1
00:00:00,900 --> 00:00:07,200
okay let's do this reviewing the RTX

2
00:00:04,140 --> 00:00:10,139
4070 is kind of a no-win situation for

3
00:00:07,200 --> 00:00:15,540
me because on the one hand everyone is furious about the unaffordability of

4
00:00:12,240 --> 00:00:17,880
modern gpus and rightly so but on the

5
00:00:15,540 --> 00:00:23,220
other hand I have to acknowledge that this is the best value proposition that

6
00:00:20,340 --> 00:00:28,560
we've seen in years then on the other other hand there's the fact that NVIDIA

7
00:00:25,680 --> 00:00:32,640
who controls all of this appears to be making this product look like a good

8
00:00:30,779 --> 00:00:36,780
value by simply making all of their other ones a terrible value I really do

9
00:00:35,520 --> 00:00:40,620
wish I could just sit this one out because by the time I'm done at least

10
00:00:38,820 --> 00:00:48,360
some of you will have misdirected your anger toward me the messenger but here

11
00:00:44,520 --> 00:00:50,399
goes the segue to our sponsor videocom

12
00:00:48,360 --> 00:00:55,140
making a presentation for worker school let video com help you spice it up by

13
00:00:52,860 --> 00:00:58,500
adding annotations webcams videos and images and much more with just a few

14
00:00:56,940 --> 00:01:04,500
simple clicks check them out at the link down below looking at the specs alone

15
00:01:00,600 --> 00:01:07,439
the RTX 4070 is pretty underwhelming at

16
00:01:04,500 --> 00:01:13,860
just 295 millimeters squared this is the smallest 70 class GPU that we've seen

17
00:01:09,780 --> 00:01:17,220
since the GTX 770 and with the same Cuda

18
00:01:13,860 --> 00:01:19,860
core count as the RTX 3070 this is the

19
00:01:17,220 --> 00:01:24,259
first time ever that NVIDIA hasn't given us more shaders generation over

20
00:01:21,659 --> 00:01:27,840
generation but but wait wait not out there are some things to get excited

21
00:01:26,520 --> 00:01:34,200
about here um boost clocks have risen a whopping 43

22
00:01:30,780 --> 00:01:37,979
compared to the RTX 3070 and the change

23
00:01:34,200 --> 00:01:40,020
to 12 gigabytes of gddr6x memory is a

24
00:01:37,979 --> 00:01:45,659
welcome upgrade it's we continue to watch vram usage in modern games go from

25
00:01:42,140 --> 00:01:47,520
unreasonable to downright idiotic to

26
00:01:45,659 --> 00:01:54,840
quantify our excitement then the lab performed a whopping 984 tests across 11

27
00:01:51,180 --> 00:01:56,280
gpus with CPUs from both Intel and AMD

28
00:01:54,840 --> 00:02:01,979
we're going to have links down below they even threw in the Intel rk770 for

29
00:01:59,040 --> 00:02:06,780
funsies now 1080p performance in rasterized games might seem kind of

30
00:02:04,380 --> 00:02:11,700
irrelevant for an Enthusiast tier GPU these days but for many competitive

31
00:02:08,880 --> 00:02:16,500
Gamers High frame rates are way more important than high resolution and if

32
00:02:13,860 --> 00:02:22,500
you're one of them the RTX 4070 is proper quick with performance Landing

33
00:02:18,540 --> 00:02:24,480
within two percent of the RTX 380. this

34
00:02:22,500 --> 00:02:30,000
might not be the leap forward that you were hoping for but given the pricing of

35
00:02:26,760 --> 00:02:31,680
even second hand 3080s it's kind of

36
00:02:30,000 --> 00:02:38,640
looking like a w kind of give me a sec at 1440p the story

37
00:02:35,819 --> 00:02:44,040
Remains the Same with the 4070 neck and neck with the RTX 3080 but what really

38
00:02:42,000 --> 00:02:49,200
stood out to us here at both of these resolutions is how flipping fast AMD's

39
00:02:47,640 --> 00:02:56,760
last gen has gotten thanks to their ongoing software updates the 6800 XT now

