whether it's the six titan views we had

for the six editor's project,

three GV100 Quadras for 12K Ultrawide gaming,

or even this unreleased mining GPU,

the CMP 170 HX.

There are not a lot of cards out there

that we have not been able to get our hands on

in one way or another,

except for one.

Until now,

the Nvidia A100,

this is their absolute top dog,

AI enterprise high performance compute,

big data analytics monster,

and they refused to send it to me.

Well, I got one anyway Nvidia,

so, deal with it.

Just like everyone's gotta deal with my segues,

smart deploy provides out of the box windows imaging support

for over 1,500 computer models.

You can deploy one windows image to any hardware model

with ease and you can get free licenses worth over $500

at smartdeploy.com/linus.

(upbeat music)

The first two questions on your mind,

are probably why we weren't able to get one of these.

and what ultimately changed

that resulted in me holding one in my hands right now.

The answer to the first one

is that Nvidia just plain

doesn't seed these things to reviewers

and at a cost of about $10,000.

It's not the sort of thing that I would just,

you know, buy.

'Cause I got that swagger.

You know what I'm saying?

As for how we got one,

I can't tell you.

And in fact,

we even blacked out the serial number

to prevent the fan who reached out offering to get us one,

from getting identified.

This individual agreed to let us

do anything we want with it.

So you can bet your butt,

we're gonna be taking it apart.

And all we had to offer in return

was that we would test Ethereum mining on it,

send a shroud, that'll allow 'em to actually cool the thing

and reassemble it before we return it.

So let's compare it really quickly to the CMP 170 HX,

which it is the most similar card that we have.

It's the silver metal

and it's not ripped for my pleasure.

Regrettable. - [Jake] Alright.

- And we actually have one other point of comparison.

This isn't a perfect one.

This is an RTX 3090.

And what would've been maybe more apt

is the Quadro or rather they dropped the Quadro banding.

But the A6000.

Unfortunately that's another really expensive card

that I don't have a legitimate reason to buy

and Nvidia wouldn't send one of those

for the comparison either.

So the specs on this are pretty similar.

We're gonna use it as a standin'

since we're not really looking

at any workstation loads anyway.

So the A100.

This is a 40 gigabyte card.

I'm gonna let at that sink in for a second.

And the craziest part,

is that 40 gigs is not even enough

for the kinds of workloads

that these cards are used to crunch through.

We're talking enormous data sets

to the point where this 40 gig model,

is actually obsolete now,

replaced by an 80 gig model.

And these NV Link bridge,

connectors on the top here,

let's go ahead and pull these off.

These, there we go,

are used to link up multiples of these cards

so they can all pull memory

and work on even larger data sets.

Now the diet, the center of it,

is a seven nanometer TSMC manufactured GPU

called the GA 100.

We're gonna pop this shroud off.

We're gonna take a look at it.

It has a base clock of just 765 megahertz,

but it'll boost up to fourteen ten.

That memory runs

at a whopping one and a half terabytes a second

of bandwidth on a massive

5,120 bit bus.

It's got 6,912 CUDA cores

and, what is it?

250 watt TDP.

Woooh.

She's packing.

- [Jake] Oh, you're just going right for it.

- I'm going right for.

- [Jake] Oh geez.

- This is Linus tech tips.

- [Jake] And basically every part of this

is identical to the CMP card.

- It kinda looks that way.

I mean the color's obviously different.

- Yeah, but it looks like the clamshell

is two pieces in the same manner.

There's no display outputs.

The fins look the same.

- Now here's something the CMP card specifically

didn't even contain the hardware for video in code,

if I recall correctly.

- Yeah, this doesn't have anything.

- Okay, so it's not that it was fused off.

It's just plain not on the chip.

- Not on GA 100, yeah.

- Okay but,

- GA102, which is like 3090.

- Yes.

- Does have it. - Ooh.

- And A6000.

- Okay, you ready?

- Oh God!

So yeah. - Hey.

- It's like exactly the same on the inside.

Same junk power connector.

- Wow.

That is super junk,

check this out guys.

It uses a single eight pin EPS power connector,

which you might think is a PCIE power connector.

So here, look, I'll show you guys.

