Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
nycesquire

Antimemetic: Scratch-built, Fully Passive, GTX 1080, SFF

Recommended Posts

Posted · Original PosterOP

14xkEs8h.png

I'd like to solicit some feedback on the cooling solution.  Can you suggest a better arrangement than the following? Imagine that all of the copper blocks are connected by heat pipes like the two central channels at the top (the chassis is laying on its side with the top facing the camera):


mJCKJt9h.png


here we are simulating an installed card, looking in from the side:


2gELIKVh.png


And the other side:


DBBITfth.png


And the top again:


TCPmT6th.png

 

Paging @Stefan1024: What do you think?

Link to post
Share on other sites

I think you have done a lot of effort to spread the heat. More than needed for a passive cooling solution. But if should work out fine.

Just be sure the surfaces of the copper blocks and the heat sink is plane and you use a lot of mounting preassure and thermal paste for all the metal parts. The surplus thermal paste will be automaticly removed when you thighen the screws and you can clean it.

 

But the most critical part is: i can't see any cooling for the VRMs (including the coils) or the VRM. Even on the 970 the VRMs reached >100 C within minutes while the core was at <70 C.

 

Make copper block to connec them also to the main heatsink.


Mineral oil and 40 kg aluminium heat sinks are a perfect combination: 73 cores and a Titan X, Twenty Thousand Leagues Under the Oil

Link to post
Share on other sites
Posted · Original PosterOP

Thanks for the feedback. I agree that I've done more than needed here. Over-engineering FTW! I think the bottom two heat blocks are overkill and will be removed.  Heatpipes don't work well in that configuration, and I have sufficient surface area for contact anyway.

 

As to the VRMs, there is currently a heat spreader (you can see it in black) covering the entire card. I will either attach a few heatsinks to the spreader or just use small copper heat sinks on each VRM individually.

 

Do you think the distance between the copper blocks is important, or will thermal equilibrium take over and make the arrangement irrelevant?

Link to post
Share on other sites

Adding a big buffer is allways a good idea when it comes to passive cooling, so overengineering is fine.

 

Also the heatspreader does only what the word implies: spreading the heat. It will be be able to dissipiate enougth and overheat. The smal copper heatsinks that are like 20x20mm are basicly useless when it comes to passive cooling. Also the natural convection will not work good in the smal gap between the card and the main heatsink.

 

Use a copper block to connect the heat spreader to the main heat sink or a big heat sink to cover the back of the card.

And don't forget the coils as well.

 

If you have any chance borrow a thermal camera to check your design after building it.


Mineral oil and 40 kg aluminium heat sinks are a perfect combination: 73 cores and a Titan X, Twenty Thousand Leagues Under the Oil

Link to post
Share on other sites
22 hours ago, nycesquire said:

- snip -

Haha, now I know why you waited so long, you switched to the 1080!

OC and temperature looks formidable.

 

How long does it take to heat up the entire heat sink? 45 minutes?


Mineral oil and 40 kg aluminium heat sinks are a perfect combination: 73 cores and a Titan X, Twenty Thousand Leagues Under the Oil

Link to post
Share on other sites
Posted · Original PosterOP

14xkEs8h.png

 

 

Admit it, you thought I had given up, didn't you?  Here's the story: I got busy with work, time passed. I started building the GPU cooling solution (which is complete - as demonstrated by the screenshot above) then three things happened:

 

1) I started getting into the SFF forums and well... I decided that Antimemetic was just too damn big.  So much wasted space on the inside!  I am not an SLI guy, so why bother with four heat sinks anyway?  Where am I going to find the space for the gull wings in my tiny apartment?

2) I realized that copper wouldn't match my new office decor.  

 

Most importantly:

 

3) I decided I wanted to keep it, not sell it, and I still wanted to automate it.