40
00:02:53,640 --> 00:02:59,940
enjoys a small lead over the 3080 with

41
00:02:56,760 --> 00:03:03,000
the 7900 XT coming in a whole 40 percent

42
00:02:59,940 --> 00:03:07,099
faster than the 4070 and nearly matching

43
00:03:03,000 --> 00:03:11,760
the RTX 4080 a GPU that costs 400 more

44
00:03:07,099 --> 00:03:14,700
then at 4K the 4070s W gets even more

45
00:03:11,760 --> 00:03:22,379
twisted and complex because outside of Forza the RTX 3080 ends up outrunning it

46
00:03:18,239 --> 00:03:24,780
by anywhere from 5 to 19 percent

47
00:03:22,379 --> 00:03:29,400
or at least it does until you turn on Ray tracing where the Gap Narrows

48
00:03:26,640 --> 00:03:35,940
significantly now the story NVIDIA wants to focus on here is not how the 4070

49
00:03:32,580 --> 00:03:38,580
beats last gen by 25 to 30 percent but

50
00:03:35,940 --> 00:03:44,220
rather on how it convincingly destroys the GTX 1080. from my point of view

51
00:03:41,700 --> 00:03:49,200
though beating a seven-year-old card is not the most incredible of Feats unless

52
00:03:47,040 --> 00:03:53,280
you name zintel it's just obvious that NVIDIA is bringing this up because they

53
00:03:51,420 --> 00:03:57,780
recognize the generation over generation upgrade it's not that big and they hope

54
00:03:55,920 --> 00:04:01,440
that it's all you Pascal owners out there that are finally gonna bite the

55
00:03:59,400 --> 00:04:07,140
bullet and upgrade moving on to productivity in our benchmarks there was

56
00:04:03,540 --> 00:04:11,040
one standout in Photoshop the Intel Arc

57
00:04:07,140 --> 00:04:14,040
a770 was able to best not only the 4070

58
00:04:11,040 --> 00:04:16,739
but also the 4070 TI and was even within

59
00:04:14,040 --> 00:04:19,799
spitting distance of the RTX 4080 I mean I'm not sure that the average user

60
00:04:18,239 --> 00:04:23,340
updating their Tinder profile is going to notice but whatever Hardware

61
00:04:21,479 --> 00:04:28,080
acceleration Intel built for Arc definitely works then outside of

62
00:04:26,160 --> 00:04:31,500
Photoshop Arc returns to its rightful place at the bottom of the stack and the

63
00:04:29,820 --> 00:04:36,360
rest of the story remains largely the same as gaming with the RTX 3080 and

64
00:04:34,080 --> 00:04:41,400
4070 Trading Places depending on the exact workload and AMD once again

65
00:04:38,820 --> 00:04:44,520
showing off that fine line driver development

66
00:04:42,600 --> 00:04:49,560
outside of blender which remains a sore spot for team red the 6800 XT put on an

67
00:04:47,520 --> 00:04:54,419
amazing show winning more titles than it lost at a lower price

68
00:04:52,259 --> 00:05:00,840
so if we're looking at the performance of an RTX 3080 but at the price of an

69
00:04:57,000 --> 00:05:02,160
RTX 370 TI given how crap GPU prices

70
00:05:00,840 --> 00:05:11,520
have been for the last couple of years this is great right well yes but saying

71
00:05:07,800 --> 00:05:14,220
an RTX 4070 for 600 is better than an

72
00:05:11,520 --> 00:05:18,900
RTX 3080 for like what would it Peak at 1600 or something like that that's like

73
00:05:16,860 --> 00:05:23,039
saying that a paddle to the ass is better than a slap to the face

74
00:05:21,240 --> 00:05:27,180
both can be fun when the kids are out of town and we do have matching couples undies

75
00:05:25,500 --> 00:05:31,139
on lttstore.com by the way to make this even more awkward but none of that is

76
00:05:28,860 --> 00:05:37,979
the point the point is that when compared to last gen sure NVIDIA gave us

77
00:05:34,199 --> 00:05:40,500
a solid increase in performance but they

78
00:05:37,979 --> 00:05:45,180
also gave us an equally solid increase in price and this is the second time in

79
00:05:43,500 --> 00:05:50,160
two generations that they've pulled this move when we went from the 1000 series

80
00:05:47,699 --> 00:05:54,539
to the 2000 series they did the same thing we got a nearly proportional

81
00:05:52,259 --> 00:05:59,220
increase in both price and performance with NVIDIA hoarding the benefits of

