WEBVTT

00:00:01.680 --> 00:00:10.480
did you guys just see the same thing i did rtx 3070 is gonna outperform a 2080 ti

00:00:07.839 --> 00:00:14.880
at less than half the price 4.99 is NVIDIA insane

00:00:13.759 --> 00:00:22.080
i mean even with all the leaks indicating that ampere was going to be super powerful i

00:00:19.439 --> 00:00:26.080
was in no way ready for this so okay

00:00:24.080 --> 00:00:32.160
deep breath there's a lot to get through here so let's get started

00:00:29.279 --> 00:00:37.440
after an f in the chat for any recent 2080 ti buyers

00:00:34.640 --> 00:00:41.280
ouch like dang oh and also our sponsor

00:00:39.600 --> 00:00:45.360
ridge wallet wants to redefine the wallet with its compact frame and rfid

00:00:43.440 --> 00:00:49.600
blocking plates check out how they can help keep your wallet bulge down and use

00:00:47.440 --> 00:00:53.640
our offer code Linus to save 10 and get free worldwide shipping

00:01:00.559 --> 00:01:06.640
now obviously i'm not ready to hand NVIDIA their crown until we see the new

00:01:05.040 --> 00:01:10.479
cards for ourselves but based on what they've shown we are

00:01:09.040 --> 00:01:14.640
in for one heck of a ride they announced

00:01:13.760 --> 00:01:21.920
well a lot and we're going to get to all of it but the big one is ampere is real and

00:01:19.520 --> 00:01:25.439
ampere is an absolute monster

00:01:22.960 --> 00:01:32.159
as it turns out the recent leaks were accurate in many if not most ways where

00:01:29.360 --> 00:01:37.040
touring based rtx 2000 gpus were manufactured on tsmc's 12 nanometer

00:01:34.960 --> 00:01:42.640
finfet process which was actually derived from their 16 nanometer process

00:01:39.759 --> 00:01:48.479
NVIDIA instead partnered with samsung on a custom eight nanometer process for the

00:01:44.960 --> 00:01:49.520
rtx 3000 series and the difference here

00:01:48.479 --> 00:01:55.920
is much bigger than we got moving from pascal to touring

00:01:53.759 --> 00:02:00.960
here's the thing for over three years now GeForce gpus have had in the

00:01:58.000 --> 00:02:05.840
neighborhood of 20 to 25 million transistors per square millimeter ampere

00:02:04.560 --> 00:02:11.680
around 60 million per square millimeter that's

00:02:09.200 --> 00:02:16.640
over double and breaking with what's become their habit lately they didn't

00:02:13.680 --> 00:02:21.120
just use this increase in density to lower their own costs and keep selling a

00:02:19.360 --> 00:02:27.520
similar performance with some new features tacked on the rtx 3080 boasts

00:02:24.080 --> 00:02:30.160
over 28 billion that's with a b

00:02:27.520 --> 00:02:36.640
transistors what does that get us well for starters pci express gen 4

00:02:33.440 --> 00:02:40.080
upgraded tents of course oh and casually

00:02:36.640 --> 00:02:42.560
more than double the cuda course across

00:02:40.080 --> 00:02:48.239
the board and here's the kicker that's not compared to the cards that each of

00:02:45.040 --> 00:02:50.160
the 3000 series GPU replaces

00:02:48.239 --> 00:02:54.239
that's compared to the previous generations step up

00:02:52.879 --> 00:03:03.599
i'm not kidding on paper every ampere card that was announced

00:02:58.160 --> 00:03:06.000
even the 30 70 is faster than a 2080 ti

00:03:03.599 --> 00:03:10.560
to find the last time that this happened a third rung down GPU being faster than

00:03:08.480 --> 00:03:16.080
the previous flagship i had to dig pretty deep through the archives here's

00:03:12.800 --> 00:03:19.599
a review from anantek showing the 6600

00:03:16.080 --> 00:03:22.560
gt beating the fx 5950

00:03:19.599 --> 00:03:27.519
that was in 2004 and according to NVIDIA things are even

00:03:24.959 --> 00:03:33.120
better than they seem they claim we are getting a 1.9 times performance per watt

00:03:30.720 --> 00:03:37.920
improvement 2.7 times the shader performance 1.7 times the ray tracing

00:03:35.120 --> 00:03:41.440
performance and 2.7 times the tensor core performance which in the real world

