HomeVideos

HW News - NVIDIA's Face ID Chip, YouTube Loses Lawsuit (lol), AMD 9950X3D2, and Laptop Scams

Now Playing

HW News - NVIDIA's Face ID Chip, YouTube Loses Lawsuit (lol), AMD 9950X3D2, and Laptop Scams

Transcript

990 segments

0:01

Everyone, welcome back to the hardware

0:02

news recap of the week. We're probably

0:04

going to have two for you this week in

0:05

rapid succession because there is a lot

0:07

of hardware news. So, in this episode,

0:09

we're talking about Intel on a tear

0:10

right now. They have their new Arc Pro

0:13

series GPUs that just came out as well

0:15

in addition to the 200 series refresh

0:17

CPUs that were pretty good. I mean, our

0:19

review of the 270K, if you haven't seen

0:20

it, 270K Plus, was pretty positive

0:23

because that CPU is very competitive

0:25

with AMD's 9950X, and it is a lot

0:28

cheaper, like half the price. So, that

0:30

was good to see. It's unfortunate that

0:31

the RAM situation is so bad. It's it's

0:33

really putting a damper on everything

0:35

for Intel having an actual CPU that's

0:37

worth paying attention to. Uh, but

0:39

nonetheless, it was cool to see

0:41

something that actually is competitive

0:43

come out on the market and isn't

0:44

stupidly overpriced. Like some of these,

0:47

you know, the $200 CPU for the 250 KP.

0:49

That's a big change from what we've been

0:50

dealing with in the RAM and GPU

0:52

situation. Also in the news, Nvidia

0:54

wants you. Nvidia's Jensen Juan is all

0:57

in on war.

1:00

That's like what he said.

1:01

>> There's a reason we went to war and I

1:03

believe at the end of the war, Middle

1:05

East will be more stable than before.

1:07

And so if we were there, if we're

1:09

considering it before, we should

1:11

absolutely be considering it after. And

1:13

so I'm 100% in on that.

1:14

>> Jensen, that's that's what he's going to

1:16

do. He's going to export AI

1:20

with the democracy we're exporting.

1:23

Also, Nvidia has facial recognition

1:27

technology that can identify people

1:28

within milliseconds of seeing their

1:30

face, even with face coverings.

1:33

Why? Why, Jensen? Why do you make me

1:36

discuss these topics on my hardware news

1:38

show? Why do you make me do it? Oh,

1:41

there is good news. YouTube lost a

1:44

lawsuit. Let's get started. Before that,

1:46

this video is brought to you by the Mont

1:48

HSO1 and HSO2 cases. The HS1 and HSO2

1:52

cases differ by way of a mesh front or a

1:54

curved glass front. We previously

1:55

reviewed these cases in our intensive

1:57

benchmarking process and found them to

1:59

be overall high build quality with

2:00

thermals comparable between the two

2:02

designs. The rear chamber of the case

2:04

uniquely brings the power supply forward

2:06

and perforates the rear wall of the

2:07

case, allowing for more air movement

2:09

internally while providing a different

2:10

look. The rear of the case uses a fan

2:12

mounting mechanism for two separate fans

2:14

that can be coupled together with a

2:16

plate, improving convenience. The HSO1

2:18

and HSO2 also invert with minimal

2:21

effort, allowing the computer to present

2:23

to the opposite side without needing to

2:25

reconfigure the entire build. Learn more

2:27

at the links in the description below.

2:28

This one's really quick. First item is

2:30

AMD has finally announced its 9950 X3D2

2:34

and this was just a day or so after

2:36

ASRock accidentally I think a second

2:38

time revealed that it was coming but

2:40

then they retracted that press release

2:42

but now it's here. So AMD has announced

2:44

the 9950X3D2. This is a dual CCD dual

2:49

Vcash solution. So unlike the 9950 X3D

2:52

that's currently out, this one adds the

2:55

extra cache also on top of the second

2:57

CCD. So that's even more cash and

3:00

they've also increased the TDP for it.