 

So I completely redesigned the thing! No copper. No hinges.  Instead, low key stealth:

 

Z1xDMzBh.jpg

 

 

ynZNCZAh.jpg

 

 

0IO7Mv3h.jpg

 

From the top:

0aY6Bo0h.jpg

 

 

And from the bottom:

7mn6TI9h.jpg

 

 

Here we are without one of the heat sinks:

s7bRvl1h.jpg

 

 

And without the other:

ECFzq7ih.jpg

 

 

Here's a sense of size:

Byi9fzth.jpg

 

 

With a case this tiny, it'll actually be quite difficult to maneuver around inside.  It is only 100mm from the inside of one heat sink to the inside of the other. Indeed, the entire width of the case, heat sinks included is less than 165mm...

 

SO HOW ABOUT WE AUTOMATE IT

 

vsuelDnh.jpg

 

 

 

5I78BeSh.jpg

 

 

The automation separates the heat sinks by a total of 200 mm for a total of 300mm of space inside to work.  It is driven by a single large linear actuator. You can see the two rods that will push the heat sinks out.  They ride on a set of drawer slides:

 

glp2UJMh.jpg

 

Let's be honest with each other though: These heat sinks are monstrously heavy. There's no guarantee that the slides will support the weight without sagging or otherwise falling apart.  If, in the sad face case, the automation is unsuccessful, It'll be a dark day, but I will settle for this alternative:

 

2TFhbtzh.jpg

 

 

An even tinier, but much simpler and sturdier case.

 

Next up, an in-depth description of the cooling solution (with pictars).  I'd love your feedback!

Link to post
Share on other sites

Just wow 


Intel i7 6700k @4.6 1.330v  cooled with H110i GT // 16Gb Kingston Fury DDR4 ram @2133MHz // Rx 480 Nitro+ // 2TB + 1 TB Hdd  // 250 GbSSD // on an Asus Z170-A // powered with Corsair RM750i // all  inside a Corsair 600C

Link to post
Share on other sites

@nycesquire

This project started way to professionell to give up in the midle ;)

SFF and passive cooling looks like a nice combination.

 

However I'm a bit worrid about the airflow inside your case. The front, toop and both sides are solid and allow no airflow. This leads to a heat trap and the big heat sinks will get conftable warm and heat up the air inside as well. The PSU, the VRMs on the motherboard as well as the RAM migth get unhappy. My PSU was in a similar configuration and peaked at 90°C.

 

Do you use heat pipes for the CPU as well? 


Mineral oil and 40 kg aluminium heat sinks are a perfect combination: 73 cores and a Titan X, Twenty Thousand Leagues Under the Oil

Link to post
Share on other sites
Posted · Original PosterOP
5 hours ago, Stefan1024 said:

@nycesquire

This project started way to professionell to give up in the midle ;)

SFF and passive cooling looks like a nice combination.

 

However I'm a bit worrid about the airflow inside your case. The front, toop and both sides are solid and allow no airflow. This leads to a heat trap and the big heat sinks will get conftable warm and heat up the air inside as well. The PSU, the VRMs on the motherboard as well as the RAM migth get unhappy. My PSU was in a similar configuration and peaked at 90°C.

 

Do you use heat pipes for the CPU as well? 

Thanks!  The front, top and back are all open, actually. the translucent panels there are just how sketchup renders mesh :)  Still, there are very good reasons to think the internal temp will be a problem. I have still not tested it with both the CPU and GPU attached together as shown. 

 

Yes, the CPU is heatpiped as well. One heat sink for each major heat generator.  My early tests of the CPU heat pipe solution shows that it performs much worse than the GPU, but still well within tolerances. No throttling even after an hour at 100% usage!  Will have pics of that soon. The problem there is that I can't press the CPU directly up against a heat sink.

2 hours ago, Str8f4c3 said:

I do like this design lots more than the original, can't wait to see it finished!

Thanks!