82
00:05:57,060 --> 00:06:03,479
their new architecture and using them to line their own Pockets rather than

83
00:06:00,960 --> 00:06:07,880
passing them along to the consumer you see the industry Norm has been that as

84
00:06:05,820 --> 00:06:14,639
our manufacturing technology improves the same amount of compute becomes less

85
00:06:11,460 --> 00:06:18,000
expensive in this case NVIDIA went from

86
00:06:14,639 --> 00:06:20,460
Samsung's 8 nanometer node to tsmc N5

87
00:06:18,000 --> 00:06:26,819
that's how they packed the same number of Cuda cores into just 295 millimeters

88
00:06:24,000 --> 00:06:30,479
squared compared to the 392 millimeter square die of the 3070 which is super

89
00:06:29,580 --> 00:06:35,340
cool but then if it supposedly didn't cost

90
00:06:33,419 --> 00:06:40,740
any less why the heck did they even bother well there are two reasons for

91
00:06:37,560 --> 00:06:43,919
that efficiency and then probably more

92
00:06:40,740 --> 00:06:45,479
importantly the RTX 4090 but let's talk

93
00:06:43,919 --> 00:06:51,840
efficiency first for the amount of performance that the

94
00:06:47,400 --> 00:06:55,259
RTX 470 delivers it sips power it has a

95
00:06:51,840 --> 00:06:56,880
tgp of 200 Watts but NVIDIA says that

96
00:06:55,259 --> 00:07:02,220
the average gaming power draw is more like 184 and in our testing that was

97
00:06:59,819 --> 00:07:06,680
basically bang on which is darn impressive given that the performance

98
00:07:03,960 --> 00:07:12,479
comparable RTX 3080 Drew an average of 296 Watts during the same test also a

99
00:07:10,620 --> 00:07:19,259
good sign is that the maximum recorded wattage of the 4070 was just 233 watts

100
00:07:15,960 --> 00:07:21,120
compared to a whopping 412 for the 3080.

101
00:07:19,259 --> 00:07:24,539
so not only will you be saving money on your monthly Energy bill but you also

102
00:07:22,919 --> 00:07:28,680
don't have to worry about a power Spike sending your aging power supply into the

103
00:07:26,460 --> 00:07:32,819
shadow Realm now you might be hoping that this means there's a lot of

104
00:07:30,000 --> 00:07:36,599
Headroom for overclocking but sadly it looks like NVIDIA has really locked this

105
00:07:35,039 --> 00:07:42,180
card down and we weren't able to meaningfully improve performance by tinkering with any of the knobs and

106
00:07:39,660 --> 00:07:46,979
dials I don't think that the sufficiency gain is why NVIDIA decided to go tsmc 5n

107
00:07:45,180 --> 00:07:51,060
though and it might actually have been better for consumers if they just took

108
00:07:49,139 --> 00:07:55,740
everything they learned about the mature eight nanometer Samsung node and then

109
00:07:53,280 --> 00:08:00,840
made better gpus with it kind of like what they did back in 2014 using the

110
00:07:57,960 --> 00:08:07,800
same 28 nanometer node for both the GTX 770 and 970. the issue though is if they

111
00:08:04,800 --> 00:08:09,419
had done that the RTX 4090 simply

112
00:08:07,800 --> 00:08:13,860
wouldn't have been feasible it would have drawn too much power and it would

113
00:08:11,280 --> 00:08:18,120
have given AMD an opportunity to have a competitive Flagship which would be

114
00:08:15,479 --> 00:08:22,379
utterly unacceptable the 4090 is a true generational leap

115
00:08:20,520 --> 00:08:26,220
coming in at just seven percent more cost than the 3090 while being over 50

116
00:08:24,720 --> 00:08:31,199
faster so then why the heck isn't that true of the rest

117
00:08:29,520 --> 00:08:36,060
of the lineup NVIDIA well because from their point of view

118
00:08:33,599 --> 00:08:41,700
they did give you a generational performance leap but through AI rather

119
00:08:38,760 --> 00:08:47,760
than raw Grunt and In fairness they aren't entirely wrong here by turning on

120
00:08:44,159 --> 00:08:51,300
dlss3 frame generation the RTX 4070 is

121
00:08:47,760 --> 00:08:53,820
able to easily dispatch the RTX 3080 and

122
00:08:51,300 --> 00:08:58,740
even deliver performance about on par with the 4080 before its AI assistance

