HomeVideos

Solving the GPU Airflow Problem | Bouncing Air in the HAVN HS 360

Now Playing

Solving the GPU Airflow Problem | Bouncing Air in the HAVN HS 360

Transcript

344 segments

0:01

Okay, we have one of the most

0:03

complicated cases for Computex. This is

0:05

the HS 360. This is from Haven. So, what

0:08

Haven's done is actually amusing, I

0:11

guess. I It should work pretty well, but

0:14

basically, there's another glass sheet

0:15

up here. Now, this sits on top of the

0:18

GPU. You can see that rubber sort of

0:20

bumper at the end that's uh allowing it

0:23

to almost directly contact the GPU. And

0:24

then if you look at the bottom front of

0:26

the video card, it's directly contacting

0:28

that glass sheet. So what's happening is

0:30

if we spin to this uh CFD simulation,

0:34

this computational fluid dynamic

0:35

simulation, it it all makes sense. This

0:38

is really fun. Like this is just it's

0:40

kind of fun science where basically what

0:43

you have is uh without the glass flow

0:47

separator here. So that's that new

0:48

piece. This is the one that previously

0:50

existed on the BF3 or the HS420, the

0:53

original HS420. Uh, and that's present

0:56

here as well, but you can see they've

0:57

added the glass separator. And then this

0:58

is the same thing, no glass separator,

1:00

still has that HS420 style uh,

1:03

deflection. And then this is with the

1:04

glass separator. So basically what

1:06

you're seeing is at least in simulation

1:09

we'll you know this is going to depend

1:10

kind of build to build um and we'll test

1:13

it but in simulation the uh for this the

1:16

hot air if you're at 100% GPU fan speed

1:18

which is very uncommon but good for kind

1:20

of a demo uh it's shooting all the air

1:23

up and out or out the bottom because the

1:25

fins are oriented vertically and then

1:27

that is getting pulled back into the

1:29

fans at that 100% speed without a flow

1:31

separator. This makes sense. This is not

1:33

like new. This isn't, you know,

1:36

revolutionary science. This is just how

1:37

it normally works. Uh, with a 55% speed,

1:40

it's kind of a different problem where

1:42

for this, still hot air exhausting top

1:45

and bottom, but the fans on the GPU are

1:47

lower speed. So, they're less likely to

1:49

pull that air in and recirculate it. And

1:52

instead, it's it's going up and out. So,

1:54

this is just a difference of positive

1:55

versus negative pressure setup for the

1:57

case, or at least for this region of the

1:59

case. And the end result is the cool air

2:01

coming in from the bottom for this one

2:03

is sometimes circumventing the GPU is

2:05

not getting pulled in. We talk about

2:06

this a lot in our testing. So you can

2:08

see this stuff in data like in thermal

2:10

data with results and we show that uh

2:12

but the simulation helps make clear kind

2:15

of what's happening. And what'll be

2:17

interesting for us to test is with this

2:19

flow separator in the two different

2:21

conditions, you when you have the

2:23

problem of the hot air getting

2:24

recirculated, this will block that

2:26

obviously cuz looking back at the card,

2:28

you these fins, they're set back towards

2:30

the CPU cooler. They're oriented

2:31

vertically. Most fins are on GPUs these

2:33

days. And uh they it's just it's not

2:36

going to be able to get back down

2:37

because there's a piece of glass

2:38

blocking it. Um other than maybe a

2:40

little bit at the back or something. And

2:42

then under the the lower flow uh

2:45

situation where it's like 55% speed

2:47

which is pretty typical for most GPUs.

2:49

Um I think we set ours to that for our

2:51

case testing in that condition. It's uh

2:54

the difference of where the the cool air

2:57

here is actually just getting blocked so

2:59

it's not going up and out and it's more

3:01

forced into the card. Pretty cool stuff.

3:04

I just I thought it was neat to see the

3:05

simulation of it. Uh we'll do the

3:07

testing of course, but that is the big

3:08

thing that Havens tried to do moving

3:10

from the HS420, which is a much larger

3:13

version of this case, uh to the HS360.

3:16

It was basically like instead of getting

3:19

rid of the concept in its entirely

3:21

entirety, let's double down and put

3:24

another piece of glass in the case. So

3:27

now we can have the full curved glass

3:29

front, the glass deflector, and the

3:31

glass separator. uh they are really

3:33

determined to make this idea work. So,

3:36

I'm looking forward to testing it, but

3:37

that's the big change. Um there's a

3:38

bunch of others, too. We'll go through

3:39

them all. Before that, this video is

3:41

brought to you by Arctic and their MX7

3:42

thermal paste. Arctic's new MX7 is a

3:45

higher viscosity paste with Arctic

3:47

claiming that the increase in viscosity

3:48

from filler content and the texture

3:50

benefit improve its performance while

3:52

reducing the risk of the pumpout effect

3:54

over time. MX7 is intended for longer

3:56

endurance applications and to be left

3:58

and used for extended service life.

