TSMC Exploring On-Chip, Semiconductor-Integrated Watercooling

[Zodiark1593]

14 hours ago, Kisai said:

Unfortunately, no "water cooling" is safe to use in any of these manners except precisely controlled ones, eg a data-center.

 

Home computers will likely never see it, and no manufacturer is going to want to offer a 10 year warranty on something water-cooled. It's pulling teeth just to get a 3 year warranty on anything.

 

Mobile devices certainly aren't going to see it.

 

Heck, what is certainly to happen is that transporting anything using liquid cooling in this matter will likely destroy the cooling system due to improper storage during shipping (think thermal expansion/freezing), where something get's left outside or in a truck overnight and it ends up below freezing. 

 

Like, water-cooling is cool, but if you start designing a chip to require it, it's like with helium in hard drives, these are not future-proof at the minimum, and self-destructing at the worst.

 

A standard water cooling loop wouldn’t be applicable here anyway. 
 

The strongest potential for this sort of tech in consumer products would come in the form of sealed vapor chambers, and heat pipes. A vapor chamber with channels in the actual die would allow for heat to spread much easier, which could then be carried away by a standard mounted air cooler or water loop, minus the aforementioned risks of water loop failure. There’s no need for a channeled cpu to be connected to a water loop. 
 

Pretty much instead of CPUs having the standard IHS and TIM or solder compound, the IHS contains the vapor chamber underneath, which will certainly get a lot more use of the water cooler you mount to it. This is what came to mind anyway when channeled CPUs were brought up.
 

I’d be quite disappointed if these CPUs were required to be part of a water loop. 

[Sauron]

I'd be very worried about gunking and, if they somehow find a way to avoid that, about the fluid leaking out of the chip and being a nightmare to replace.

[rcmaehl]

I'm ready I'm ready. We've gone from simple prototypes to A-B testing. Enterprise adoption when?

 

[Bombastinator]

16 hours ago, Sauron said:

I'd be very worried about gunking and, if they somehow find a way to avoid that, about the fluid leaking out of the chip and being a nightmare to replace.

These aren’t problems with current heat pipes.  They might not be able to use the current sterilization system of just boiling water inside them which is pretty effective.  Not many critters can withstand boiling AND vacuum.  Plus there’s no light and nothing to eat in there.  Requires that the chips actually be able to take it and I’m not sure they can.  How well do silicon ICs stand up to radiation?  That’s another fast quick and cheap method. Wouldn’t need to be powered.  Just come out the other side with traces intact.

Edited July 15, 2021 by Bombastinator

[Bombastinator]

Another thought for a possible cooling solution is graphene layering.  Iirc they were trying that at one point to use laptop cases as  radiator housings.  I suspect it wouldn’t work all that well because the stuff can be fragile.  An IC wouldn’t have to even bend though.

[Sauron]

3 hours ago, Bombastinator said:

These aren’t problems with current heat pipes.

Either these chips come soldered to the entire heatsink or there will have to be a way to connect them to one.

[Bombastinator]

7 hours ago, Sauron said:

Either these chips come soldered to the entire heatsink or there will have to be a way to connect them to one.

Like is currently the case and why thermal paste exists?

[Sauron]

5 minutes ago, Bombastinator said:

Like is currently the case and why thermal paste exists?

...you can't do that if you need to carry the water from inside the chip to a radiator...

[Bombastinator]

Just now, Sauron said:

...you can't do that if you need to carry the water from inside the chip to a radiator...

You don’t need to carry the water, you need to carry the heat. You can do that by carrying the water but you don’t HAVE to, it’s just easier.

[Sauron]

17 minutes ago, Bombastinator said:

You don’t need to carry the water, you need to carry the heat. You can do that by carrying the water but you don’t HAVE to, it’s just easier.

then what is the point of using water in the first place?

[pas008]

cube/box cpus?

5 sides silicon top side waterblock ports?

think this would work?

[Bombastinator]

Just now, pas008 said:

cube/box cpus?

5 sides silicon top side waterblock ports?

think this would work?

Boxes have 6 sides. Work and work best are different things.  My suspicion is not work best but I could be wrong.  Happens a lot.

[pas008]

47 minutes ago, Bombastinator said:

Boxes have 6 sides. Work and work best are different things.  My suspicion is not work best but I could be wrong.  Happens a lot.

top side is for waterblock ports

bottom and 4 sides are silicon  for i/o gpu etc just a 3d chiplet in cube/box form with waterblock mounted to top to transfer heat out

[Bombastinator]

25 minutes ago, pas008 said:

top side is for waterblock ports

bottom and 4 sides are silicon  for i/o gpu etc just a 3d chiplet in cube/box form with waterblock mounted to top to transfer heat out

Would have to be some small ports or make the ihs bigger so a different socket.  I think that when/if this comes out there may be no ports at all.  Heat will be transferred through the ihs. 