123
00:08:57,480 --> 00:09:05,760
is kicked in now frame generation may increase input

124
00:09:01,920 --> 00:09:08,160
latency but NVIDIA reflex does go a long

125
00:09:05,760 --> 00:09:13,200
way towards bringing it back down is it as good as Simply Having a faster GPU

126
00:09:10,800 --> 00:09:17,880
and not using any of these tricks that's a topic for another video sadly

127
00:09:15,959 --> 00:09:20,640
our Labs team really does need to do things like sleep even if it's in their

128
00:09:19,740 --> 00:09:27,420
tents adding to NVIDIA's argument currently there are dozens of games that support

129
00:09:24,540 --> 00:09:32,100
DLS S3 and the list is rapidly expanding so you might as well just turn it on but

130
00:09:30,000 --> 00:09:37,320
detracting from NVIDIA's argument there are also tens of thousands of games that

131
00:09:34,800 --> 00:09:43,320
do not support dlss3 and will perform better on the 6800 XT making the 4070s

132
00:09:41,100 --> 00:09:49,200
position more precarious is the fact that pricing of 30 70s and 3080s on eBay

133
00:09:46,500 --> 00:09:52,680
has been in free fall Pro tip by the way don't use buy it now because auctions

134
00:09:51,060 --> 00:09:58,620
have been completing literally a hundred dollars lower now it's worth noting that the 4070 has

135
00:09:56,640 --> 00:10:03,480
some non-gaming generational advantages it's new so it'll have driver support

136
00:10:01,680 --> 00:10:08,100
for longer and it comes with a warranty and it features av1 encoding and

137
00:10:05,580 --> 00:10:12,120
decoding which is genuinely great and as a bonus this thing has been shockingly

138
00:10:10,500 --> 00:10:16,260
stable for being so early in its life cycle AMD could learn a thing or two

139
00:10:13,860 --> 00:10:20,220
this is a really good product but that doesn't mean that we are unreservedly

140
00:10:17,940 --> 00:10:24,839
happy about it NVIDIA is whining about how they simply can't pass along any

141
00:10:22,980 --> 00:10:29,220
silicon cost savings just doesn't hold a ton of water

142
00:10:26,880 --> 00:10:34,740
development costs for new process nodes have increased substantially but I also

143
00:10:32,339 --> 00:10:38,040
think that NVIDIA does have some room to move here if everyone else's pricing is

144
00:10:36,660 --> 00:10:43,380
anything to go by one day Arc will save us

145
00:10:41,339 --> 00:10:48,360
but it is not this day this day I'm going to tell you about our

146
00:10:46,019 --> 00:10:52,860
sponsor vessi it's April and you know what that means rain that's where Vesti

147
00:10:51,120 --> 00:10:57,480
comes in they claim their shoes are 100 waterproof keeping your feet dry during

148
00:10:54,959 --> 00:11:00,540
those unexpected downpours take them on and off with ease vessi has designed

149
00:10:59,160 --> 00:11:04,500
their shoes to be stretchy and convenient to wear their slip-ons yes

150
00:11:02,459 --> 00:11:08,160
with a wide range of styles available you're sure to find a pair that fits

151
00:11:06,420 --> 00:11:12,660
your lifestyle even better their shoes are made from cruelty free products

152
00:11:10,140 --> 00:11:15,240
right down to the glue no more checking the weather right before you head out a

153
00:11:13,920 --> 00:11:18,959
pair of Besties will ensure your feet are taken care of rain or shine snow or

154
00:11:17,339 --> 00:11:22,320
Grime there is always an occasion to throw on a pair of bessies not to

155
00:11:20,339 --> 00:11:27,300
mention their styles are modern and trendy AF treat your feet check out

156
00:11:24,480 --> 00:11:31,560
Bessie at vessey.com Linus Tech tips and use code lines Tech tips for 15 off if

157
00:11:30,180 --> 00:11:35,579
you guys enjoyed this video you might enjoy our coverage of the CPU that you

158
00:11:33,779 --> 00:11:42,019
should pair with really any of these gpus the 7800x 3D it is flipping unreal

159
00:11:39,540 --> 00:11:42,019
for gaming