3:59

Arctic also now has a counterfeit paste

4:02

checking website because believe it or

4:04

not, knockoff paste from non- firstparty

4:06

sources is actually relatively common

4:08

for the big brands to experience. So,

4:10

they have a site to check the validity

4:12

of it. Get a tube of Arctic's MX7

4:14

thermal paste at the link in the

4:16

description below. All right, so the HS

4:18

360 is a relatively difficult case for

4:20

me to memorize and go over at a show uh

4:23

because there's a lot going on with it,

4:24

which is a good thing. But the Haven

4:27

series cases, we've overall been

4:28

positive on the BF360 and the HS420 both

4:32

overall tested pretty well. We did

4:34

obviously have various criticisms of

4:36

them like every case. One of them was on

4:38

the HS420 it was a I think the technical

4:42

term is gigantic pain in the ass to deal

4:45

with the fan in the back of the case and

4:47

they have fixed that. So, this is just a

4:50

fan mount or bracket that goes in the

4:52

back here. And, uh, rather than having

4:55

to, uh, be careful about what order you

4:59

install the rear fans in, now you just

5:01

access it from the backside. So, that's

5:03

a lot better. They're still doing the

5:05

thing that they did with the previous

5:06

cases, which is putting rubber at

5:08

basically every single contact point of

5:10

the case, where metal might touch metal.

5:12

There's rubber all over the place. So,

5:14

that's something we've been positive on.

5:16

We've liked with the Haven cases. It's a

5:17

really small touch. Um I don't know,

5:19

maybe like sometimes it might help with

5:21

vibrational noise. It's more of just

5:22

kind of a nice uh attention to detail

5:24

thing though. And then some other

5:27

improvements and changes. Um so the was

5:30

it over here on the So we did talk, this

5:33

is on the previous uh case that we

5:36

reviewed, but I did just want to give an

5:38

update for everyone. Previously

5:41

we did talk about the recommended uh

5:44

cable paths and they have fixed that. So

5:48

so Haven has just been working really

5:50

hard at improving everything that we've

5:53

criticized from from spelling to the

5:54

mechanical design. Um some other

5:56

interesting stuff with this. So for the

6:00

new case, this is their attempt at

6:02

trying to solve the fact that anyone who

6:04

supports both a back connect board and a

6:07

standard board uh ends up creating a

6:10

cheese grater motherboard tray. And uh

6:13

basically you can see the kind of the

6:15

holes for both here. So this is what

6:17

everyone's shipping when they try to

6:18

support both. Looks kind of ugly. And

6:22

they're just doing a separable plate.

6:23

Now this is currently done with a bunch

6:25

of screws. I would love to see this

6:27

implemented in a way where there's a a

6:31

simple mechanism either it's you know

6:33

plastic clips or metal socket and ball

6:35

or whatever um to get it to be a little

6:37

quicker to pull out. Now you should

6:40

probably only be doing this once as an

6:42

end user but uh it would be nicer.

6:44

They're going to ship this in a standard

6:45

configuration by default. And then if

6:47

you want a back connect, you would

6:49

basically pull it out, flip it like

6:51

we're doing here, and then that would

6:53

give you the the whole pattern without

6:55

having holes through the entire

6:56

motherboard tray. Um, and then there are

6:59

the grommets that we've also come to

7:02

expect from Haven. So, they've got kind

7:03

of the stringy kind where you can run

7:05

the cables through them and separate

7:06

them. Uh, those can go in if it's

7:08

standard layout. If it's back connect,

7:10

you probably don't want that for obvious

7:11

reasons. Um, but those come out with

7:13

back connect and then they also have

7:14

these cable combs that uh that go in the

7:17

case, too. So, a lot of really small

7:19

like attention to detail touches.

7:21

Pricing on these is supposed to be $150

7:23

for base and then $230 for the vertical

7:27

GPU model. Uh, and then there, this is a

7:31

different case. is this HS420, but um

7:33

this is also going to be in the uh uh

7:36

upper 200s for the technically not

7:39

Mercedes I guess a AMG Patronis Formula

7:42

1 team. I've been told in no uncertain

7:45

terms that I am not to say that this is

7:48

a Mercedes collaboration. This is not a

7:51

Mercedes case.

7:54

This case with that logo is not a

7:58

Mercedes case. It is an AMG Patronis

8:02

Formula 1 team case. I hope we are all

8:05

clear that this is not a Mercedes case,

8:08

especially the lawyers at Mercedes.

8:11

Okay. All right. So, another thing that

8:12

Haven's working on is on the mechanical

8:14

design where uh this corner, they wanted

8:16

to strengthen that. And if you look

8:18

internally at the older HS420 here, the

8:21

main thing to notice is that this is

8:22

just kind of a flat bar for this top

8:25

corner. Um there's been some changes to

8:26

the way the panel secures, too. on the

8:28

new HS360.

8:30

This is a far more complicated uh set of

8:33

steel here. So you can see it kind of

8:35

bends and this like punched out section.

