Re: Intel publishes Larrabee paper
- From: Quadibloc <jsavard@xxxxxxxxx>
- Date: Thu, 5 Mar 2009 08:31:51 -0800 (PST)
On Mar 4, 11:29 am, Rick Jones <rick.jon...@xxxxxx> wrote:
I do not seek to diminish the issues, just clarify a point - does
there _have_ to be a mechanical pump? Don't heatpipes work via
convection?
It is true that heat pipes _do_ work by convection. The heated fluid
moves out of the area being cooled, and then gives up heat somewhere
else; this could be simply by conduction and radiation, or maybe even
by the more elaborate refrigerator-like processes cited in another
reply.
But while that does let them remove more heat than a conductor would,
this still removes less heat than liquid cooling involving a pump,
because then you can bring huge quantities of cool liquid to the
affected area, much more than the power from the heat being dissipated
can move by itself.
So, if you're trying to control the temperature of a chip within
narrow limits, and if the geometry is such that you only have narrow
channels in which to bring cooling fluid to the relevant area
(requiring more force to push the fluid through) then a mechanical
pump is beneficial.
John Savard
.
- Follow-Ups:
- Re: Intel publishes Larrabee paper
- From: Rob Warnock
- Re: Intel publishes Larrabee paper
- From: Larry Elmore
- Re: Intel publishes Larrabee paper
- References:
- Re: Intel publishes Larrabee paper
- From: Tom Knight
- Re: Intel publishes Larrabee paper
- From: nmm1
- Re: Intel publishes Larrabee paper
- From: Kim Enkovaara
- Re: Intel publishes Larrabee paper
- From: nmm1
- Re: Intel publishes Larrabee paper
- From: Rick Jones
- Re: Intel publishes Larrabee paper
- Prev by Date: Re: When good computers go bad...
- Next by Date: Re: Intel publishes Larrabee paper
- Previous by thread: Re: Intel publishes Larrabee paper
- Next by thread: Re: Intel publishes Larrabee paper
- Index(es):
Relevant Pages
|