Link to post
Share on other sites
Posted · Original PosterOP

14xkEs8h.png

 

Enough with the design. Let's talk engineering.  This is how I am cooling the GPU, which is an EVGA 1080 Super Clocked. First, I start with the HDPlex H5 GPU heatsink system.  Shout out to the folks at   logo.png for hooking me up with a few at a discount.  They advertise the device as only being able to dissipate 95 watts.  In truth, the block and eight heat pipes can dissipate significantly more than that.  Much, much, much more than that. They are just hooking it up to the modest heat sinks on the side of their H5 Chassis, which limit the cooling capability.  I don't have that problem. I'm drilling into 300mm beast. Behold:

 

5bhIzdzh.png

 

gYN3eeDh.png

 

For the evaporator end (the end mounted to the GPU) I use two of the base H5 units back to back, sandwiching eight of my own 6mm heat pipes within.

 

I8o520wh.png

 

j79v685h.png

 

For the condenser end (the end mounted to the heat sink), I use the full h5 unit with the base and fins. Grease them up nicely:

 

 

4OauQvLh.png

 

et voila, the complete GPU core cooling system:

 

gzbydB3h.png

 

It raises the card off the heat sink enough to leave vertical room for the power connector, PCIE extender, board components, etc.

 

Add a big block of aluminum to connect the VRM to the heat sink:

 

y0k0bqhh.png

 

9pJwiu0h.png

 

4adjXGDh.png

 

Drill the necessary holes to securely attach said big block of aluminum to the heat sink (rube-goldberg-esque drill press setup incoming):

 

yavPivgh.png

 

Add some standoffs for ensuring good mounting pressure:

 

m4ckOxsh.png

 

And you have (what might be?) the world's first passive cooling for the Nvidia GTX 1080:

 

m4ckOxsh.png

 

No fans, no water, no pumps, nada.

 

Here's a vertical shot:

 

2LwykiTh.png

 

and here we are testing on the frankenbench (sorry for potato quality):

 

AdPCHXph.png

 

Speaking of testing... RESULTS ARE GOOD. In my ambient room temp of 24 degrees C, here's temps after browsing in chrome for an hour or so:

 

Ia25hAgh.gif

 

And of course, the temps that matter - after an hour or so of Heaven:

 

mTdkUEGh.gif

 

1911 is the boost clock I was getting with stock cooling, so there's no loss due to the cooling solution.  You might note, though, that it is thermal throttling (pink bars).  Indeed, all GTX 1080s will keep clocking up until they either throttle due to max voltage or max thermals. So, I may not have lost any performance, but my headroom for overclocking is probably diminished considerably, unless I'm willing to push past 84 degrees.  Right now, I'm happy with it.  I have some ideas to revamp the cooling solution, in which case I may subject it to the 1080ti when it's released. Guess we'll wait and see :)

 

Next up, the CPU cooling solution!  Until then, what do you think?  As always, I welcome your feedback.

 

 

Link to post
Share on other sites
Posted · Original PosterOP

14xkEs8h.png

 

The GPU side is complete now, and performing famously. Let's talk CPU/Mobo/Ram, power supply and SSDs.  I've made a motherboard tray out of black acrylic. Here we are cutting it to size:

 

GSdEXvnh.png

 

 

And cutting a notch for it to slide cleanly into the aluminum extrusion

 

6zzHt61h.png

 

Drilling each motherboard and SSD standoff

 

u6BFJGmh.png

 

Here we are all laid out

 

EtAP89Wh.png

 

Here, you can see how the tray slides cleanly into the extrusions that hold the case together.  Also, you can see the placement of the PSU above where the SSDs will sit

 

U65LiR7h.png

 

The CPU, an 88 Watt part (4690k) is connected to the heat sink with six heat pipes, all with only minor bending:

 

VlgL291h.png

 

It looks almost organic, doesn't it?  Note the reflections that look like kinks. It's not really kinked much at all actually. Just bright lights.

 

Here we are all sealed up

 

nSoSb3Fh.png

 

and on the desk plugged in and running (a sketchup model of itself!)

 

hVRFYfFh.png

 

Not bad for zero fans! It's under 80 degrees with your usual (non Prime95 voltage voodoo) stress testing and runs under 60 degrees in Heaven and in games.  Thoughts? Feedback?

 

...For those who want to know about totally unrealistic, "never happen in the real world" scenarios, here we are under Prime95's ridiculous stress test in a very warm (25c) room.

 

HM8A8IGh.png

 

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×