WEBVTT

00:00:00.160 --> 00:00:05.200
Intel's been the industry's favorite punching bag for the last few years

00:00:03.679 --> 00:00:09.760
we've certainly taken our fair share of cheap shots but that changes today

00:00:07.600 --> 00:00:14.240
because chips like this with two different kinds of cores in them or more

00:00:12.240 --> 00:00:18.720
are probably the direction the entire industry is going to go that's right

00:00:16.400 --> 00:00:24.080
Intel jumped on the bandwagon with ARM and apple for their best generational

00:00:21.119 --> 00:00:28.000
upgrade in at least a decade but let's back up a sec

00:00:25.760 --> 00:00:32.399
this isn't going to be another furnace like past generations is it and if

00:00:30.400 --> 00:00:37.360
they're cheaping out juicing up their spec sheets with these

00:00:34.480 --> 00:00:42.800
low performance efficiency cores how could they possibly compare to AMD

00:00:40.079 --> 00:00:46.559
who ships only high performance cores our sponsor Corsair doesn't know thanks

00:00:45.120 --> 00:00:51.280
Corsair for sponsoring this video Corsair's xenion gaming monitor is an

00:00:48.640 --> 00:00:55.840
ultra slim 32 inch qhd gaming monitor with up to 165hz refresh rate one

00:00:53.760 --> 00:00:59.719
millisecond response times and more learn more at the link below

00:01:06.720 --> 00:01:12.880
12th gen core codename alder lake is a paradigm shift for desktop processors

00:01:11.280 --> 00:01:16.560
but it's also one that we've had in mobile for quite some time and the tl dr

00:01:15.520 --> 00:01:21.759
is this instead of stuffing as many full power

00:01:19.040 --> 00:01:26.240
cores into the CPU as possible then turboing them down when they're all

00:01:23.600 --> 00:01:32.400
active the cores are now split up between performance and efficiency cores

00:01:29.280 --> 00:01:35.439
called p and e-course respectively the

00:01:32.400 --> 00:01:37.280
launch lineup ranges from 10 to 16 total

00:01:35.439 --> 00:01:41.200
cores with the main differences between them being of course clock speed and

00:01:39.439 --> 00:01:45.280
that the performance cores are capable of running two threads at once via hyper

00:01:43.360 --> 00:01:49.280
threading there's more to the platform than that but the first question you

00:01:47.119 --> 00:01:54.799
guys have is how does all of this impact performance great question but we need

00:01:51.680 --> 00:01:56.720
to talk about Windows 11 first it's kind

00:01:54.799 --> 00:02:00.960
of a prerequisite for testing elder lake's radical new design because the

00:01:58.320 --> 00:02:05.119
thing is Windows 10 just doesn't have a scheduler that's capable of taking full

00:02:02.880 --> 00:02:10.160
advantage of this new architecture this is a huge problem because up until now

00:02:07.520 --> 00:02:13.680
Windows 11's new scheduler hasn't been kind to AMD

00:02:11.760 --> 00:02:17.760
thankfully the fixes for performance issues on ryzen came out just as we were

00:02:15.840 --> 00:02:21.520
starting our benchmarks so our numbers should be fair to both sides in spite of

00:02:20.080 --> 00:02:25.360
the fact that we're benchmarking on Windows 11. we do need to address

00:02:23.280 --> 00:02:31.360
another big variable here though alder lake supports both ddr4 and ddr5 which

00:02:29.200 --> 00:02:35.120
each have their own trade-offs we chose ddr5 for our launch review meaning that

00:02:33.280 --> 00:02:39.599
our numbers are going to represent a best-case scenario for performance

00:02:37.519 --> 00:02:45.280
but maybe won't represent a best-case scenario for value due to ddr5's higher

00:02:42.319 --> 00:02:50.239
cost f1 2021 starts things off with a bang for Intel with the core i9 12900k

00:02:47.760 --> 00:02:55.599
coming in at 5 to 10 faster than its ryzen nine competitors and the core i5

00:02:53.040 --> 00:03:00.800
just a hair's breadth away can Intel keep it up with forza oh yeah they can

00:02:58.720 --> 00:03:04.000
that is anywhere from eight to eleven percent faster depending on the

00:03:02.080 --> 00:03:08.800
parameter you're looking at and whatever advantage Intel has here probably clock

00:03:06.159 --> 00:03:15.040
speed it's paying off in far cry 6 too with an even bigger lead it is up to 25

00:03:12.800 --> 00:03:20.000
faster than the competition and never dipped below 100 frames per second and

