1
00:00:00,640 --> 00:00:06,960
NVIDIA's top of the line rtx 30

2
00:00:04,480 --> 00:00:06,960
90.

3
00:00:09,120 --> 00:00:16,560
no no there's no time for that we've got to talk about this thing now

4
00:00:14,480 --> 00:00:22,000
if you haven't watched our rtx 3080 review here's the background you need

5
00:00:18,640 --> 00:00:24,240
rtx 3000 also known as ampere is what

6
00:00:22,000 --> 00:00:29,279
NVIDIA is calling the biggest generational leap in performance in the

7
00:00:26,480 --> 00:00:35,200
company's history now if you go back far enough i'ma press x here

8
00:00:31,760 --> 00:00:38,800
but he's also not joking these cards

9
00:00:35,200 --> 00:00:41,440
unlike the touring based rtx 2000 series

10
00:00:38,800 --> 00:00:46,640
are smoking fast right out of the gate with our rtx 3080 reaching speeds in the

11
00:00:43,680 --> 00:00:50,239
neighborhood of 75 to 100 faster than its predecessor depending on

12
00:00:49,120 --> 00:00:55,600
the task that combined with NVIDIA's super

13
00:00:52,399 --> 00:00:58,719
impressive dlss 2.0 upscaling technology

14
00:00:55,600 --> 00:01:00,719
means that 4k gaming went from

15
00:00:58,719 --> 00:01:02,960
someday to today

16
00:01:01,760 --> 00:01:07,280
right now but wait there's even more like this

17
00:01:05,680 --> 00:01:10,880
message from our sponsor team group team group's extreme argb ddr4 gaming memory

18
00:01:09,600 --> 00:01:15,200
features clock speeds up to 4000 megahertz and a stunning mirror design with RGB check it out on their launch

19
00:01:13,840 --> 00:01:20,720
event showing off their three brands and you can run a 3 000 pc at the link below

20
00:01:17,600 --> 00:01:23,759
but the 3080 is last week's news in my

21
00:01:20,720 --> 00:01:26,640
hands right now is what jensen ceo of

22
00:01:23,759 --> 00:01:31,840
NVIDIA lovingly referred to as the bfgpu

23
00:01:28,799 --> 00:01:35,280
and spoiler alert he's right

24
00:01:31,840 --> 00:01:37,840
this is an rtx titan for scale heck

25
00:01:35,280 --> 00:01:42,640
here's an rod strix oc for scale this thing is easily the largest not to

26
00:01:40,479 --> 00:01:48,960
mention heaviest conventional graphics card that i have ever seen which means

27
00:01:46,000 --> 00:01:53,680
that anyone who buys one needs to also make some additional investments

28
00:01:51,200 --> 00:02:00,079
first up you're gonna need a case with a solid GPU mount and enough clearance

29
00:01:57,680 --> 00:02:05,280
being both taller and thicker than previous flagships it simply won't fit

30
00:02:02,880 --> 00:02:11,440
in most small form factor cases that is unless you hot rod them

31
00:02:07,520 --> 00:02:12,680
second you will need a fat power supply

32
00:02:11,440 --> 00:02:19,040
rated at 350 watts of power consumption and with

33
00:02:15,920 --> 00:02:21,280
NVIDIA's new 12 pin connector the 3090

34
00:02:19,040 --> 00:02:24,640
could easily push your previous unit to its limits

35
00:02:22,480 --> 00:02:29,760
last but not least you're going to need the fastest gaming CPU in the world

36
00:02:26,879 --> 00:02:33,599
because quite frankly with anything else you are going to be leaving performance

37
00:02:31,920 --> 00:02:38,879
on the table which is something you don't want to do when

38
00:02:35,280 --> 00:02:41,760
you're paying 1500 us dollars for a

39
00:02:38,879 --> 00:02:48,319
graphics card yes my friends compared to the rtx 3080 the world's first 8k gaming

40
00:02:45,680 --> 00:02:54,959
GPU gets you more than double not to mention faster video memory 20 more cuda

41
00:02:51,680 --> 00:02:58,080
tensor and rt cores and

42
00:02:54,959 --> 00:03:00,560
even nv link support via a new connector

43
00:02:58,080 --> 00:03:05,760
that was presumably shrunk to minimize wasted space on the physically cut down

44
00:03:02,720 --> 00:03:08,640
pcb that makes NVIDIA's new flow-through

45
00:03:05,760 --> 00:03:12,800
cooler possible but wait double the video memory