8:38

Uh all of this is intentional where they

8:40

say it's basically meant to improve the

8:42

structural strength. So Haven ran these

8:46

simulations as well where this is a

8:47

force application scenario. So this is 5

8:49

kg uh of force pushing down on that

8:52

front edge without a glass panel

8:54

underneath it. So without any support

8:55

under that front corner. Um, and then

8:58

what you're looking at is deformation.

9:00

So this is in millimeters deformation.

9:02

This is pretty cool, uh, from a software

9:04

standpoint. So they're seeing like

9:05

almost.5 millimeters of deformation on

9:08

the old design with the HS420. And then

9:11

the H360, which is the new design that's

9:12

got that reinforced

9:14

uh, piece underneath the uh, the fan

9:17

tray panel. There's, you know, 0.3 mm

9:19

deformation max. So pretty cool stuff.

9:22

It's like this is what we liked with the

9:24

BF360 announcement last year at Computex

9:27

was having some actual data available.

9:30

We can go through some of the other

9:31

stuff. So on this one where they were

9:33

doing the simulation of flow, uh the big

9:36

thing to notice on the right side for

9:37

the GPU fans at 55%. The GPU temperature

9:41

is 60.2° without the flow separator,

9:44

52.7 with it. So that's pretty big

9:47

difference. It's like 7 degrees is a

9:49

gigantic difference for a GPU. At the

9:51

100% fan speed, they're not really

9:53

seeing any difference, which I would

9:54

expect just because uh when you're

9:57

blasting the fans that much, you're kind

9:59

of overpowering a lot of the um more

10:02

nuanced differences anyway. And then on

10:05

this one, so this is a liquid cooler in

10:08

the HS 360 with a VGPU

10:10

and uh I've seen a couple degrees of

10:12

performance or one almost two degrees of

10:14

performance improvement. noise

10:16

normalized on the left image with the uh

10:19

with and without the flow separator. And

10:21

then for the HS420, they also did

10:23

testing with the flow separator, even

10:25

though the focus today is on the 360,

10:26

but see it added here. That improvement

10:29

is pretty big. Then Haven says they are

10:31

planning to do a flow separator as a

10:33

separate component for the HS420s in the

10:36

future. So this is Haven's uh

10:38

competitive analysis chart. These are

10:40

not necessarily the cases I would choose

10:42

to compare against. The Corsair 3500X

10:45

sucks. Uh the Lean Lee 011 is old and

10:49

the NZXTH9 is from NZXT. Um so we would

10:52

compare against probably different

10:53

cases, but at least versus this Eclectic

10:56

mix, they you know they're showing well.

10:58

We'll we'll test that in our own

10:59

benchmarking of course, more noise

11:01

stuff, but anyway, that's kind of some

11:03

of the basics on the data. Um the plan

11:06

for Haven is they want to launch around

11:08

September for the HS360

11:12

and uh pricing as I said is 150 base no

11:15

fans um or 230 for the VGPU option which

11:20

uh we tested both the VGPU and the base

11:22

models for the HS420 if you're curious

11:24

how that kind of all looks and works.

11:26

But the general design, I guess, look is

11:29

very similar to the uh the past cases

11:33

where they're still doing these really

11:35

large um slats. And on on the past Haven

11:39

cases we've looked at, the spacing for

11:42

the uh supports has typically been

11:45

pretty good in terms of especially on

11:46

the side panel in terms of minimizing

11:48

how much overlap you have for the

11:50

support on top of the fan blades. So,

11:52

that's been one of Haven's strong points

11:54

is trying to align like on the side

11:56

panel the supports with the fan hub

11:58

where the hub is a dead zone for fans.

12:00

You're not going to get really any air

12:02

going through there. Um, and that's all

12:04

been kept with the newer case. So,

12:07

mostly just updates and refinement for a

12:09

smaller uh HS series case that they are

12:13

taking some some learnings from and

12:15

improving on. I think we've kind of

12:16

covered most of it. The rest of this

12:18

stuff is I mean it's like the same as

12:20

the past with the uh drive cages that

12:22

can mount on the side versus a a cooler

12:25

fan. So, um I think that pretty much

12:27

covers this stuff and otherwise we're

12:29

looking forward to getting the 360 in

12:31

for testing. Uh there's going to be a

12:33

lot of AB testing to do with the

12:34

separator and the glass. I think that

12:36

pretty much covers it though. So far, we

12:38

haven't seen a ton of cases yet, but out

12:39

of the ones I've seen, this is the most

12:41

interesting one to me. So, we'll see how

12:42

the rest of the show goes and check back

12:44

for more. As always, uh subscribe for

12:46

the rest of the news. will be here all

12:47

week and we'll see you all next

Interactive Summary

The video highlights the new Haven HS360 PC case, focusing on its innovative design choices, particularly the inclusion of a 'glass separator' meant to improve thermal efficiency by preventing hot air recirculation within the GPU region. The host details various mechanical refinements over previous models, such as improved rear fan access, increased structural strength, and modular motherboard tray designs. The case is scheduled for a September release with pricing starting at $150.

Suggested questions

3 ready-made prompts