3:02

Uh it has a total 28 megabytes of onchip

3:06

cache and totality. AMD's official

3:08

product page lists its latest Zen 5

3:10

processor as featuring 16 cores, a 4.3

3:13

GHz base clock, up to 5.6 GHz boost, 200

3:17

watt TDP, and a 192 megaby L3 cache. So,

3:21

that's just the L3 for that number,

3:23

which improves upon the previously

3:25

released 9950X3D's

3:27

128 megabytes of L3 cache. And TDP also

3:30

increases by 30 watts, which will aid in

3:32

maintaining clocks, but of course means

3:33

it'll pull more power. Regarding its

3:35

performance, AMD had some claims for

3:37

firstparty testing. They claimed the

3:39

9950X3D2

3:41

is 5 to 10% improved against the

3:43

original 9950X3D,

3:46

not 2. The funniest part though of this

3:48

announcement is not the announcement

3:50

itself, but the fact that over 80% of

3:51

the comments are things like release the

3:53

FSR4 files, people flooding AMD and

3:56

bullying them to try and get support on

3:58

older devices, which we know is

4:00

possible. The 9950 X3D2

4:03

uh should mostly be interesting in

4:06

situations where one obviously the cache

4:08

is already beneficial. So that's going

4:09

to be gaming for the most part in our

4:10

test suite. We don't see a lot of uplift

4:12

from the extra cache on the X3D

4:14

solutions in our non-gaming tests. Not

4:16

to say it never happens, but for the

4:18

workloads we test, we just we don't

4:20

really see it. So, uh, mostly we see

4:22

that in gaming. Adding more cash. To me,

4:24

it's academically it's really

4:26

interesting. Meaning, not necessarily at

4:27

a product level, but just I I'm curious

4:30

to see which games will show a benefit

4:34

and if we're kind of at diminishing

4:36

returns at some point. It just it seems

4:37

like more cache on the chip is always

4:39

going to be better because you're just

4:40

you're reducing hits out to system

4:42

memory by that much. But probably when

4:43

you're going from something like 128 to

4:45

192 somewhere in there is going to be

4:46

diminishing returns I think. So anyway,

4:48

it'll be really interesting to test. The

4:51

performance versus the price is probably

4:52

going to be a problem because if they're

4:54

claiming 5 to 10%

4:57

the price is going to be kind of rough.

4:59

This likely was fasttracked. It seems

5:01

like maybe to try and bury Intel's

5:04

relatively successful Aerolake refresh.

5:07

Uh it seems like that got buried on its

5:09

own thanks to the RAM pricing which I

5:11

mean I don't blame anyone. But Aerake

5:13

refresh overall fairly exciting for

5:15

Intel. They had a good week where 270 KP

5:18

in particular, the 270K Plus is

5:20

competitive with the 9950X specifically

5:23

in production workloads and in gaming

5:25

workloads, but it's really the

5:26

production ones that are troubling for

5:28

AMD where it's like half the price or

5:30

pretty close to it anyway and uh very

5:33

competitive in performance. So, um this

5:35

may be AMD's way of trying to follow up,

5:37

but it's going to be a higher price

5:39

processor. I really doubt the

5:41

performance will change very much in

5:42

production applications. Obviously,

5:43

we'll test it, but it's probably not

5:45

going to move the needle a lot there.

5:46

So, it doesn't really change much the

5:47

conclusion of the 270 KP. Uh gaming,

5:50

however, will be interesting again from

5:51

an academic standpoint. So, anyway, AMD

5:54

has not yet confirmed the MSRP for the

5:55

solution. They put out a 5-minute video.

5:57

They announced it exists. That is about

5:59

all they did. And they say that this is

6:02

supposed to be coming soon in quarter 2,

6:04

2026.

6:05

Maybe Computex for the announcement.

6:07

That would be first week of June, end of

6:09

May, somewhere in there. Uh, but

6:11

obviously as soon as they actually send

6:13

out samples or we're able to buy one or

6:15

whatever the case may be, we'll do some

6:16

testing on it. Let's start with just the

6:18

computer hardware stuff today. We'll

6:19

start easy and we'll ease into the part

6:21

where Nvidia's Jensen Juan is talking

6:24

about the war in Iran and they're making

6:26

technology that would be extremely good

6:28

for domestic spying. Uh after a strong

6:31

week with the 250K Plus and 270K Plus,

6:33

Intel also launched some GPUs and the

6:37

company has most recently added the B65

6:40

and B70 to its Pro lineup. The ARC Pro

6:42

B65 spec defines 20 XE cores, 20 RT

6:46

units, 197 GPU peak tops, whatever

6:49

that's worth, and a 200 watt TVP or

6:51

total board power. While the B70 lists

6:54

32 XE cores, 32 RT units, and 367 GPU

6:58

peak tops, and a 160 to 290 watt TBP, so

7:01

a wide range here depending on board

7:03

partners. Both cards will offer 32 GB of

7:06

GDDR6 at 68 GB per second memory

7:08

bandwidth and support up to four

7:10

displays. Intel positions its ProGPUs as

7:13

what they call AI workstation cards with

7:15

its marketing claiming quote seamless

7:17

support across multiple GPUs end quote

7:19

and writing that they can quote combine

7:21

the power and memory of multiple GPUs to

7:23

execute AI models in Linux that take 100

7:25

plus GB of VRAM end quote to which

7:27

Nvidia said 100 gigabytes best I can do

7:30

is nine. The Ark Pro B70 is listed for

7:33

pre-order on Newegg listing an April

7:35

24th release date and $950 MSRP at the

7:38

time of writing. Although the B65's

7:40

launch price and date remain unclear at

7:42

the time of writing this. Our friend

7:43

Wendell over at Level One Tech has been

7:45

talking about these cards as well. So,

7:46

if you're interested in these, we'd

7:48

recommend checking out his video. It's

7:49

on his channel. Wendell always does

7:51

great work, so happy to send you his way

7:53

for more info on these cards. Possibly

7:55

do something with them at some point,

7:56

but I'm not sure yet. Uh, he definitely

7:57

is though. Up next, Nvidia's always on

8:00

facial recognition technology. Are you

8:03

ready for probably like 10 minutes of

8:06

great news about Nvidia? because we've

8:09

got it for you, but you should know

8:11

about it. Especially Jensen's feelings

8:14

on war.

8:16

While Nvidia professed its love for

8:18

surveillance company Palanteer at GTC in

8:20

March, Nvidia was already busy creating

8:22

more spy capable technologies earlier

8:25

this year. Nvidia unveiled a new facial

8:27

recognition system at an e conference in

8:29

February. According to Ile Spectrum, the

8:32

engineering outlet reported that Nvidia

8:33

has developed an always on low power

8:35

facial recognition system for use in

8:37

robotics, consumer devices, naming

8:39

laptops as one of them, autonomous

8:40

vehicles, and drones. Great

8:43

technology for drones, facial

8:46

recognition technology, no less. And

8:48

Nvidia's partnered with this guy.