46
00:03:10,959 --> 00:03:18,319
why well aside from giving it much better compute

47
00:03:16,000 --> 00:03:23,200
capabilities and more importantly the ability to render much more complex 3d

48
00:03:20,800 --> 00:03:30,720
scenes in software like blender as it turns out gaming at 8k also genuinely

49
00:03:27,200 --> 00:03:35,840
demands this much video memory i mean

50
00:03:30,720 --> 00:03:37,200
think about it ak is 33 megapixels at 32

51
00:03:35,840 --> 00:03:40,599
bits per pixel that means to display just a single

52
00:03:39,599 --> 00:03:47,120
frame 128 megabytes of video memory is needed

53
00:03:43,840 --> 00:03:48,879
which doesn't sound so bad but add in

54
00:03:47,120 --> 00:03:55,920
textures models screen space effects and of course rtx real time ray tracing i mean come on

55
00:03:53,599 --> 00:03:59,599
and you've got yourself many times that ideally 60 times per second or more

56
00:03:57,920 --> 00:04:06,480
wolfenstein youngblood clocked in at over 11 gigs of memory usage at 8k

57
00:04:03,920 --> 00:04:12,239
sounds fun but at the end of the day it all boils down to performance doesn't it

58
00:04:09,599 --> 00:04:17,759
so by request we built up a core i9 10900k test bench to minimize our CPU

59
00:04:14,720 --> 00:04:21,280
bottlenecks and dug out our rtx 3080

60
00:04:17,759 --> 00:04:23,440
along with a 2080 ti and titan rtx for

61
00:04:21,280 --> 00:04:28,880
comparison starting off with minecraft rtx we're seeing about a 22 average

62
00:04:26,160 --> 00:04:36,720
frame rate improvement over the rtx 3080 and a whopping 51 to 60 percent over the

63
00:04:32,320 --> 00:04:37,840
titan rtx both with the lss on and off

64
00:04:36,720 --> 00:04:44,800
not bad wolfenstein youngblood meanwhile gives us a much more modest performance

65
00:04:41,520 --> 00:04:46,400
improvement over the 3080 at around 15

66
00:04:44,800 --> 00:04:52,160
but while still pulling off a respectable 50 improvement over the

67
00:04:48,960 --> 00:04:53,360
titan rtx a card that costs a thousand

68
00:04:52,160 --> 00:04:57,919
dollars more so there you have it guys

69
00:04:55,440 --> 00:05:02,080
buy yourself an rtx 3090 and a water bottle with all that money you saved am

70
00:04:59,360 --> 00:05:06,000
i right lttstore.com csgo and microsoft flight simulator tell a very similar

71
00:05:04,240 --> 00:05:10,800
story in 4k but interestingly the gap narrows

72
00:05:08,600 --> 00:05:17,919
significantly for flight simulator at 1440p because of CPU bottlenecks

73
00:05:14,400 --> 00:05:20,000
that's right despite the op CPU and

74
00:05:17,919 --> 00:05:25,360
ultra graphic settings finally the parallel n64 emulator which

75
00:05:23,120 --> 00:05:31,120
does most of the graphics and sound emulation on the GPU rather than the CPU

76
00:05:28,479 --> 00:05:36,720
and recently added upscaling support shows us that the extra oomph brings us

77
00:05:33,280 --> 00:05:38,720
up to 99th percentile frame times near

78
00:05:36,720 --> 00:05:43,360
60 FPS that eight times upscale by the way

79
00:05:40,560 --> 00:05:46,479
works out to more pixels than 5k emulated

80
00:05:45,440 --> 00:05:50,800
but then there's the elephant in the room

81
00:05:49,280 --> 00:05:56,800
8k gaming as tom's hardware reported NVIDIA's

82
00:05:53,520 --> 00:05:59,919
demos only showed 8k aaa gaming with

83
00:05:56,800 --> 00:06:02,880
dlss 2.0 in what they're calling

84
00:05:59,919 --> 00:06:07,680
ultra performance mode now before dlss 2.0 offered performance and quality

85
00:06:05,759 --> 00:06:14,639
scaling options with performance rendering at 1080p like the lss 1.0 and

86
00:06:11,280 --> 00:06:17,360
quality rendering at 1440p before

87
00:06:14,639 --> 00:06:22,240
upscaling usually to 4k now

88
00:06:18,240 --> 00:06:25,280
ultra performance still uses 1440p

89
00:06:22,240 --> 00:06:27,520
but instead of performing a reasonable

90
00:06:25,280 --> 00:06:32,800
scaling jump up to 4k it's trying to take that data and go all

