[Sidefx-houdini-list] hardware innovations

Peter Bowmar pbowmar at gmail.com
Wed Dec 14 13:23:02 EST 2005


I would go for as many CPUs as possible, then as much RAM as possible,
and a really cheap graphics card. I still run an FX 5200 ($40 or less,
these days) which lets me run interactively just as well as a FX
Quadro 1400 that I have at school. On Linux of course, the Windows OGL
performance is crap, at least on the FX 5200.

One really important point that was touched on by Mark but needs
elaboration: With DOPs, you want to spend time with simulations.
However, unless you're a big studio, you can't send the DOPs sim off
to the queue. A dual or better system lets you run multiple instances
of Houdini/Hscript on a single license, which means a dual system
saves you significant money!

I have a dual Athlon 2000mp system (1.6 ghz) at home, with the cheap
FX 5200 card, and at school a single 3.6ghz P4 with FX Quadro 1400
card, and frankly, I can get more work done (on the exact same
project) at home. This changes when I am doing lighting since IPR
doesn't work multi-threaded :( and the faster CPU makes a big
difference.

Cheers,

Peter B

On 14/12/05, Peter Robbinson <probbins at corefa.com> wrote:
> I'd aim for stability over speed. You waste way more time with crashing
> hardware and software than you do somewhat slower screen refreshes.
>
> Directors are always going to find ways to choke the workflow to a
> crawl, no matter how fast your machines are.
>
> I'm looking towards off-loading as much of the work onto other hardware
> and keeping my workstations for the immediate work at hand.
> So, moving as much processing to the graphics card and finding the most
> efficient ways of sending renders to a farm.
>
> Part of my consideration have to do with power consumption as well, it's
> only going to get more expensive.
> Much of the work we do doesn't require a crt screen, so I'd shift to
> flat screens. Many machines don't necessarily have to be
> big iron power monsters, so I'm looking at desktop machines like the
> ShuttleX.
>
> PeterR
>
> Dan Seddon wrote:
>
> > John Coldrick wrote:
> >
> >> On Wednesday 14 December 2005 10:23, Jed Schwartz wrote:
> >>
> >>
> >>> I'm finding with the latest houdini builds (especially running
> >>> elaborate DOPs simulations-etc) I'm spending more and more time
> >>> waiting for things to cook- What is the best overall recipe for a
> >>> high- performance "houdini" system. Are the duel core processors a
> >>> huge advantage to sys performance beyond render-?
> >>>
> >>
> >>
> >>     Only with compositing.  AFAIK none of the other parts of Houdini
> >> are multithreaded except for mantra.  Certainly DOPs won't speed up.
> >> Personally,
> >>
> > Yeah - I've been wondering about that, it seems that Intel an AMDs
> > strategys are to go towards multicore processors in the future, as
> > they reach the upper limit on what they can raise clock speeds to.
> > Which sounds really bad for packages like Houdini, which aren't
> > multithreaded, as I gather it's very difficult migrate parts of the
> > application towards being multithreaded - especially solvers, which
> > you'd really hope would get faster as time goes on..... Looks like
> > long nights are here to stay!
> >
> > Dan
> > _______________________________________________
> > Sidefx-houdini-list mailing list
> > Sidefx-houdini-list at sidefx.com
> > https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> >
> >
> _______________________________________________
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>



More information about the Sidefx-houdini-list mailing list