[pas008]

16 minutes ago, Bombastinator said:

Would have to be some small ports or make the ihs bigger so a different socket.  I think that when/if this comes out there may be no ports at all.  Heat will be transferred through the ihs. 

would have to have bigger socket for all the pins cause there would be much needed communication

imagine something like 4 x 5600x around vertically an i/o and gpu capped on top with ihs like with waterblock ports could also have the whole thing surrounded by copper/aluminum for stability

not sure how else i could describe it but I have wondered about this for awhile

[Bombastinator]

2 hours ago, pas008 said:

would have to have bigger socket for all the pins cause there would be much needed communication

imagine something like 4 x 5600x around vertically an i/o and gpu capped on top with ihs like with waterblock ports could also have the whole thing surrounded by copper/aluminum for stability

not sure how else i could describe it but I have wondered about this for awhile

If one makes the ihs bigger it also needs to be thicker to prevent denting.  Might be room for that.  I don’t know.  Putting water block ports on the IHS would basically preclude everything but watercooling though.  Seems unlikely. Watercooling requires more maintenance and big companies who buy the lions share of CPUs simply aren’t set up for it.  I suspect that if a way to leave the final cooling system without a pump or infectable water is possible it will be done even if it’s more complicated.  Maintenance is a major cost for businesses. Having to worry about water stuff is not something they would voluntarily do. If there was a different system that could fit their needs that didn’t work as well and cost more they might do it simply because TCO would still be lower.

Edited July 15, 2021 by Bombastinator

[pas008]

1 hour ago, Bombastinator said:

If one makes the ihs bigger it also needs to be thicker to prevent denting.  Might be room for that.  I don’t know.  Putting water block ports on the IHS would basically preclude everything but watercooling though.  Seems unlikely. Watercooling requires more maintenance and big companies who buy the lions share of CPUs simply aren’t set up for it.  I suspect that if a way to leave the final cooling system without a pump or infectable water is possible it will be done even if it’s more complicated.  Maintenance is a major cost for businesses. Having to worry about water stuff is not something they would voluntarily do. If there was a different system that could fit their needs that didn’t work as well and cost more they might do it simply because TCO would still be lower.

if they would incorporate something like quick disconnect like what ek and hospitals use could be very effective

4-5x epycs like this would be killer lol

[Bombastinator]

20 minutes ago, pas008 said:

if they would incorporate something like quick disconnect like what ek and hospitals use could be very effective

4-5x epycs like this would be killer lol

That’s already assumed. Wouldn’t even be possible without them.  How the heck are you going to get fingers or a wrench in there?   Still a potential major problem because of the majority users thing.  Also those things are expensive

[Mark Kaine]

On 7/13/2021 at 7:02 PM, Bombastinator said:

Also they were all at least 20 years old so they often used technology not even in use anymore.  They did tend to be quite loud. 

No shit haha. But hdds can last a long time (actually i have many approaching or exceeding 20 years still fully working…)

My favorite harddrive type: (love the sound lol)

"8bit IDE XT" 

 

[Bombastinator]

1 hour ago, Mark Kaine said:

No shit haha. But hdds can last a long time (actually i have many approaching or exceeding 20 years still fully working…)

My favorite harddrive type: (love the sound lol)

"8bit IDE XT" 

 

That’s the thing though: can.  There are model Ts that have lasted near a hundred.  Most didnt.  There is an argument made that the model T was a well made car because of these survivors.  I would call it a does not follow argument.  Not all HDDs are built the same. Some will wear well some won’t. One seems to be able to expect a HDD to last about 5 years.  That is either larger of smaller than an SSD depending on how much it is written to.  There are other issues that start to come into play after large amounts of time.  Outgassing I suspect is going to be a big one.  Plastics get brittle as they outgas.  It wouldn’t surprise me at alll if some of the older simpler HDDs with thicker metal and things last occasionally much much longer than SSDs simply because they’re mostly plastic.  There’s this thing in art called archival quality.  A lot of great masters are great masters because their work still exists and their work still exists because it was saved and treated carefully, but also because a lot of the materials that were used are archival.  Heavy metals used as pigments remain bright, etc..

[cj09beira]

@leadeater talking about hard drives does orientation matter for drive reliability?, my 1 min of though process seems to indicate that vertical orientation could be better for bearing lifespan as there is more surface area distributing the load, a quick google search doesn't seem productive "either orientation will keep error rate within spec".

[Quackers101]

3 hours ago, cj09beira said:

@leadeater talking about hard drives does orientation matter for drive reliability?, my 1 min of though process seems to indicate that vertical orientation could be better for bearing lifespan as there is more surface area distributing the load, a quick google search doesn't seem productive "either orientation will keep error rate within spec".

Would think both have their cons or pros. So long it stops vibration damage etc. But if its slanted or / I wonder if that wouldnt be so great.