Re: SGI using ATI Chips? - SGI

This is a discussion on Re: SGI using ATI Chips? - SGI ; Ian Mapleson wrote: > I do hope the PR improves though. At the moment SGI still isn't boasting about > so many things it could. I often do a simple demo on my SGIs when friends > visit, running 100+ ...

+ Reply to Thread
Page 2 of 2 FirstFirst 1 2
Results 21 to 25 of 25

Thread: Re: SGI using ATI Chips?

  1. Re: SGI using ATI Chips?

    Ian Mapleson wrote:
    > I do hope the PR improves though. At the moment SGI still isn't boasting about
    > so many things it could. I often do a simple demo on my SGIs when friends
    > visit, running 100+ apps at once while doing a CD write and multiple disk
    > operations. Never crashes, never harms the CD write, etc. And one can close
    > down all the apps afterwards without causing chaos. This is just flatly
    > impossible on a PC. In practical terms, it means being able to do multipe
    > things at once without the system slowing to a crawl and with the confidence
    > that the machine isn't going to go bellyup.


    This is the perfect example of why SGI Workstations ROCK! When I got my O2
    (180Mhz R5kPC) I was amazed by how smooth everything was. Really the only
    problem with it is that when playing MP3's, if it's doing to much it gets
    choppy, but lets face it the R5kPC CPU sucks. The O2 R12k/270Mhz I have is
    really a dream, I just wish I had more of an excuse to use it. I tried
    switching to it for my main system, but the Apps just aren't there, but then
    I'm more into DTP than 3D graphics. I just wish I had the room to give it
    it's own monitor, if I did, I'd use it a *lot* more than I do.

    I've got to admit, I'm wondering how Apple's new G5 systems will handle a
    *heavy* load. Still, although I'm finally learning to like Mac OS X, IRIX
    beats it hands down. There are just two many things that Mac OS X does
    wrong (though 10.2 is better than previous versions).

    Zane

  2. Re: SGI using ATI Chips?

    "Zane H. Healy" wrote in message news:...

    > I tried switching to it for my main system, but the Apps just aren't there, > but then I'm more into DTP than 3D graphics. I just wish I had the room to > give it it's own monitor, if I did, I'd use it a *lot* more than I do.


    You wouldn't have that much fun with 3D on an O2 - even my 300MHz MXE
    Octane slows down when handling complex wireframe objects in
    Softimage¦3D (which is still based on native Irix code unlike that
    godawful emulation piece of **** XSI). Same for Maya imo: purely from
    a subjective gut-feel point of view that software peaked speed- and
    usability-wise with version 4. The later versions seem to behave a bit
    gooey.

    As for DTP apps, have you tried Framemaker? There is still a
    reasonable new version available for Irix (5.5.6+SGML) and afaik Adobe
    will still issue a license for you. Imho the best WYSIWYG program for
    complex layout tasks with lots of formulas. But if you're serious
    about proper layout, you would use TeX anyway

  3. Re: SGI using ATI Chips?

    Ian Mapleson wrote:
    >

    [...]
    >
    > The modeling people I support at work do not need 48bit colour or OGL ARB
    > imaging. They need *speed*.


    I think the Onyx4 datasheet says it supports ARB_imaging. Nvidia doesn't
    accelerate it yet, at least not in 5200FX with latest drivers (it
    supports it in software and, btw, it's surprisingly slow for the speed
    of current PC processors). I don't know about ATI, but I believed they
    also emulated ARB_imaging in software. So, either the ATI cores used in
    Onyx4 have this as a new feature, or SGI implemented it through the
    programmability of such cores (another possibility would be that SGI
    supports it in software, but that would be a big disappointment).

    Regarding the 32bit framebuffer, I agree it's not a problem for the
    people you mention, but it's a limitation for the work I do (I use
    IMPACT often for rendering the presentations I do with Pegamento, just
    for the extra 4 bits it supports if the composition doesn't require an
    alpha channel, although it's of course limited to 32bit when I need
    destination alpha).

    Onyx4 supports FP offscreen buffers, but... if it uses the same
    extension as Nvidia, it can be useless: Many features like blending and
    texture filtering are not available for FP buffers (the "workaround" is
    to implement such features in non-standard ways through
    programmability).

    However, IIRC the Quake development team said they'd like to have
    framebuffers deeper than 32bit. If Quake needs it, the world will need
    it, so it could perhaps be implemented in future Nvidia/ATI generations.

    César
    cesarblecua@ono.com

  4. Re: SGI using ATI Chips?

    > This is a bit different. Sgi using Ati graphics is like HP selling Epson
    > printers. If they'll fit the Tezro with Ati graphics, why should a customer pay
    > $20,000 when they can get a PC with the same graphics for much less ?
    > Using commodity hardware hasn't worked well for Sgi so far.
    >
    > I sure hope they'll still do their own graphics. It's what makes it a Silicon
    > Graphics workstation. VPro V12 may be getting old, but I can't imagine that
    > they haven't done anything in the 2 years that it was released.
    >
    > Mark


    Don't forget you are paying for the benefits of the whole architecture
    of an sgi system
    It's not going to be some radeon 9xxx in there, I'm sure!
    IMO I would of been happier to see an nVidia chip in there, but that's
    down to personal taste

  5. Re: SGI using ATI Chips?

    > On the other hand, SGI's 12bit/component is 100% standard OpenGL. No dirty
    > workarounds needed.


    I was talking about the hardware (I program it with DirectX 9.0), not
    OpenGL, and mipmapping is definitely supported. You are right about alpha
    blending, and the explanation is _NOT_ detailed at all why it isn't
    supported. The only reason I can see why alpha blending isn't supported, is,
    that the source and destination factors (rgb and a) are not between 0.0 and
    1.0, since the values are allowed to be negative AND larger than 1.0 -- so
    infact it's IMPOSSIBLE that the traditional blending formulas WOULD WORK AT
    ALL (the way they are intended to work).

    Vertex- and Pixelshaders are easy to develop, and very, very flexible and
    powerful. I'm not familiar with GL fragment programs, but I would be keen to
    believe that same feature set more or less is supported.


    --
    Totally Gay
    =(|)= vs. (_*_)



+ Reply to Thread
Page 2 of 2 FirstFirst 1 2