8:50

>> Well, I love the idea of getting a drone

8:52

and having light fentinyl laced urine

8:55

spraying on analysts who tried to screw

8:57

us. At least now the fentinol piss will

8:59

be able to accurately hit my face. Wait,

9:02

no, that didn't that's the editor's cut.

9:04

The new computer vision system can quote

9:06

detect human faces in less than a

9:07

millisecond. End quote. After which

9:09

point, their partner Palunteer can spray

9:11

piss with fentinol in it on them. This

9:13

saves power by storing data locally,

9:15

too. Nvidia engineer Ben Keller

9:17

presented this new system on behalf of

9:18

the company. ILE E Spectrum summarized

9:21

Nvidia's technical presentation,

9:22

writing, quote, according to the

9:24

researchers, this kind of vision

9:25

processing typically requires about 10

9:27

watts, but that's too much power to

9:28

leave a face detection system on

9:30

continuously. The Nvidia system on chip

9:32

uses less than 5 matt with a frame rate

9:34

of 60 fps. The system refreshes to

9:36

process a new image every 16.7

9:38

milliseconds and is only fully powered

9:41

on for 5% of that time with E Spectrum

9:44

citing Keller for that. The quote

9:45

continues, quote, within 787

9:47

milliseconds, the SOC calls on a deep

9:49

learning accelerator to determine

9:51

whether or not a human face is present

9:53

with about 99% accuracy. End quote. Just

9:55

a note, but processing 16.67

9:57

milliseconds for each image is the same

9:59

as the 60 fps rate. Says the same thing.

10:01

Nvidia is currently working on major

10:02

laptop engineering efforts. These

10:03

include their new ARM CPUs that aren't

10:04

announced yet, but are supposed to be

10:06

probably called the N1X, and that'll

10:09

likely be launched uh or announced at

10:10

least at Computex more formally. And

10:13

Nvidia's also with its 4% stake in Intel

10:16

got the RTX chiplet laptop solutions

10:18

coming out in addition to whatever

10:20

networking solutions it's doing and

10:21

enterprise solutions. So they are like

10:24

the war in Iran all in on laptops right

10:26

now and uh this is the perfect time to

10:29

roll out a facial recognition technology

10:31

that can go into them. Much of the SOC

10:33

is powered off by default in Nvidia's

10:35

new system except for a subsystem called

10:37

Alpha Vision. In an abstract from the

10:39

research paper, Nvidia said that Alpha

10:41

Vision is quote always on low power

10:44

subsystem for DNN inference-based vision

10:46

tasks and edge SOC's end quote. And the

10:48

subsystem also doesn't allow external

10:50

access to memory. An Nvidia graphic

10:52

shows that Alpha Vision remains on in

10:54

sleep mode before powering the rest of

10:56

the SOC when it detects a face in active

10:58

mode. Ile E Spectrum said the required

11:01

data is stored locally on the SRAMM and

11:03

that the quote face recognition system

11:06

rushes through its work and then quickly

11:08

puts the SRAMM into a low power sleep

11:10

mode end quote which Nvidia refers to as

11:12

race to sleep. This is also what I

11:14

called the end of my Crimson Desert

11:16

benchmarking process. Nvidia's

11:17

investment in facial recognition

11:19

technology isn't new. Back in 2012 or

11:21

2013 they were already moving on it. In

11:22

fact, I remember going to Nvidia's then

11:25

HQ probably about 2012 or 13 and uh they

11:28

gleefully showed us their facial

11:30

recognition technology where I think the

11:32

demo was showing people walking through

11:34

an airport with a camera system

11:37

autographing and identifying faces and

11:39

that was over a decade ago. Now in July

11:41

2013, Nvidia launched the GeoInt

11:43

Accelerator which the company called the

11:44

quote world's first GPU accelerated

11:46

geospatial intelligence platform end

11:48

quote. In a blog post, Nvidia said the

11:50

platform quote provides defense and

11:52

homeland security analysts with tools

11:54

that enable faster processing of facial

11:56

recognition and surveillance video. End

11:57

quote. And that was 2013. Additionally,

11:59

in April 2016, Hera Security, a company

12:01

specializing in facial recognition,

12:03

talked about how Nvidia GPUs accelerated

12:05

their solutions on the NVIDIA developer

12:07

YouTube channel.

12:07

>> We are focused in law enforcement

12:09

projects. We have to be able to detect

12:12

faces in crowds and we have to solve the

12:14

problem of uh very uh large resolutions.

12:16

Let's say we're talking about 4K

12:18

resolutions. We make use of u GPUs uh to

12:22

go very fast to be able to detect faces

12:25

very fast and to be able to process

12:26

images coming especially from video

12:28

streaming and you have to scan these

12:30

faces and uh in real time.