91
00:06:30,240 --> 00:06:38,560
the way up to 8k i mean that's a lot of guesses to make in terms

92
00:06:35,440 --> 00:06:40,800
of hmm what are those smeary pixels

93
00:06:38,560 --> 00:06:46,080
so does it look any good well on a 4k display there's little

94
00:06:44,160 --> 00:06:50,639
point in using it since the only real effect is similar to anti-aliasing which

95
00:06:48,560 --> 00:06:54,800
can be had with less performance penalty in other ways

96
00:06:52,080 --> 00:06:58,560
but if you are one of the privileged few with an 8k display today

97
00:06:56,960 --> 00:07:05,919
here are some shots of what you can expect to see from 4k native versus 8k

98
00:07:01,759 --> 00:07:09,039
dlss ultra performance versus 8k native

99
00:07:05,919 --> 00:07:11,440
you can see clearly that 8k dlss ultra

100
00:07:09,039 --> 00:07:16,000
performance delivers significantly more detail than 4k does

101
00:07:13,360 --> 00:07:21,280
even if it isn't quite up to 8k native standards i mean check out the text and

102
00:07:18,639 --> 00:07:27,440
distant details in this shop for example 8k dlss ends up much closer to 8k native

103
00:07:24,639 --> 00:07:30,960
here even if it's still not perfect and the same deal with these grates on

104
00:07:29,120 --> 00:07:37,199
the floor in this shot they're almost completely wiped out in 4k but dlss

105
00:07:34,400 --> 00:07:43,120
restores that now performance is roughly 25 less than 4k but

106
00:07:40,319 --> 00:07:47,039
that still puts us above the magic 60 FPS mark on average which is pretty

107
00:07:45,599 --> 00:07:54,160
sweet if you shelled out for one of those new 8k tvs with HDMI 2.1

108
00:07:50,479 --> 00:07:57,360
even better is that not every game even

109
00:07:54,160 --> 00:08:00,800
aaa games necessarily needs to use dlss

110
00:07:57,360 --> 00:08:03,360
we ran do maternal at native ak

111
00:08:00,800 --> 00:08:08,080
with very high detail settings for what was honestly the most mind-bending

112
00:08:05,520 --> 00:08:13,599
experience i've had with this GPU so far and we barely saw dips below 60 FPS also

113
00:08:12,160 --> 00:08:19,039
impressive during that session with the 88 inch tv was HDR gameplay recording

114
00:08:16,879 --> 00:08:23,919
with shadowplay now we had wanted to include a big section about performance

115
00:08:21,360 --> 00:08:27,280
and image quality but the story is sort of already in the footage that you're

116
00:08:25,680 --> 00:08:31,360
looking at it just worked the performance hit was

117
00:08:29,360 --> 00:08:35,519
reasonable and it looks as good as you could expect given the 100 megabit per

118
00:08:33,360 --> 00:08:39,200
second bit rate at such a massive resolution

119
00:08:36,719 --> 00:08:43,200
our entire 8k gaming experience video yesterday was shot with it running in

120
00:08:41,440 --> 00:08:46,240
the background without any hiccups whatsoever

121
00:08:44,560 --> 00:08:51,760
great job productivity though is where the rtx

122
00:08:48,399 --> 00:08:53,519
3090 both gets to stretch its legs and

123
00:08:51,760 --> 00:08:57,600
also get a little bit dicey

124
00:08:55,760 --> 00:09:02,959
going up against the titan and blender we saw anywhere from a 50 improvement

125
00:08:59,920 --> 00:09:05,760
with the CPU optimized gooseberry render

126
00:09:02,959 --> 00:09:11,120
all the way up past 100 more performance in the pavilion barcelona optics render

127
00:09:08,560 --> 00:09:16,000
yes my friends gooseberry a benchmark designed for cpus can run on all of

128
00:09:13,920 --> 00:09:19,839
these cards thanks to their ridiculous amounts of vram

129
00:09:17,760 --> 00:09:24,000
other rendering benchmarks like v-ray redshift and octane bench all provide

130
00:09:22,320 --> 00:09:30,240
similar numbers across the board with the rtx 3090 hovering around the

131
00:09:26,640 --> 00:09:33,200
15 to 20 percent mark over the 3080 and

132
00:09:30,240 --> 00:09:37,920
50 to 80 percent over the titan then there's spec you perf where we can

133
00:09:35,279 --> 00:09:46,080
see that the rtx 3090 is sorry wait what no no go back to that

