On Fri, Sep 09, 2005 at 10:25:38AM -0400, Timothy Miller wrote:
> On 9/9/05, Jack Carroll <[EMAIL PROTECTED]> wrote:
> > On Thu, Sep 08, 2005 at 11:26:51PM -0400, Timothy Miller wrote:
> > > On 9/8/05, Jack Carroll <[EMAIL PROTECTED]> wrote:
> > > > If you have heat sinks on the board, you need space around them
> > > > for
> > > > convection. If you have fans on the heat sinks, you need even more
> > > > clearance for the air to enter the fans. In either case, you're not
> > > > going
> > > > to have the component side facing another board, unless you want to use
> > > > a
> > > > blower and skinny ductwork to deliver the airflow parallel to the plane
> > > > of
> > > > the board.
> > >
> > > The problem we have to deal with is that while some systems have good
> > > airflow around the PCI slots, others seem to have practically none,
> > > and we need to be able to deal with both cases.
> > >
> > > Also, we'd like to sell it without a fan but allow one to be installed
> > > for overclocking.
> >
> >
> > Ah, so. In the latter situation, you'd have to plug the OGD into
> > the end PCI slot, and not put a board in the AGP slot next door. Of course,
> > overclocking might not be that much of an issue for the OGD, since it's not
> > intended for maximum performance in the first place. And an OGC1A goes in
> > an AGP slot, which does have clearance for a fan intake.
>
> The problem is that this card WILL get installed in system with NO
> airflow, and no matter how "non preformance oriented" it is, it WILL
> overheat. There's a certain popular PC manufacturer we often buy from
> who has absolutely NO airflow in the PCI area.
Dunno what to tell you. We could probably figure out something, and
bench-test it to prove that it cools the chips. If we were to put a
centrifugal blower outside the cramped area, and use flat ductwork to carry
the air to the heat sink, it would probably have to be some kind of
cut-to-fit ductwork that could be easily configured in the field to fit
where it has to go. Another possibility might be a heat sink with cooling
pins instead of fins, and increase its area to get the volume up even with
limited thickness. That might work with natural convection flow in any
direction, if the total heat load isn't too large. There still has to be a
way for air to get in and out of the area, though, or nothing will work.
I've done some of this kind of design, but only in well-understood
industrial situations. I'm no expert on forced-air cooling, and I don't
know the component suppliers for blowers that can run from motherboard fan
headers. Maybe Aavid or Wakefield Engineering has some application notes or
tutorials that could educate us. I can at least look.
If this threatens to become a big deal, we might want to mount an
analog temperature IC on the heat sink, so the developer can tell what's
really happening. There are some that can be read out with a common
hand-held voltmeter. I could look in a couple of catalogs if you want me
to.
I'm so thankful I picked the builder of my big dual Athlon machine
for their expertise in airflow.
I guess the thing that has to be done fairly soon, before parts
placement can get under way, is decide the space envelope allocated to the
on-board heat sink, and the method of attachment and conductive heat
transfer. Screw holes may have to be placed in the board for heat sink
mounting, and their sizes, locations, and keep-out zones must be included in
the dimension drawing at start of design. Don't forget to analyze the
tolerance stack-ups around the holes, and make the keep-out zones large
enough to contain screw heads, lock washers, and spacers at maximum
misalignment plus .015" clearance to the nearest live copper. I always
connect mounting hole pads to the ground plane.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