00:03:40.000 --> 00:03:48.400
translates to sorry you'll have to give me a second here

00:03:45.040 --> 00:03:50.879
one metric butt ton NVIDIA showed

00:03:48.400 --> 00:03:58.080
another marvel's fully retraced graphics demo but while touring ran it at 720p

00:03:55.040 --> 00:04:01.680
ampere managed it at 1440p with more

00:03:58.080 --> 00:04:04.400
scene complexity that's huge and well to

00:04:01.680 --> 00:04:08.480
be fair they had a little help from dlss and from a new learning algorithm that

00:04:06.319 --> 00:04:13.680
derives missing color data from an incomplete image to improve performance

00:04:11.120 --> 00:04:17.440
optimizations as a concept are nothing new and the results here

00:04:15.840 --> 00:04:22.000
speak for themselves NVIDIA's also got a new memory signaling

00:04:20.000 --> 00:04:26.639
technique which carries four possible values per cycle compared to only two

00:04:24.560 --> 00:04:31.520
effectively doubling the amount of memory bandwidth while also increasing

00:04:29.040 --> 00:04:37.360
the frequency significantly so with the rtx 3080 you're going to get 10 gigs and

00:04:34.720 --> 00:04:42.960
it's going to run 320 watts of rated power with the rtx 3070 getting eight

00:04:40.000 --> 00:04:48.400
gigs of a more traditional gddr6 at a mere

00:04:44.199 --> 00:04:52.160
220 watts put that in context a 2080 ti

00:04:48.400 --> 00:04:52.160
was only 250 watts

00:04:52.639 --> 00:04:58.960
and NVIDIA claims that that rtx 3070

00:04:56.000 --> 00:05:06.160
even though it consumes less power is faster than the 1200 2080 ti with the

00:05:02.639 --> 00:05:10.720
new rtx 3080 doubling the performance of

00:05:06.160 --> 00:05:12.960
the 2080 and this is at 4.99 and 699

00:05:10.720 --> 00:05:17.120
respectively there's nothing else i can say

00:05:14.800 --> 00:05:24.720
i'm absolutely mind blown they even light up but the question is why

00:05:21.280 --> 00:05:26.000
why after all these years of driving GPU

00:05:24.720 --> 00:05:32.000
prices up has NVIDIA decided to

00:05:28.160 --> 00:05:33.919
bestow upon us these benevolent prices

00:05:32.000 --> 00:05:39.520
we've got a couple of theories one is that rtx 2080 and 2070 sales were

00:05:37.600 --> 00:05:43.199
something of a disappointment i mean it's not that they were terrible cards

00:05:41.360 --> 00:05:47.520
they just weren't a very compelling upgrade for anyone who already owned a

00:05:45.280 --> 00:05:51.360
high-end 1000 series so looking back then in

00:05:50.240 --> 00:05:56.960
whatever the past version of a crystal ball is

00:05:53.600 --> 00:05:59.440
it seems like NVIDIA priced touring high

00:05:56.960 --> 00:06:03.759
due to uncertain yields with tsmc and then hoped that rtx ray tracing would be

00:06:02.160 --> 00:06:09.440
enough to carry it to the new price point so then now because samsung has

00:06:06.479 --> 00:06:14.240
worked with NVIDIA to make this custom eight nanometer process

00:06:11.280 --> 00:06:18.400
presumably it's more reliable which means that despite the cost per wafer

00:06:16.479 --> 00:06:22.639
almost certainly being higher as they tend to be as we keep shrinking the

00:06:20.479 --> 00:06:27.280
process lower they're getting more usable chips out of them working out to

00:06:24.800 --> 00:06:31.199
a much lower cost which is great for consumers i think it's fair to say that

00:06:29.120 --> 00:06:35.759
this is shaping up to be an absolute unicorn of a GPU release make sure you

00:06:34.080 --> 00:06:39.759
guys are subscribed by the way so you don't miss our review

00:06:38.000 --> 00:06:45.280
and there's also another angle to the pricing with launch date set for the end of q3

00:06:43.280 --> 00:06:50.720
and beginning of q4 it's pretty clear who NVIDIA sees as the

00:06:47.840 --> 00:06:55.680
real competition for 3000 series and it's not AMD

00:06:53.120 --> 00:07:01.120
or at least not directly anyway it's sony and microsoft with the xbox series