00:03:18.480 --> 00:03:25.519
if you've been paying close attention to the graphs the core i5 has never been

00:03:22.959 --> 00:03:30.959
far behind making life really difficult for our previous value gaming darling

00:03:28.319 --> 00:03:34.959
the ryzen 5 5600x moving on to microsoft flight simulator

00:03:33.200 --> 00:03:41.200
well this is a bit of a mess isn't it the ryzen 9 5900x beats out the core i9

00:03:38.799 --> 00:03:46.640
with the higher turboing not to mention higher core count 5950x

00:03:43.760 --> 00:03:50.480
falling behind even the ryzen 5. given how lightly threaded flight simulator

00:03:48.400 --> 00:03:53.920
has been so far this might just be an example of the new Windows 11 scheduler

00:03:52.560 --> 00:03:58.560
running up against something that it didn't expect in spite of the fixes in

00:03:56.319 --> 00:04:03.439
fact we almost threw this test out due to the weirdness but we felt like

00:04:00.879 --> 00:04:07.840
whatever's going on this is one scenario where Intel's new silicon isn't as

00:04:05.360 --> 00:04:11.840
dominant as they'd like it to be and is possibly also an area of investigation

00:04:09.760 --> 00:04:16.239
for AMD and microsoft then you've got situations like cs csgo where Intel

00:04:13.920 --> 00:04:20.239
treats us to even more ludicrous frame rates in the dust 2 benchmark and

00:04:18.000 --> 00:04:26.160
civilization 6 where our turn times are a good 7 to 10 percent shorter than on

00:04:23.040 --> 00:04:28.400
team red which could make that one more

00:04:26.160 --> 00:04:32.400
turn before bed go by a little quicker overall our testing has firmly

00:04:30.320 --> 00:04:36.960
demonstrated that Intel has reclaimed its crown for gaming performance with

00:04:34.080 --> 00:04:41.680
the core i9 enjoying a solid 7 overall lead over ryzen 9 and the core i5

00:04:39.520 --> 00:04:46.080
pulling ahead even further compared to the ryzen 5. man

00:04:43.840 --> 00:04:51.040
this kind of competition is what i signed up for when ryzen first launched

00:04:48.479 --> 00:04:55.280
now Intel famously thinks that cinebench is not really a real world test but

00:04:53.759 --> 00:04:59.840
they've got cost to celebrate these numbers for a change cinebench tends to

00:04:57.680 --> 00:05:04.800
scale linearly with core count so the fact that a CPU that has only eight

00:05:02.320 --> 00:05:10.720
proper performance cores could dethrone the ryzen 9 is an achievement perhaps

00:05:07.840 --> 00:05:15.199
fueled in part by its ddr5 memory it also demonstrates that those

00:05:12.320 --> 00:05:18.479
efficiency cores ain't your grandma's efficiency course they clearly pack some

00:05:17.600 --> 00:05:25.600
punch while the firefox compile did take a little longer on the core i9 versus the

00:05:22.240 --> 00:05:28.320
5950x Intel still won out in the bursty

00:05:25.600 --> 00:05:31.919
bmw blender render although they lost ground on the higher endurance

00:05:29.759 --> 00:05:36.240
gooseberry test showing that for really serious multi-core work AMD is still the

00:05:34.639 --> 00:05:43.039
more compelling option and aside though here that core i5 be

00:05:38.800 --> 00:05:44.560
looking more like a 5800x than a 5600x

00:05:43.039 --> 00:05:48.400
keep that in mind for when we discuss pricing later

00:05:46.080 --> 00:05:53.360
now strangely corona benchmark seems to come back in favor of AMD's additional

00:05:50.720 --> 00:05:58.160
course which shows that 3d rendering may be becoming more than simply a raw how

00:05:55.919 --> 00:06:01.759
many cores do you have game puget bench for its part benefits from

00:06:00.000 --> 00:06:06.319
what Intel's got on tap with a healthy lead especially in photoshop over AMD

00:06:04.240 --> 00:06:11.840
although Intel falls short in spec workstation where the only consistent

00:06:08.319 --> 00:06:14.080
wins are against the 5900x check out the

00:06:11.840 --> 00:06:18.160
financial services and energy tests guys where the 5950x still lives in a league

00:06:16.800 --> 00:06:23.120
of its own if you look at all the productivity benchmarks together though you'll see that the core i9 averages

00:06:21.360 --> 00:06:29.759
about three percent lower than the more expensive 5950x and 10 higher than the