134
00:09:41,040 --> 00:09:46,080
katia ran over 40 faster on the titan

135
00:09:46,320 --> 00:09:53,680
how does that make sense okay so as it turns out

136
00:09:50,399 --> 00:09:57,120
this card while it has more raw

137
00:09:53,680 --> 00:10:00,320
hutzpah than the titan rtx and was

138
00:09:57,120 --> 00:10:02,880
introduced as if it were a titan-ish

139
00:10:00,320 --> 00:10:07,200
it isn't a titan when we spoke to NVIDIA they explained that there are driver

140
00:10:04,720 --> 00:10:12,560
optimizations for the titan that have not been enabled for the rtx 3090 and

141
00:10:10,240 --> 00:10:16,959
presumably won't and they confirmed that these results are 100 expected behavior

142
00:10:15,839 --> 00:10:21,680
which kind of sucks because if this thing is

143
00:10:20,160 --> 00:10:27,839
mostly just a giant overgrown

144
00:10:24,160 --> 00:10:30,480
3080 ti at 1500

145
00:10:27,839 --> 00:10:35,200
what's the ampere titan going to cost three grand anyway

146
00:10:33,200 --> 00:10:40,399
back to spec view perf we did get several wins for both the rtx 3090 and

147
00:10:38,160 --> 00:10:45,120
the titan rtx depending on the workload with siemens nx not even being in the

148
00:10:42,560 --> 00:10:48,800
same area code let alone ballpark that's kind of rough for professionals who are

149
00:10:46,880 --> 00:10:52,800
looking to juice up their workstation for cad work by day and gaming at 8k by

150
00:10:52,000 --> 00:10:58,880
night which do they choose and there's more bad news remember that

151
00:10:56,800 --> 00:11:04,160
bit about needing a new power supply well our test bench sure did even a high

152
00:11:01,040 --> 00:11:07,480
quality 850 watt couldn't keep up with

153
00:11:04,160 --> 00:11:10,880
our rtx 3090 routinely breaking

154
00:11:07,480 --> 00:11:13,120
375 watts with NVIDIA's pcap power meter

155
00:11:10,880 --> 00:11:17,680
connected throughout our testing with a peak power draw of over

156
00:11:15,640 --> 00:11:22,720
450 watts the titan rtx meanwhile barely broke 300

157
00:11:20,480 --> 00:11:28,079
watts the whole time thankfully the cooler as with the rtx

158
00:11:25,360 --> 00:11:31,440
3080 was well up to the task and it didn't even break a sweat through spec

159
00:11:29,760 --> 00:11:36,240
view perf with thermals never pushing past 70 degrees also core clocks were

160
00:11:34,160 --> 00:11:41,360
ridiculously stable at around 2 gigahertz under load so i guess all that

161
00:11:38,880 --> 00:11:45,839
weight and metal is doing its job and then some NVIDIA was clearly really

162
00:11:43,920 --> 00:11:50,480
concerned with thermals this time around and they just did a great job there's no

163
00:11:48,560 --> 00:11:55,040
other way of putting it and hopefully that means some overclocking headroom as

164
00:11:52,480 --> 00:12:00,640
well early reports on the rtx 3080 seemed to suggest that the GDDR 6x

165
00:11:57,519 --> 00:12:02,640
memory is pretty easily bumped up but

166
00:12:00,640 --> 00:12:06,000
we've also seen reports that NVIDIA went for lower speed RAM due to thermal

167
00:12:04,480 --> 00:12:10,399
limitations so as always stress test your overclocks

168
00:12:08,240 --> 00:12:14,079
with a prolonged load before considering them tournament ready and speaking of

169
00:12:12,480 --> 00:12:18,639
overclocking we have a video coming where we look at the differences between

170
00:12:16,160 --> 00:12:21,839
aftermarket gpus so make sure you get subscribed so you don't miss it

171
00:12:20,880 --> 00:12:27,120
for now these results leave me i mean there's no

172
00:12:24,880 --> 00:12:32,720
real word for it other than conflicted about the rtx 3090 on the one hand it's

173
00:12:30,320 --> 00:12:36,639
a beast of a graphics card but on the other hand

174
00:12:34,560 --> 00:12:40,880
it's clear that NVIDIA is holding it back so they don't eat into their titan

175
00:12:38,959 --> 00:12:44,000
or quadro lineups kind of like how Intel runs things with