00:06:58.479 --> 00:07:06.479
x and playstation 5. and by getting out ahead of those launches NVIDIA seems to

00:07:03.919 --> 00:07:10.560
be trying to steal some of the thunder hopefully preventing gamers from being

00:07:08.400 --> 00:07:13.759
lured away from pc where NVIDIA dominates over to consoles where they

00:07:12.800 --> 00:07:17.840
will i don't know i guess there's nintendo switch

00:07:18.000 --> 00:07:26.880
one big hint here is that just like the ps5 the rtx 3000 gpus have the ability

00:07:23.599 --> 00:07:28.639
to decompress data directly from an SSD

00:07:26.880 --> 00:07:34.080
attached to the system using direct storage this means load times can be

00:07:31.520 --> 00:07:37.840
much shorter and the CPU doesn't need to get bogged down decompressing the

00:07:35.919 --> 00:07:41.039
gigantic assets that make up today's game worlds

00:07:39.039 --> 00:07:45.120
just like on the playstation 5 then developers who leverage this technology

00:07:42.960 --> 00:07:50.160
can achieve more seamless loading and asset streaming than was previously

00:07:47.120 --> 00:07:53.199
possible it also means that NVMe and

00:07:50.160 --> 00:07:55.599
even pci express gen 4 might actually

00:07:53.199 --> 00:08:00.960
become useful for gamers where load times are notoriously CPU bound in spite

00:07:58.639 --> 00:08:05.680
of the storage speed of modern systems NVIDIA is also hyper focused on content

00:08:03.440 --> 00:08:11.039
creation here looking to lure streamers and machinima enthusiasts with NVIDIA

00:08:07.680 --> 00:08:13.120
broadcast and NVIDIA omniverse machinima

00:08:11.039 --> 00:08:18.879
NVIDIA broadcast includes a number of tools to apply effects to the microphone

00:08:15.360 --> 00:08:21.120
input camera and speakers using tensor

00:08:18.879 --> 00:08:25.280
cores in what looks like a continuation of what rtx voice started

00:08:23.280 --> 00:08:29.440
so in addition to the noise removal that was possible before we can now without a

00:08:27.759 --> 00:08:35.599
green screen automatically remove backgrounds and apply a blur effect or

00:08:32.080 --> 00:08:37.360
backdrop and more with what looks like a

00:08:35.599 --> 00:08:41.440
very convincing effect compared to current solutions

00:08:38.719 --> 00:08:45.279
NVIDIA omniverse machinima or nom i guess

00:08:43.760 --> 00:08:49.200
wash it down with a drink with a water bottle from ltdstore.com

00:08:47.680 --> 00:08:53.760
no okay whatever nom lets you import game libraries and

00:08:51.440 --> 00:08:58.320
tools apply materials physics simulations poses captured from webcams

00:08:56.720 --> 00:09:03.839
automatic facial animations from recorded audio and more to create a

00:09:01.680 --> 00:09:09.680
wholly unique scene for doing your own machinima it's kind of like ansel except

00:09:07.519 --> 00:09:13.920
that you get full control over the scene instead of just the camera they provided

00:09:12.240 --> 00:09:17.519
a pretty great looking example of a short clip of a siege to give us an idea

00:09:16.000 --> 00:09:21.519
of what can be accomplished and quite frankly i wouldn't be surprised if

00:09:19.519 --> 00:09:24.399
developers or even animators end up drafting scenes this way it's pretty

00:09:23.519 --> 00:09:31.360
cool so then rtx 3000 much less of a middle finger

00:09:29.040 --> 00:09:35.120
directly to Radeon technologies group and much more of a middle finger to the

00:09:33.760 --> 00:09:40.240
consoles which isn't to say though that NVIDIA hasn't done some extra credit work

00:09:38.080 --> 00:09:44.399
against rtg as well do you remember Radeon anti-lag and how NVIDIA

00:09:42.560 --> 00:09:48.399
originally claimed that it was just a tweak that users could already do to

00:09:46.080 --> 00:09:51.120
their drivers and then realized oops it's actually something more and then

00:09:49.760 --> 00:09:55.200
they've released NVIDIA ultra low latency well now they're putting the

00:09:53.519 --> 00:09:59.600
cherry on top with what they're calling NVIDIA reflex a driver path optimization