12:33

>> In a related blog post, Nvidia said her

12:35

securityurities technologies quote make

12:37

it possible to record subjects

12:38

automatically through on the-fly video

12:40

capture and works correctly even when

12:42

the subject is wearing glasses, a hat or

12:44

their face is partially concealed. It

12:46

even works with changes in facial

12:47

expression, difficult lighting

12:49

conditions, and slight rotations of the

12:50

face. End quote. Nvidia announced more

12:52

surveillance company partnerships for

12:53

its AI city initiative later in 2016. In

12:56

a now deleted blog post that we

12:58

recovered via the way back machine,

13:00

Nvidia said its highkey vision, Dahua,

13:02

and Univ partners are quote the top

13:04

three market share leaders for video

13:06

surveillance products. End quote. Nvidia

13:07

also partnered with Sense Time, which it

13:09

described as a quote leader in facial

13:12

and object recognition end quote. The

13:14

type of recognition that these companies

13:15

are leaders in include patents for

13:18

ethnic profiling via facial recognition

13:20

technology.

13:22

Nvidia partnered with a company that

13:25

filed a patent that explicitly named it

13:28

is capable of identifying people by

13:33

ethnicity which only has good uses.

13:37

There's no reason that would ever be

13:39

abused. Notably, Nvidia partner Daha has

13:42

been federally restricted in the past

13:44

due to quote involvement in human rights

13:46

abuses and persistent cyber security

13:48

vulnerabilities and quote per Nebraska's

13:51

attorney general. We also recognized

13:53

sense time from the past Silverlake

13:55

investment prior to its divevestature

13:57

which we covered in our story about

13:59

Jared Kushner and Saudi Arabia investing

14:01

in EA games by which we mean buying it.

14:04

The US government sanctioned Sensime in

14:06

2021 for its connections to quote human

14:08

rights abuse and repression and quote a

14:10

press release from the Treasury

14:12

Department said Sensime quote has

14:13

developed facial recognition programs

14:15

that can determine a target's ethnicity

14:17

with particular focus on identifying

14:19

ethnic Uggers when applying for patent

14:21

applications. Shenzhen Sensime

14:23

Technology Co- Limited has highlighted

14:25

its ability to identify Ueagers wearing

14:27

beards, sunglasses, and masks. End

14:29

quote. The sanctions prevented US

14:30

citizens and companies from investing in

14:32

sensime and required their

14:34

divevestature. Gee, I wonder why Nvidia

14:36

seemingly deleted the blog posts. It

14:39

really has nothing to do with any of

14:40

this. The reason we had to dig it up

14:42

through archive.org because the internet

14:44

never forgets. But anyway, two for two

14:45

on partnering with companies accused of

14:47

human rights abuses specifically through

14:50

surveillance and facial recognition

14:52

technology which Nvidia is accelerating

14:54

the development of per these

14:55

announcements. So, uh, Nvidia openly

14:58

showcases partnerships with companies

15:00

like Palanteer. Now,

15:01

>> Palanteer, Palanteer, Palanteer.

15:03

>> And if you want to learn more about this

15:04

stuff, we have a deep dive video on

15:06

allegations that Palanteer has enabled

15:07

predictive policing in the past, aka

15:10

pre-rime arrests, some minority report

15:13

[ __ ] and also generally what we think

15:16

is war profitering from the company.

15:17

Now, Nvidia has repeatedly emphasized

15:19

its love of Palunteer, and Jensen Juan

15:22

has turned into a a bit of a war pig as

15:24

he's further embedded himself with

15:27

companies like Palanteer and the

15:28

government. And decision to support

15:30

surveillance companies and invest in

15:31

facial recognition technology isn't new.

15:33

Obviously, that's why we went back to

15:34

2013 to tell you about it. Uh this is

15:36

despite clear dangers or maybe because

15:39

of clear dangers that are created by

15:41

this type of technology deployed

15:43

indiscriminately or in some cases

15:45

discriminately on mass. Wrongful arrests

15:48

done through facial recognition happen

15:50

all the time now. And facial recognition

15:52

can be prone to biases like racial

15:54

biases and otherwise. Nvidia was right

15:56

when it said it's not a gaming graphics

15:58

company anymore because apparently now

15:59

it's a surveillance state company. You

16:01

got to keep the bubble inflated somehow.

16:02

I guess there's the quote QuiGon Jen.

16:05

There's always a bigger fish and in this

16:07

case that fish is the government and

16:09

they need to keep the bubble inflated.

16:11

The second of the two major Nvidia

16:12

stories before we get back to other

16:14

stuff is Nvidia CEO Jensen Juan

16:15

indicated he sees opportunity in the

16:17

Middle East for AI market opportunity

16:20

there and uh he thinks that war will

16:23

bring and this is not me editorializing

16:25

it. This is what he said. thinks war

16:27

will bring stability in the Middle East

16:30

and which I think I I'm pretty sure that

16:33

George Orwell literally wrote those

16:35

words when he said quote war is peace

16:38

end quote in 1984 but okay sure after

16:41

his GTC keynote Nvidia CEO and man who

16:44

personally wrestled a crocodile for this

16:46

leather jacket Jensen Juan had some more

16:49

to say and he returned to one of his

16:51

beloved pastimes which is rejoicing with

16:54

podcasters fashion episode this week.

16:57

We've preempted the weekly show. And

17:00

there's only three people we preempt the

17:02

show for. President Trump, Jesus, and

17:06

Jensen.

17:07

>> I

17:08

What even What even What is this? What

17:12

What are we doing? Jensen joined the

17:14

All-In podcast to discuss everything

17:16

under the AI umbrella, including Nvidia

17:19

strategy, robotics, open source,

17:20

geopolitics, data centers, and global

17:22

supply chains. Jensen ran through many

17:24

of his favorite talking points. I can

17:26

prove it that the $50 billion factory

17:28

will generate for you the lowest cost

17:31

tokens. We call that omniverse. It

17:33

wasn't until chat GPT put a user

17:37

interface around it, made it easy for us

17:38

to use that generative AI took off. Yes,

17:41

every job will be will be transformed.

