:06:27.120
covered by massdrop and you can order outside the US but a $25 fee will be

00:06:24.400 --> 00:06:31.240
assessed to all International orders so if you want to check them out and grab a

00:06:28.520 --> 00:06:33.199
pair of ruby red k7xx headphones head over to the link in the video

00:06:32.240 --> 00:06:37.080
description today anyways let me know what you guys

00:06:35.759 --> 00:06:42.240
thought about it comments down below don't forget to like it if you thought I did a good job dislike it if you thought

00:06:40.319 --> 00:06:45.840
I did do a bad job but don't forget to leave the constructive criticism down

00:06:43.840 --> 00:06:50.000
below I would like to know what I can do to be better also be sure to subscribe

00:06:48.800 --> 00:06:55.599
if you want to see more of our videos use our Amazon affiliate code to buy stuff cuz like that's cool I don't know

00:06:53.960 --> 00:06:58.280
that's pretty awesome also we have shirts not this one but they're in the

00:06:56.960 --> 00:07:03.560
link in the description down below you can check out some shirts that we have those are also pretty cool and we have a

00:07:01.160 --> 00:07:06.479
forum that's like awesome and has new software on it and looks super cool and

00:07:05.199 --> 00:07:11.039
like works better and stuff and we're still making tweaks to it there's definitely some things that can still be

00:07:08.960 --> 00:07:15.120
improved the store is a little a little broken it's my bad anyways check out

00:07:13.400 --> 00:07:21.720
this video there's more information up there about like cool computer stuff I

00:07:17.160 --> 00:07:24.840
haven't picked one yet uh but it'll be

00:07:21.720 --> 00:07:30.039
cool Daren put something cool up there

00:07:24.840 --> 00:07:30.039
feel free to put your own video okay bye