Jouni osmala (fname.sname.delete@this.aalto.fi) on August 16, 2019 9:46 am wrote:
> Adrian (a.delete@this.acm.org) on August 16, 2019 12:12 am wrote:
> > Jouni Osmala (fname.sname.delete@this.aalto.fi) on August 15, 2019 5:58 am wrote:
> > > Adrian (a.delete@this.acm.org) on August 15, 2019 3:47 am wrote:
> > >
> > > > In conclusion, while for small computers I also always attempt to use only fanless computers,
> > > > for large computers I do not believe that it is worthwhile to accept the much larger
> > > > size & weight that is required for a fanless computer at any given TDP.
> > >
> > >
> > > I disagree about the size but not about the weight, which is mostly dependent on personal fitness. I
> > > measured the extra volume my cramped midtower case made by nexus takes to insulate better for noise.
> > > Convection heatsink of that size can emit 25W per degree of difference between ambient and the heatsink.
> > > Assuming ambient around 20C, you can emit atleast 750W with reasonable temperatures for the components.
> > > This assumes fanless system design, not trying to put fanless individual components inside the case,
> > > but use heatpipes to transfer the heat to the case side panel which operates as heatsink.
> > > Compulab airtop3 has high enough total power in a very small form factor that scaling
> > > its gpu 160w heatsink to 250w and upgrading CPU heatsink to that wouldn't make it particularly
> > > large design. Here's its comparison to other small form factor PC:s.
> > > https://www.youtube.com/watch?v=35OyZzCvG0g&t=1m22s
> >
> >
> >
> > Where I live there is no doubt about climate warming, so I must assume ambient
> > of 40C, as such temperatures became not infrequent during summers.
>
> Gee. I got movable air conditioning unit because 25C heatwave was so terrible. And yes it
> is noisy but I only operate it when temperature would be worse problem than its noise.
>
> > However, for now, Compulab Airtop is a unique fanless case on the market. All the other
> > available fanless cases, including those shown at the link provided by you in your previous
> > post, are much larger for a given TDP than a case with active cooling.
>
>
> Well those fractal design cases people recommend for air cooled quiet high power threadripper
> take significantly more floor space than the passively cooled threadripper system from
> my link. The case is taller, but who cares it still would fit well underdesk.
>
> > Also, if I would be willing to spend enough time, I could make in my home a case equivalent
> > to or better than any other fanless case that I have ever seen, except Airtop, by
> > cutting, milling and assembling extruded aluminum or copper profiles.
>
> That is my main point also. There should be better fanless cases commercially available,
> since they are possible to make. That German boutique vendor that I linked for threadripper,
> probably thought the same and did it but he just thinks that people would just make terrible
> mess if he sold it as a kit. Every other part seems to be standard except cooling solution.
> And cooling solution based on pictures could be tweaked to fit other systems.
> The downside is that components that you would really want to reuse limit the power consumption and he seems
> to have designed it with 240W CPU+240W GPU limit by the heatpipe size and number of locations where it can
> be attached to the side panel. Unfortunately latest threadrippers and 2080 ti both have 250W TDP . The difference
> is small enough that it probably doesn't matter and causes just tiny increase in temperature.
> I love what that german computer made in concept, but would prefer slight increase
> in potential cooling capacity for CPU and GPU, and bigger power supply.

Have you considered buying an aquarium or just a plastic tub, filling it with Novec, and dropping your servers in without the sides on the case? Novec isn't cheap but if you put a lid on it you are good.

https://www.semiaccurate.com/2014/04/09/3m-intel-sgi-immersion-cool-high-density-servers-novec/

-Charlie