17:43

Um, some jobs will be eliminated.

17:45

President Trump wants American industry

17:48

to lead. I would love that the American

17:50

tech stack is 90% of the world. It's

17:52

very likely that you're going to have a

17:54

lot of pain and suffering. So, you

17:55

better enjoy it.

17:56

>> Nvidia CEO parided his usual AI

17:58

propaganda. And in one notable segment,

18:01

an interviewer who just finished dining

18:03

on a most exquisite boot at the Jensen

18:06

restaurant asked Jensen about money and

18:08

specifically how much Nvidia is spending

18:10

on tokens for its engineering teams. In

18:12

response, Jensen Juan said that

18:14

engineers should be spending a token

18:16

budget worth approximately half of their

18:19

base pay. So, are you spending in Nvidia

18:22

a billion2 billion dollars on tokens

18:24

from your engineering team right now?

18:25

>> We're trying to Let me give you the

18:26

thought experiment. Let's say you have a

18:28

software engineer or AI researcher and

18:30

you pay them $500,000 a year. We do that

18:33

all the time.

18:34

>> Yeah.

18:34

>> Okay. This is happening all over the

18:36

time. Um, that $500,000 engineer at the

18:39

end of the year, I'm going to ask them

18:41

how many token how much did you spend in

18:43

tokens? And that person said, $5,000, I

18:46

will go ape something else.

18:48

>> Yes.

18:48

>> Right. If that if that $500,000 engineer

18:51

did not consume at least $250,000 worth

18:54

of tokens, I am going to be deeply

18:57

alarmed.

18:58

>> In other words, man selling shovels says

19:00

you need more shovels. News at 11. But

19:03

Jensen keeps making loftier promises

19:05

about how AI will drive growth. He's

19:07

even equating GPU compute power to GDP

19:11

these days as at a Morgan Stanley event.

19:13

So the first thought is that compute

19:17

equals revenues. Now the big idea of

19:20

course compute equals GDP that we also

19:23

know. Compute equals a country's GDP.

19:28

>> But talking about how they think you

19:29

need to spend half of a base salary on

19:32

tokens and talking about developing

19:34

things that have potential for global

19:36

surveillance applications. It's it's

19:38

kind of dark. Let's move into a a

19:40

lighter topic. I was also asked you know

19:42

given what's happening in the Middle

19:44

East is that an area where we believe

19:47

that we can expand artificial

19:49

intelligence too. Um I believe that

19:52

there there's a reason we went to war

19:54

and I believe at the end of the war

19:56

Middle East will be more stable than

19:58

before and so if we were there if we're

20:01

considering it before we should

20:02

absolutely be considering it after and

20:04

so I'm 100% in on that.

20:07

>> Okay. All right.

20:09

not a lighter topic. To be clear, Jensen

20:11

Juan says he's all in on war in Iran. He

20:14

thinks, but did not specify that there's

20:16

a reason for it. We assume the reason is

20:19

I like money,

20:20

>> maybe selling weapons, things of that

20:22

nature. A lot of money to be made in

20:24

war, but he also sees opportunity there,

20:27

he says. Best of all, though, was when

20:29

he said that he thinks this will bring

20:31

stability to the Middle East. Yes. Yes.

20:33

Jensen, tell us more about your

20:35

opinions, please, on war, so that you

20:37

put all the rest of us into a position

20:39

where we also have to comment on your

20:41

opinions on them. Uh, clearly, this is

20:43

the time that it'll work. War is, in

20:46

fact, peace. This follows Iran

20:48

reportedly declaring Nvidia a legitimate

20:50

military target in the region alongside

20:52

other companies fueling AI. Nvidia's

20:55

been pro-war for a while now, we think.

20:57

In fact, we've got a montage for you

20:59

showing just how desperately they're

21:01

trying to integrate with the

21:02

military-industrial complex.

21:03

>> The wealthier our nation,

21:06

the more we can fund the mightiest

21:08

military on the planet.

21:10

>> I'm happy that that uh our military is

21:14

going to use AI technology for defense.

21:16

>> One of the things that that we just

21:18

always have to remember is that the

21:20

world's mightiest military uh is

21:22

supported by the world's mightiest me

21:24

economy. And so the wealth that we

21:27

generate um brings jobs home, creates

21:30

prosperity in the United States, um

21:32

provides for tax revenues, and

21:34

ultimately funds the mightiest military

21:37

on the planet.

21:37

>> That we want America to be the

21:39

wealthiest country so that we can fund

21:40

the mightiest military. And I I think

21:43

that that is our way of contributing to

21:45

national security.

21:46

>> And now with Iran reportedly declaring

21:48

Nvidia, I can't believe I'm even saying

21:50

these words. This is like

21:52

should be should be talking about gaming

21:54

GPUs. Anyway, with Iran reportedly

21:56

declaring Nvidia a legitimate military

21:58

target, Jensen, it seems you've gotten

22:01

your wish. Good job. You're finally a

22:03

real boy. Now you're part of the

22:05

military-industrial complex.

22:07

Congratulations. I'm sure all your

22:08

employees in the region are thrilled.