00:06:26.880 --> 00:06:33.280
less expensive 5900x which puts it in kind of weird

00:06:30.960 --> 00:06:39.360
might be worth it might not be worth it territory depending on what you're doing

00:06:35.600 --> 00:06:42.160
the core i5 on the other hand is over 30

00:06:39.360 --> 00:06:48.000
faster than AMD's competing ryzen 5 for just 10 dollars more at retail making it

00:06:45.360 --> 00:06:51.919
very compelling that is at least if you were already down to spend the extra to

00:06:49.919 --> 00:06:56.479
move to ddr5 we're going to have a follow-up comparing ddr4 to ddr5

00:06:54.639 --> 00:06:59.759
performance at various clock speeds coming so make sure that you're

00:06:57.759 --> 00:07:04.800
subscribed for that how is Intel doing all this though well first is the Intel

00:07:02.240 --> 00:07:08.160
7 manufacturing process it's basically just a rebrand of their 10 nanometer

00:07:06.720 --> 00:07:13.199
process that we've already seen in laptops but this is its first appearance

00:07:10.639 --> 00:07:17.599
without thin and light power constraints and boy did they ever unconstrain it

00:07:16.000 --> 00:07:22.639
because another big part of the equation is Intel effectively making support for

00:07:20.000 --> 00:07:27.120
multi-core enhancement official what this does is throw the processor

00:07:25.120 --> 00:07:32.800
base power rating completely out the window in favor of always running at

00:07:29.680 --> 00:07:34.960
maximum turbo power letting the CPU be

00:07:32.800 --> 00:07:39.520
unrestrained as long as there's thermal and power budget available it makes such

00:07:37.759 --> 00:07:43.440
a dramatic difference to performance that motherboard manufacturers who

00:07:41.120 --> 00:07:47.360
enabled this by default in the past were accused of cheating at benchmarks

00:07:46.240 --> 00:07:52.319
of course like any overclocking it comes at a cost

00:07:50.160 --> 00:07:58.240
the processor base power for the core i9 12900k is 125 watts same as the rest of

00:07:55.919 --> 00:08:02.199
the lineup however when loaded down with a long blender render this thing easily

00:08:00.720 --> 00:08:08.080
sucked back over 230 watts that is nearly double what our

00:08:05.520 --> 00:08:12.800
ryzen 9 chips drew thankfully however this doesn't carry over to gaming where

00:08:10.160 --> 00:08:19.440
f1 2021 stayed within the processor base power of 125 watts often remaining lower

00:08:16.240 --> 00:08:21.919
than ryzen so as it turns out those

00:08:19.440 --> 00:08:26.879
efficiency cores can make a big difference to lighter workloads as for

00:08:23.840 --> 00:08:29.599
the core i5 12600k it stayed within the

00:08:26.879 --> 00:08:34.560
125 watt processor base power on both the blender workload and our f1 2021

00:08:32.399 --> 00:08:39.200
gaming test where it's only a little bit more power hungry than the ryzen 5 5600

00:08:36.959 --> 00:08:42.800
x continues to get even more compelling doesn't it something to note though is

00:08:40.959 --> 00:08:48.480
that you won't see anything close to this boost forever behavior on a non-k

00:08:46.000 --> 00:08:52.399
chip so don't assume that just because you see two cpus with a similar model

00:08:50.720 --> 00:08:57.120
number that you will see similar performance especially not at the very

00:08:54.880 --> 00:09:02.240
high end to keep thermals under control under these circumstances Intel thinned

00:08:59.519 --> 00:09:06.080
out their die even further and even thinned out the soldered thermal

00:09:03.839 --> 00:09:10.640
interface material but in spite of their efforts the fact is that anything

00:09:08.240 --> 00:09:16.080
drawing that much power that is that small is going to get hot

00:09:13.120 --> 00:09:21.600
our nhd15 strained to keep our CPU package below 90 degrees and after about

00:09:19.200 --> 00:09:27.600
8 minutes all pretenses that thermals might be manageable were completely gone

00:09:24.160 --> 00:09:29.360
for context guys the nhd15 is the top

00:09:27.600 --> 00:09:33.760
performing heatsink on the market and it beats out many water coolers at least

00:09:32.000 --> 00:09:39.200
the efficiency cores only got to around 70 or 80 degrees so that's a plus i

00:09:36.560 --> 00:09:46.560
guess but compared to even the ryzen 9 59 50 x a 16 performance core CPU

