WEBVTT

00:00:00.280 --> 00:00:07.600
simultaneous multi-threading a feature that Intel calls hyperthreading has

00:00:04.720 --> 00:00:11.880
become a real mainstate on consumer CPUs hyper threading allows a single CPU core

00:00:09.639 --> 00:00:16.359
to process two threads at one time and while this isn't quite as powerful as

00:00:14.280 --> 00:00:20.000
just having two physical cores certainly improves performance in many

00:00:17.880 --> 00:00:24.519
applications becoming important enough since it was originally introduced in

00:00:21.600 --> 00:00:30.000
2002 to spread to a huge number of chips as far down as the lower-end core I3

00:00:27.160 --> 00:00:34.760
series but get this Intel is planning to ditch hyperthreading after 22 years with

00:00:32.800 --> 00:00:38.719
the launch of its lunar Lake and aerol Lake processors later this year why

00:00:37.120 --> 00:00:43.879
would they get rid of what seems like such a useful feature I mean do they not

00:00:41.200 --> 00:00:48.199
want to beat AMD a big part of this is the hybrid chip design that Intel

00:00:46.039 --> 00:00:52.640
started rolling out with its aler L CPUs in 2021 instead of having all the cores

00:00:50.480 --> 00:00:58.199
be basically identical team blue instead gave us P or performance cores and e or

00:00:56.280 --> 00:01:01.519
efficiency cores especially important for laptops where battery life is a

00:01:00.000 --> 00:01:06.080
Paramount concern and you don't want to waste energy putting processes through

00:01:03.719 --> 00:01:09.600
more power hungry cores if you just don't need the extra performance

00:01:07.560 --> 00:01:14.240
according to Intel current peores give you about 30% more average throughput

00:01:12.400 --> 00:01:18.400
when hyperthreading is enabled but at the cost of a 20% increase in power

00:01:16.759 --> 00:01:22.439
consumption and even if you're not running off a battery this can matter a

00:01:20.320 --> 00:01:25.759
great deal in places like data centers where you're trying to move as many

00:01:24.000 --> 00:01:30.880
threads as possible through your chips meaning power consumption becomes an

00:01:27.320 --> 00:01:33.040
important concern and a big cost so how

00:01:30.880 --> 00:01:37.040
is Intel trying to solve this power problem without losing that magic

00:01:35.439 --> 00:01:40.600
hyperthreading performance we'll give you the answer right after we thank

00:01:38.439 --> 00:01:44.280
delete me online privacy isn't just personal it's a family matter because of

00:01:42.320 --> 00:01:48.399
that delete me is now offering seamless protection for your entire family with

00:01:46.159 --> 00:01:51.960
their family plans with individual data sheets tailored to each member their

00:01:50.439 --> 00:01:56.399
privacy first design ensures personalized removal of personal

00:01:53.520 --> 00:02:01.200
information from online databases from kids to adults everyone stays safe from

00:01:58.360 --> 00:02:04.600
unwanted exposure and scams simplified management means peace of mind for all

00:02:03.119 --> 00:02:09.200
so checko delete me at the link below and Safeguard your family's digital

00:02:06.439 --> 00:02:13.879
world today although ecores obviously don't have the same processing muscle as

00:02:11.360 --> 00:02:18.360
peores Intel says it's much more efficient to run extra threads on those

00:02:15.920 --> 00:02:23.000
ecores instead of using hyperthreading on peores with the way the operating

00:02:20.480 --> 00:02:28.480
system schedules threads it doesn't invoke hyperthreading until all cores

00:02:25.920 --> 00:02:32.959
both P and D have been populated in order to maximize power efficiency Intel

00:02:30.920 --> 00:02:38.800
says that by foregoing hyperthreading and instead just having more ecores you

00:02:35.239 --> 00:02:41.480
get a 15% increase in IPC that's

00:02:38.800 --> 00:02:47.720
instructions per clock per unit of power and a 10% IPC increase per unit of die

00:02:45.159 --> 00:02:52.920
area hyperring is still better in terms of IPC per die area when you don't have

00:02:51.000 --> 00:02:57.239
those additional ecores but seeing as Intel seems committed to their hybrid

00:02:55.159 --> 00:03:02.480
chip design which also allows them to pack more cores onto their CPUs than in

00:02:59.920 --> 00:03:05.720
the past team blue seems confident that users won't particularly Miss

00:03:03.959 --> 00:03:10.440
hyperthreading remember that the feature was originally designed for a CPU that

00:03:07.959 --> 00:03:14.040
had just one core Intel's benchmarks back this idea up with lunar Lake

00:03:12.519 --> 00:03:17.959
beating the previous gen meteor Lake chips fairly comfortably despite

00:03:16.159 --> 00:03:21.040
consuming a similar amount of power of course we'll have to wait and see what

00:03:19.560 --> 00:03:24.799
independent benchmarks say once the chips actually hit the market but the

00:03:22.680 --> 00:03:29.080
focus on efficiency isn't surprising considering Intel is trying to stay

00:03:26.879 --> 00:03:33.080
ahead of the arm-based offerings from both apple and and Qualcomm which are

00:03:31.159 --> 00:03:37.319
touting power efficiency as a major selling point but if you really do end

00:03:35.360 --> 00:03:41.879
up missing hyperthreading it could be a great time to invest in a really high

00:03:39.159 --> 00:03:46.360
performance sewing machine AO I'm sorry I'm sorry everyone thanks

00:03:44.640 --> 00:03:52.040
for watching guys if you like this video check out this video on another type of

00:03:48.640 --> 00:03:53.760
CPU Intel killed called itanium comment

00:03:52.040 --> 00:03:57.239
with video suggestions and don't forget to subscribe and follow