22:10

Iran has previously targeted AWS data

22:12

centers in the UAE and Bahrain with

22:14

drone strikes causing cloud service

22:16

disruptions in the region. AWS said the

22:18

strikes have quote caused structural

22:20

damage, disrupted power delivery to our

22:22

infrastructure, and in some cases

22:24

required fire suppression activities

22:26

that resulted in additional water

22:27

damage. End quote. We emailed Nvidia's

22:29

Matt Webblin, who previously refused to

22:31

get on a recorded call to talk about how

22:33

they were uh attempting, we thought, to

22:36

change our editorial direction on the

22:37

LSS, and asked him about specifically

22:40

what Jensen Juan meant when he said he

22:43

thinks there's a reason for war, and

22:46

specifically what that reason was. We

22:48

also asked them uh why it would bring

22:51

stability this time and more

22:53

specifically and pointedly if Nvidia is

22:55

selling any military of any nation

23:00

technologies and AI services that would

23:02

be applicable in this war and they did

23:06

not respond. So

23:09

I'm going to go with yes. I don't know.

23:11

I think they didn't respond. I just

23:13

that's I just have to I just have to

23:15

assume. But we don't need NVIDIA PR and

23:17

communications to confirm that Nvidia

23:19

has interest in the Middle East because

23:21

NVIDIA has sold thousands of data center

23:23

GPUs in the region. Last May, Nvidia

23:25

announced several AI initiatives with

23:26

the Saudi Arabian government, including

23:28

a partnership with Humane to deploy

23:30

18,000 GB300 chips. Humane is also owned

23:32

by Saudi Arabia's public investment

23:34

fund, which is the same fund that

23:35

partnered with Jared Kushner, who is the

23:37

son-in-law to Donald Trump and married

23:38

to Ivanka Trump, to purchase EA games.

23:41

Further, in November 2025, Microsoft

23:43

said it had received export licenses

23:45

from the Commerce Department from both

23:46

the Biden and Trump administrations to

23:48

ship thousands of NVIDIA GPUs to the

23:50

UAE, including the A100, H100, H200, and

23:53

GB300. The point is, Nvidia is expanding

23:55

its presence in the region. The

23:57

company's planning to build new research

23:58

and development campuses in Israel.

24:00

According to the Times of Israel, the

24:02

outlet said, quote, "The campus is

24:03

expected to accommodate up to 8,000

24:05

employees and will significantly expand

24:07

Nvidia's presence in northern Israel."

24:08

End quote. Nvidia understandably wants

24:10

to protect its assets and hopefully its

24:13

people. Maybe not other people, but at

24:15

least its people, I would hope.

24:18

But it is uncommon I I can't say I've

24:21

ever heard it in this industry for a

24:25

tech CEO, at least a computer tech CEO

24:28

who I would be familiar with, to

24:30

explicitly say that they support war.

24:32

That was I mean that's there's not

24:34

really a lot of ways to read what he

24:36

said and uh that's the way we're reading

24:38

it. This is it's abnormal for someone

24:42

for a tech CEO to just straight up be

24:43

like, "Yep, we support that." And no

24:48

further elaboration on it whatsoever. Uh

24:50

but militarism is not new to Jensen

24:52

Juan. He's ratcheted up his military

24:53

interest for a while now and he is uh

24:56

dragging everyone else with him. But

24:57

let's get on to some actually lighter

24:59

hardware topics like um

25:03

a laptop scam. That's a that's actually

25:06

that's a big relief. That's just a

25:08

simple scam, you know, just like the the

25:10

old stuff, the good old days when you

25:12

got ripped off for buying something.

25:14

There's you know, it's we should just

25:17

let the scammers do it at this point for

25:18

entertainment. I think Notebook Check

25:20

called out laptop manufacturer Chewy for

25:22

selling devices with older, worse

25:24

performing CPUs than advertised. I mean,

25:26

seriously, actually, what a relief this

25:29

story is. All right, let's get into our

25:31

uh our icebreaker story. Our our relief

25:34

story. It's like comedic relief, except

25:36

it's not funny really, but it's okay,

25:39

let's continue. Despite Chewy marketing

25:42

its Corebook X and Corebook Plus as

25:44

powered with the Ryzen 5 7430U, it

25:46

actually shipped them with a 2 years

25:48

older Ryzen 5 5500U, which notebook

25:50

check claims has an up to 20% worse and

25:53

quote performance in its benchmarking

25:54

tests. Oddly, the publication also found

25:56

that the 5500U presents as a 7430U in

26:00

BIOS and software. So, it's

26:01

misrepresenting itself. Notebook check

26:03

noticed an inconsistent L3 cache, boost

26:05

clock, and code name, which helped tip

26:08

them off to the scam. The only way to

26:09

truly identify the CPU would be to tear

26:11

down the laptop, which of course they

26:13

did next. AMD commented on the situation

26:15

posted via HKC in Hong Kong and machine

26:19

translated. We'll paraphrase sections of

26:21

the auto translation that are machine

26:22

errors to make it clearer. They said

26:24

quote with some paraphrasing recently

26:25

Chewy has come to our attention which

26:27

has misrepresented AMD Ryzen 55500U

26:30

products as Ryzen 57430U AMD has never

26:32

authorized confirmed or acquiesed to

26:34

this behavior in any way nor has it

26:36

participated in the labeling or

26:37

promotional decisions of related

26:39

products and is completely unaware of

26:40

this matter. AMD has always placed great

26:43

importance on the accuracy and

26:44

transparency of product information and

26:46

is committed to maintaining a fair and

26:47

orderly market environment and

26:49

protecting the legitimate rights and

26:50

interests of our users. Our company has

26:52

taken this matter very seriously and

26:54

reserves the right to pursue legal

26:55

action against the relevant parties. End

26:57

quote. Interestingly, Chewy has since

26:59

edited its Corebook Plus and Corebook X

27:01

listings to replace any mentions of the

27:03

Ryzen 57430U with instead Ryzen 5

27:07

processor 5 days after AMD's response.