176
00:12:42,639 --> 00:12:48,320
xeon it runs really cool with a fan that's

177
00:12:45,920 --> 00:12:54,399
not nearly as loud as the rtx 2000 series but it guzzles power and i can't

178
00:12:52,079 --> 00:12:58,480
help feeling like a doofus for buying it for workstation or cad use knowing that

179
00:12:56,800 --> 00:13:06,880
there could be a titan coming down the line that puts the 3090 squarely in prosumer

180
00:13:03,279 --> 00:13:09,200
slash enthusiast territory so like super

181
00:13:06,880 --> 00:13:13,839
high-end gamers content creators and especially 3d modelers are going to love

182
00:13:11,600 --> 00:13:17,519
it which isn't a bad thing it's just

183
00:13:15,200 --> 00:13:22,079
a niche thing and i can't help but feel that NVIDIA could have done more to make

184
00:13:19,600 --> 00:13:24,959
the 3090 shine do you think that's fair

185
00:13:25,760 --> 00:13:32,639
Anthony do you think that's fair yeah

186
00:13:30,639 --> 00:13:36,240
how about you jensen i know you watch our videos

187
00:13:34,480 --> 00:13:40,639
and i also know that you don't like it when you think i'm not being fair which

188
00:13:38,480 --> 00:13:46,000
is kind of funny to me because i get to sit here imagining a man who

189
00:13:42,880 --> 00:13:48,000
casually buys 40 billion companies and

190
00:13:46,000 --> 00:13:51,519
changes computing as we know it who still gets butt hurt when some nobody

191
00:13:49,680 --> 00:13:54,320
youtuber doesn't like his gaming widget enough

192
00:13:52,880 --> 00:13:58,240
but on the other hand you know what jensen

193
00:13:56,000 --> 00:14:02,560
i actually really like that about you it's that kind of passion that makes

194
00:14:00,639 --> 00:14:07,519
NVIDIA so great so i've resolved

195
00:14:04,639 --> 00:14:12,160
to keep making you mad because evidently when you got mad that i didn't like your

196
00:14:09,600 --> 00:14:16,320
rtx 2080 ti we got the 30 80.

197
00:14:14,480 --> 00:14:19,360
so i might go even harder on you guys just to see how far you can push pc

198
00:14:18,240 --> 00:14:23,760
gaming no i'm just kidding jensen

199
00:14:21,440 --> 00:14:29,920
sort of since i have your attention please unlock the titan driver paths and

200
00:14:26,000 --> 00:14:32,240
sri ov at least for the rtx 3090 this is

201
00:14:29,920 --> 00:14:37,040
a niche enthusiast card and it should really enjoy niche enthusiast features

202
00:14:34,800 --> 00:14:41,120
don't you think just like our viewers enjoy my segways

203
00:14:39,279 --> 00:14:46,399
to sponsors like privacy and one password two services

204
00:14:43,680 --> 00:14:49,760
that prioritize security and convenience starting now privacy users can

205
00:14:48,079 --> 00:14:54,240
automatically add their virtual cards into their one password vault with a

206
00:14:51,839 --> 00:14:58,880
single click and one password users will be able to create virtual cards in their

207
00:14:56,160 --> 00:15:02,320
browser and store them in one password when creating a new virtual card with

208
00:15:00,480 --> 00:15:07,519
one password users can set a spending cap and adjust settings for a one-off

209
00:15:04,399 --> 00:15:09,680
payment monthly or annual limits or a

210
00:15:07,519 --> 00:15:13,120
total amount users can limit cards so they can only be used at a single

211
00:15:11,120 --> 00:15:16,880
service to help keep things locked down and when it's time to enter payment

212
00:15:14,399 --> 00:15:20,480
details one password will show users any cards associated with that website

213
00:15:18,959 --> 00:15:25,680
including any privacy cards you've generated so head to privacy.com Linus

214
00:15:23,519 --> 00:15:29,839
and sign up for an account new privacy customers will automatically get five

215
00:15:27,440 --> 00:15:33,920
dollars to spend on their first purchase or if you're a new one password user

216
00:15:32,000 --> 00:15:37,600
head to the link below to also receive five dollars when signing up

217
00:15:35,920 --> 00:15:40,560
good stuff so thanks for watching guys if you're

218
00:15:39,040 --> 00:15:45,199
looking for something else to watch go check out our review of the rtx 3080

219
00:15:43,120 --> 00:15:49,600
where we happen to use a ryzen bench instead of an Intel bench does it

220
00:15:46,959 --> 00:15:53,199
perform better with PCIe gen 4 i'll see you over there