00:09:44.080 --> 00:09:51.519
this thing packs a ridiculous amount of heat into a tiny area as much or more

00:09:49.839 --> 00:09:55.760
than last gen rocket lake with multi-core enhancement enabled again

00:09:53.440 --> 00:10:01.279
though it's far more manageable with gaming where f1 2021 pushed our core i9

00:09:58.480 --> 00:10:05.519
into only the mid 60s our core i5 for its part never got hotter than around 70

00:10:03.440 --> 00:10:10.240
degrees even in our all-out blender render while only hitting 60 in games

00:10:08.399 --> 00:10:14.320
it almost feels like Intel sent us the core i5 because

00:10:12.000 --> 00:10:18.079
that's the CPU that they really wanted us to review but they knew we'd want to

00:10:16.399 --> 00:10:22.640
look at the core i9 so they threw that in the box as well like how much extra

00:10:20.320 --> 00:10:27.120
heat is that for what six more FPS from a gaming perspective save your money buy

00:10:24.880 --> 00:10:31.120
a water bottle at lttstore.com and it gets even worse for the core i9

00:10:29.360 --> 00:10:34.880
these results mean that if you intend to run the CPU at the red line a lot maybe

00:10:33.279 --> 00:10:39.519
you're a content creator who does a lot of rendering or you're a developer who

00:10:37.120 --> 00:10:44.240
runs a lot of large compile operations you're going to want to invest in a big

00:10:41.440 --> 00:10:47.440
cooler both for your CPU and for your computer room during the summer months

00:10:45.760 --> 00:10:52.399
and that's a cost that you do have to consider another cost to consider is the

00:10:49.680 --> 00:10:57.920
platform we already mentioned that ddr5 won't come cheap but the same is true of

00:10:55.120 --> 00:11:01.519
the z690 chipset motherboards that 12th gen cpus require

00:10:59.760 --> 00:11:06.880
now there will be long-term benefits to taking the plunge pci express gen 5

00:11:04.160 --> 00:11:10.959
which doubles bandwidth over PCIe gen 4 is going to unlock the full capabilities

00:11:08.560 --> 00:11:16.880
of the next generation z of graphics cards and the new dmi 4.0 link between

00:11:14.160 --> 00:11:20.640
the CPU and the chipset means more high speed networking more high speed

00:11:18.480 --> 00:11:23.920
peripherals more storage devices you guys can get the full details on this in

00:11:22.480 --> 00:11:27.279
our platform overview which we're going to have linked below what you need to

00:11:25.680 --> 00:11:32.320
know now though is should you pull the trigger if an op cooler is in your

00:11:29.440 --> 00:11:36.720
budget the core i9 12900k would get the nod from us for both

00:11:34.240 --> 00:11:40.399
gamers and for power users that is assuming that gaming was your top

00:11:38.240 --> 00:11:48.240
priority but the really interesting chip today is the core i5 12600k

00:11:43.519 --> 00:11:51.760
at 320 bucks it might be only a 6 core

00:11:48.240 --> 00:11:53.440
CPU like a ryzen 5 but those extra 4

00:11:51.760 --> 00:11:57.279
efficiency cores boy do they ever pick up a lot more

00:11:55.360 --> 00:12:03.279
slack than we assumed they would i mean this is a legit generational leap over

00:12:00.240 --> 00:12:05.120
even the 5600x and in my opinion it is

00:12:03.279 --> 00:12:09.440
the big winner today just like our sponsor is the big winner

00:12:07.200 --> 00:12:12.240
jackery jackery makes solar ready portable power stations that are

00:12:10.720 --> 00:12:16.720
designed to improve outdoor life by providing power outdoors whenever and

00:12:14.480 --> 00:12:20.160
wherever you need it the explorer 1500 portable power station has enough juice

00:12:18.240 --> 00:12:26.079
to keep all your devices powered and connected it's huge 1500 watt hour

00:12:22.720 --> 00:12:27.920
capacity and 1800 watt output rate

00:12:26.079 --> 00:12:31.519
allows up to seven devices to be plugged in simultaneously it takes only four

00:12:29.920 --> 00:12:36.000
hours to recharge from zero to eighty percent and you can get ten percent off

00:12:34.079 --> 00:12:39.440
with code Linus tech tips at the link down below if you guys enjoyed this

00:12:37.519 --> 00:12:42.959
video go check out our full overview of the platform because there's a lot more

00:12:40.880 --> 00:12:47.279
juicy details in there that just didn't make the cut for our performance review