27:09

Chewy released its first statement

27:11

claiming, quote, "Due to a production

27:13

error, a limited number of CoreBook X

27:15

and Corebook Plus units were assembled

27:18

with incorrect processors." Got to

27:19

interrupt the quote here for a second,

27:21

but to also misrepresent those

27:23

processors in software seems

27:26

intentional. There's not really a lot of

27:29

ways that can accidentally happen, but

27:30

let's continue. Quote, "If you have

27:32

received a device that does not match

27:33

the specifications you ordered, please

27:35

return it for a full refund. You may

27:37

initiate the return through your

27:38

original purchase channel or contact our

27:40

official support team." They give an

27:42

email address for assistance. End quote.

27:44

Chewy's return guidelines we looked up.

27:46

note, quote, "Please ensure the device

27:48

is in original condition with all

27:49

accessories included." End quote. With

27:51

the return period ending May 31st, 2026.

27:55

Now, to jump it again, that's not how it

27:57

works. It's one thing to have a return

27:59

window for a product you bought and then

28:01

you decide you don't want it or

28:03

whatever. Uh, but

28:06

you can't just you can't scam someone

28:09

and be like, "There's a limited window

28:10

where you can return." Like, that's not

28:12

how that works. The window is called the

28:14

statute of limitations and it's for

28:16

scamming them. So I it would be better

28:19

to just extend the window into uh I

28:21

don't know if you bought it and we lied

28:23

or misrepresented it then you get your

28:25

money back because please don't sue us.

28:27

Thank you. That seems more reasonable

28:28

but okay sure May 31st. Chewy statement

28:31

doesn't mention the Ubox 7430U mini PC

28:34

that HKPC claimed was also affected. It

28:37

doesn't explain why or how the CPU was

28:39

mislabeled in BIOS and software. It

28:41

doesn't describe how to identify the

28:43

processor, which is especially rich

28:44

considering it allegedly threatened

28:46

notebook check legally for telling

28:48

people how to identify the scam. And it

28:50

only gives affected customers about two

28:51

months to return the device for a

28:53

refund. Good company. But speaking of

28:54

good companies,

28:56

Meta and YouTube. Now, California jurors

29:00

ordered Meta and Google to pay a woman

29:03

$3 million in compensation after

29:05

establishing that quote, "Meta and

29:07

Google's YouTube were to blame for the

29:08

depression and anxiety of a woman who

29:10

compulsively used social media as a

29:12

small child." End quote. As reported via

29:14

NPR. In the hearing, which CNN notes is

29:16

quote, "The first of more than 1500

29:18

similar cases," end quote, making the

29:19

verdict especially noteworthy as a

29:21

precedent, the plaintiff cited numerous

29:23

mental health issues caused by the

29:25

platforms. Regarding the ruling, NPR

29:27

explains, quote, "The jury was not

29:28

tasked with deciding whether Meta and

29:30

Google had created Kayle's mental health

29:32

wos, but rather if her compulsive social

29:35

media use was a substantial factor in

29:38

her struggles, and if the defective

29:39

design of the platforms was the direct

29:41

cause of the distress," end quote. Now,

29:43

to us, the design is not defective. That

29:47

is what I mean, they're they know what

29:48

they're doing. They're building social

29:51

media platforms to cause people to get

29:54

the quick hits. The game is to keep

29:57

people on the platforms as long as

29:58

possible. Look, we're uploading to

30:00

YouTube. I know that how this [ __ ]

30:01

works. Like, it's pretty transparent. I

30:03

think everyone watching knows how these

30:05

social media platforms work at this

30:07

point. Uh, and it's the same for

30:09

Facebook and Instagram and Tik Tok and

30:10

everything else where regardless of who

30:14

uploads what, the platforms are built

30:17

specifically to trap people in a loop

30:19

and some of them put them in an echo

30:20

chamber to help perpetuate that loop.

30:22

And seeing a a case about this is big.

30:25

This is good news. If any of these

30:28

companies do anything to try to change

30:31

how they operate for fear of fines,

30:33

unfortunately, it seems unlikely. the

30:35

money you make is going to likely be a

30:37

lot larger than the fines. But still,

30:39

one can hope that the system works as it

30:41

should, uh, just like these platforms

30:43

are working as they were designed to

30:45

work. According to PBS, quote, Meta

30:47

consistently argued that Kaye had

30:49

struggled with mental health separate

30:50

from her social media use, often

30:52

pointing to her turbulent home life and

30:54

quote, "Google argued that quote, this

30:56

case misunderstands YouTube, which is a

30:59

responsibly built streaming platform,

31:01

not a social media site." End quote.

31:04

That's right. They're arguing semantics.