This is an eight pin,

like normal GPU connector,

but watch, cannot go in.

But if we take the connector

out of our CPU socket on the motherboard,

There you go although,

the clips are interfering a little bit.

I mean, what the heck is going on here,

ladies and gentlemen?

- You need more power.

- Yeah exactly.

- So you can combine two PCIE connectors into that.

- [Andy] Can't remember how to get it outta here.

I see the fingerprint of the technician

who assembled the card though.

- I think we have to unclip this part first.

Oh, there's a little screw, right?

- There's a little screw.

- Haha, third type of screws.

- [Andy] Yourself. - Didn't see that one, nerd.

- [Andy] You're a nerd.

- [Jake] Your face is a nerd.

- [Andy] Your but nerd.

- [Jake] Whoa.

- It's not coming off, Jake.

- What? You gotta like tilt it out, buddy.

Whoa, whoa, whoa.

Don't pull the cooler off.

- See?

It's like it's caught back here.

- Hey ho.

Hey, how you doing?

- Jesus.

- Stressful.

Look, maybe if we break it,

you'll actually have to buy one.

- I don't wanna buy one.

That's not the goal. - What?

- I thought you put your hand up for a high five.

I was like, "well, what are you talking about?

I don't want to buy one."

- Why not?

Whoa, what is going on here?

You see that?

- It looks like there was a thermal pad there or something,

but there isn't, its like greasy.

- It actually,

no, look at it closer.

It's not greasy.

It's, you see how this is brushed almost.

Or looks somebody sandblasted it.

- That part's not.

I don't remember that on this card.

- Alright, so the spring loading mechanism

is just from the bend of the back plate, that's kinda cool.

- [Jake] So I checked the CMP thing.

Doesn't look like it.

- [Andy] I wonder why they wouldn't have like a map.

- [Jake] This doesn't look brushed at all.

What did we, last time we twisted?

- [Andy] No, I don't think we did.

- Yeah we did.

- [Andy] I'm pretty sure I just rimmed on it.

- [Jake] Oh God! No.

You were against rimming on it.

And then we were like, just twist a little.

- [Jake] Oh.

God.

Ah.

It has an IHS.

It looks basically the same.

- [Andy] Yeah.

- [Jake] We're gonna have to clean that off and see

there's not much alcohol.

- [Jake] No, I like to go in dry first.

So yep, that's the same thing, alright.

I mean, this isn't the first time Nvidia

has used the same Silicon in two different products

with two different capabilities.

We see the same thing

with their Quadro lineup versus their GForce lineup

where things will just be disabled

through drivers or fusing off different functional units

on the chip.

What I wanna know then

is besides the lack of NV Link connectors on this one.

- Well, they are in there.

They're just not accessible and they probably don't work.

- Right.

What is the actual difference

in function between them?

(Jake sighs)

- Well, this one doesn't have full PCIE 16X,

- Right? - It does less memory.

I think it has way less transistors,

but it is still a GA100.

- Yeah, so the transistors are there.

- Yeah, they're probably just not functional.

Let me see what the chip number is on that one.

- Yeah, 'cause were we not even able

to find a proper Nvidia.com reference to this one anyway.

So we're just relying on someone else's spec sheet.

So the transistor count could just be wrong.

- Okay, so this is so the CMP card was a GA.

- Look at this guy?

- Yeah.

What a weirdo.

GA 100-105F.

And this is a GA100-833.

- If it's a GA100,

I guess it could be a different GA100.

I don't know.

- I mean, it used to be back in the day,

you would assume that it's just using the same Silicon

as the GForce cards because Nvidia's data center business

hadn't gotten that big yet,

but nowadays, they can totally justify,

an individual, like new guide design

for a particular lineup

of enterprise product.

- And interestingly enough,

the SXM version doesn't have an IHS

at least it seems that way.

But the SXM version is also like 400 Watts.

And this is like 250.

- [Andy] Yeah, totally different classes

of capabilities, alright?

Let's put it back together then, shall we?

- I got your new goop. - Goop me.

- I brought two goops.

- We're going for the no look catch.

- Oh yeah baby. - Yes.

X marks the spot, baby.