00:09:58.080 --> 00:10:05.120
technique that should be available to all gtx 1000 and newer users to reduce

00:10:03.040 --> 00:10:09.440
total system latency by up to roughly half now the exact mechanisms that they're

00:10:07.680 --> 00:10:14.720
exploiting to do this are currently unknown but presumably there's a cost of

00:10:11.920 --> 00:10:18.560
some sort somewhere and you'll probably lose a bit of driver level control in

00:10:16.880 --> 00:10:24.320
the process finally there's the the one more thing

00:10:21.360 --> 00:10:29.839
rtx 3090 is the formal name of this generation's titan but jensen lovingly

00:10:26.720 --> 00:10:33.600
referred to it as the bf GPU it sucks

00:10:29.839 --> 00:10:36.959
back a rated 350 watts hence the fancy

00:10:33.600 --> 00:10:40.959
new 12-pin power connector has over 10

00:10:36.959 --> 00:10:44.640
000 cuda cores and 24 gigs of GDDR 6x

00:10:40.959 --> 00:10:47.279
memory and comes in at 14.99 meaning

00:10:44.640 --> 00:10:51.680
it's certainly priced like a titan but unlike the last couple of generations of

00:10:49.200 --> 00:10:56.480
titans it also appears to be specked like a titan and it might even be worth

00:10:54.079 --> 00:11:02.720
that obscene price tag they rolled out an lg 8k OLED tv to game on and

00:10:59.440 --> 00:11:05.760
demonstrated 8k 60fps gameplay using

00:11:02.720 --> 00:11:07.680
dlss and rtx

00:11:05.760 --> 00:11:12.560
not only did the beefed up tensor cores appear to handle 8k upscaling without

00:11:10.000 --> 00:11:17.760
breaking a sweat the rtx 3090 can apparently even record 8k HDR gameplay

00:11:16.160 --> 00:11:21.279
with shadowplay in fact there's bits and pieces in the

00:11:19.680 --> 00:11:26.399
whole presentation that allude to upgrades across the board av1 codec

00:11:24.079 --> 00:11:31.440
support for decode is now available and nvenk is presumably no different on the

00:11:28.560 --> 00:11:34.880
rest of the ampere cards making 4k HDR shadow play a thing now

00:11:33.360 --> 00:11:40.640
to go with it it looks like we're finally getting HDMI 2.1 and probably

00:11:38.000 --> 00:11:44.160
more but we can figure all of that out in our review for now all that's left to

00:11:43.200 --> 00:11:49.440
say is boy is it ever good to see the effects of more competition in the gaming

00:11:47.519 --> 00:11:53.760
hardware space even if it came from a bit of an

00:11:51.040 --> 00:11:57.360
unexpected direction unlike my sponsor segways you guys

00:11:55.680 --> 00:12:00.880
always expect those this time for freshbooks freshbooks is the small

00:11:59.440 --> 00:12:04.880
business cloud accounting solution that's built for owners with freshbooks

00:12:03.120 --> 00:12:08.000
you can work anywhere thanks to their mobile apps allowing you to create

00:12:06.160 --> 00:12:12.079
professional looking invoices in seconds on the go snap pictures of your receipts

00:12:09.920 --> 00:12:16.480
so you don't lose them and this is new on iOS even automatically track your

00:12:14.560 --> 00:12:20.399
mileage so you can be organized and ready to maximize your deductions when

00:12:18.320 --> 00:12:23.600
tax time comes with freshbooks you'll never miss an update when a client has

00:12:22.160 --> 00:12:27.120
viewed your invoice or even when it becomes overdue it'll let you know and

00:12:25.440 --> 00:12:33.279
you can get fifty percent off your first three months of fresh books when you sign up for a paid plan at

00:12:29.920 --> 00:12:34.720
freshbooks.com forward slash tech tips

00:12:33.279 --> 00:12:38.800
if you guys enjoyed this video and you'd like a bit of a refresher on what a bad

00:12:36.800 --> 00:12:43.120
GPU launch looks like maybe go check out our review of the rtx 2080 seems like

00:12:41.279 --> 00:12:47.360
NVIDIA has learned a thing or two since then

00:12:45.279 --> 00:12:53.360
huh or maybe check out our 8k gaming video ah that was that was terrible

00:12:49.600 --> 00:12:53.360
that's gonna be way better now