31:06

The best possible argument when you have

31:08

no other arguments to argue. Google is

31:12

saying, "Nuhuh. You're social media." I

31:16

like, "What's the play here?" Like, it's

31:18

no, no, no. It's what? We did all of

31:20

that stuff. Okay. But we're not a social

31:23

media platform. We're a responsibly

31:25

built streaming service. Go [ __ ]

31:27

yourself, YouTube. It's a social media

31:29

platform. I mean, come on. You know,

31:31

like you got posts. They have community

31:34

things now. You can create like a forum

31:36

basically on your YouTube channel. Hey,

31:38

come on. It's the same thing. The ruling

31:41

was in favor of the plaintiff with NBC

31:43

reporting that Meta's assessing its

31:45

legal options while Google is already

31:47

planning to appeal. There's a lot of

31:49

other stories, too. So, we have a second

31:51

hardware news episode coming up. Um,

31:54

we've got some more. I mean, AMD Lisa

31:57

Sue is talking about how she's joined

31:59

the, as I'm calling it, the Council of

32:00

Corruption. We'll talk about that. Uh,

32:03

Epic Games had massive layoffs. We're g

32:04

to find a bunch more hardware news as

32:06

well. I'll put it at the front of the

32:07

episode, just like we did with this one

32:09

as the as the icebreaker. Um, you know,

32:13

I don't know. It's

32:15

It's a mess. This industry is is [ __ ]

32:19

Like, it's just in terms of the

32:20

corruption and uh it's really the Nvidia

32:25

stuff. And now, you know, we'll talk

32:26

about Lisa Sue in the next one like we

32:28

did recently in the her special video,

32:30

but in this episode specifically, SIN

32:32

video stuff where I just read it in the

32:34

news. I'm like, what the [ __ ] happened?

32:36

Like, what is this? It's just I remember

32:39

doing hardware news at the house before

32:40

we moved to the first office and then

32:42

moved here where uh the topics it's like

32:46

you would never talk about this stuff

32:48

with Nvidia. It was whatever gaming GPU,

32:50

whatever driver they launched. They

32:51

weren't even [ __ ] them up yet. That

32:53

came later. That was a new innovation.

32:55

uh updates to things. This was pre-RTX

32:58

as well. So there's updates to things

32:59

like I think it was GameWorks at the

33:01

time, stuff like that. AMD with shitty

33:04

drivers and now that's turned around. So

33:05

that's at least good news. Intel's

33:07

turned around. That's good news for CPUs

33:09

that is. I mean, but you know, it's just

33:12

I don't understand

33:15

how we got here as an industry where now

33:18

you have the gaming GPU guy who one

33:20

describes GeForce as his greatest

33:22

marketing campaign ever and says that

33:24

now GeForce gamers can be proper

33:26

customers by being customers for data

33:28

center products. Something he actually

33:30

said, by the way, that's not I'm not

33:31

editorial. That's in the GTC keto. Go

33:34

watch it if you want. But he's saying

33:36

that and then he's turning around and

33:38

he's talking about war and his belief in

33:40

it and it's just like what what

33:42

happened.

33:45

I don't It's I don't I don't get it. I I

33:47

mean well I get it conceptually I can

33:49

follow it. Logically I can follow it

33:50

exactly. I've been here so like I get

33:53

it. I just I don't understand.

33:56

I don't know man. All right. Okay. We'll

33:59

close this one out. Uh we had a lot of

34:01

reviews this past week. So there's some

34:02

good stuff. We got the two CPU reviews

34:04

up. Really interesting stuff. Clean of

34:06

all the [ __ ] It's just straight up

34:08

benchmarks. It's fun. You'll like it.

34:09

It's like the old times before the

34:11

companies did all this other BS. Uh 270

34:13

KP in particular was an interesting one

34:15

because that one I think super

34:16

competitive with AMD. The Nocta case

34:19

review we did a I think about 10 days

34:21

ago or something. That's super

34:22

interesting as well because that's got

34:24

some of our fan testing charts in it. So

34:26

if you've been looking for those, really

34:28

interesting stuff. And then our Crimson

34:29

Desert benchmarks had our first sort of

34:32

real animation error, as we're renaming

34:34

it, simulation time error charts with

34:36

actual like head-to-head comparisons of

34:38

GPS. It is awesome. So, if you want just

34:40

technical stuff, you got plenty of it on

34:41

the channel. Go check it out. And we'll

34:43

have another hardware news coming up

34:44

shortly with some hardware news that is

34:47

pure hardware and some hardware news

34:48

that is these companies doing the stuff

34:50

they do. Thanks for watching. Subscribe

34:51

for more. Go to store.gameex.net.

34:53

patreon.comdexes and we'll see you all

34:55

next

Interactive Summary

This hardware news recap covers several significant developments: Intel's competitive new Arc Pro series GPUs and 200 series refresh CPUs, which are noted for being cheaper than AMD's counterparts despite RAM issues. AMD announced its 9950X3D2 CPU, featuring increased cache and TDP, with claims of 5-10% performance improvement over its predecessor. A major focus is on Nvidia's expanding involvement in facial recognition and surveillance technology, including past partnerships with companies linked to human rights abuses, and CEO Jensen Huang's controversial stance on war and AI's role in the Middle East. The video also highlights a laptop scam by Chewy, which sold devices with older, misrepresented CPUs, and a landmark lawsuit where Meta and Google's YouTube were ordered to pay compensation for contributing to a woman's mental health issues due to compulsive social media use.

Suggested questions

6 ready-made prompts