AMD demonstrating Havok on Radeon GPUs at GDC

Discussion in 'GPGPU Technology & Programming' started by Kaotik, Mar 20, 2009.

  1. Kaotik

    Kaotik Drunk Member
    Legend

    Joined:
    Apr 16, 2003
    Messages:
    8,254
    Likes Received:
    1,940
    Location:
    Finland
    When did "nvidia decide to quit" havokfx? as far as I can remember, it was intel that pulled the plug after buying havok
     
  2. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    yeah, though Havok's physics would be limited (all sm3.0 shaders) it would still allow for faster processing of anything that is normally done on the cpu client of havok and that served me well wherever it's applied.

    whatever killed HFX.. I honestly don't know for sure.
     
  3. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    Oh, I dunno, seems as valid as outsiders arguing over what motivates AMD's strategic direction :razz:

    Pissing in the wind, passing the time of day, learning new things occasionally.

    I've already made a comment in this thread about how technically inferior the blue coat looks (look Ma! you can see the squares!). My view has got nothing to do with the artistry - though I do have a soft spot for dance and the colour red :lol:

    Jawed
     
  4. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,436
    Likes Received:
    443
    Location:
    New York
    Yeah I saw that comment and that's what I referred to by the human factor. Or are you implying that PhysX itself is limited in terms of control point density in the mesh?
     
  5. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    PhysX appears to be using a much coarser mesh, yes.

    Do you remember the original PhysX videos, particularly the liquids? Do you remember how the liquids looked like gloopy pools of spheres rather than liquids (bringing back memories of traditional porridge in my case)?

    The red dress looks a bit gloopy, but the blue coat looks notably patchy. Both demos have 25 figures and the dress appears to have considerably more points in its mesh.

    Jawed
     
  6. aaronspink

    Veteran

    Joined:
    Jun 20, 2003
    Messages:
    2,641
    Likes Received:
    64
    Sure there are artistic differences, but the differences in the meshes/fluidity of the cloth is miles apart. The physx cloth looks chunky, while the havok cloth looks like it should.
     
  7. CNCAddict

    Regular

    Joined:
    Aug 14, 2005
    Messages:
    290
    Likes Received:
    2
    Hmmm, this the best cloth I've ever seen. The self interaction is just mind blowing (watch till the end)

    http://www.matthiasmueller.info ->videos ->Position Based Dynamics

    He's got a nice paper on it too, but it's beyond me :(
     
  8. Davros

    Legend

    Joined:
    Jun 7, 2004
    Messages:
    15,035
    Likes Received:
    2,381
    bit-tech.net reports NVIDIA is considering porting PhysX to OpenCL to allow their physics technology to be used by other manufacturers. They quote Nadeem Mohammad, NVIDIA director of product management for PhysX, saying: "In the future it’s a possibility that we could use OpenCL," noting that a later comment about future plans includes the quote: "as and when we do move to OpenCL." Mohammad points out NVIDIA has worked closely with The Khronos Group to help shape the OpenCL specification.
    http://www.bit-tech.net/news/hardware/2009/03/27/nvidia-considers-porting-physx-to-opencl/1
     
  9. CJ

    CJ
    Regular

    Joined:
    Apr 28, 2004
    Messages:
    816
    Likes Received:
    40
    Location:
    MSI Europe HQ
    Pssst. You have a habit of posting news that has already been posted before don't you? ;)

    http://forum.beyond3d.com/showpost.php?p=1281572&postcount=85
     
  10. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    Yeah those cloth effects look much better. The tearing is particularly nice and the "washing machine".

    The thing that bothers me in these demos (physics in general) is that objects often seem to have an unrealistically low mass/weight and consequently much less inertia than we're used to experiencing with real life objects.

    The bit in one of the Havok demos where a tank crashes into huge bolders and the boulders go flying like bits of foam is particularly crap.

    With the cloth demos it seems there's a complete lack of damping in the "springs", which makes them behave like sheets of rubber, not woven cloth.

    Davros, this is becoming a bad habit. Read the thread :lol:

    Jawed
     
  11. CNCAddict

    Regular

    Joined:
    Aug 14, 2005
    Messages:
    290
    Likes Received:
    2
    Yeah, I agree. Surely they can just tweak some of the material properties to more closely model real life. It would also be easy to apply drag properties to materials so light materials don't fly 600 feet from a grenade going off, and heavy ones don't bounce like a rubber ball after hitting the comparatively soft ground.

    We're getting to an uncanny valley when it comes to physics. The better the graphics look, the more we notice the crappy animation and physics.
     
  12. nexus_alpha

    Newcomer

    Joined:
    Nov 25, 2006
    Messages:
    44
    Likes Received:
    0
    The demo must have several parameters that can be tweaked acheive more believable results.
     
  13. MfA

    MfA
    Legend

    Joined:
    Feb 6, 2002
    Messages:
    6,836
    Likes Received:
    481
    Has PhysX incorporated this method yet though? (It was impulse based to begin with right?)

    It's funny to see the movement from velocity to impulse to location ... each step makes seems to make it more hacky and rule based, also it seems to make things look better at each step.
     
    #113 MfA, Mar 31, 2009
    Last edited by a moderator: Mar 31, 2009
  14. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,436
    Likes Received:
    443
    Location:
    New York
    Yeah I agree with both you and Jawed on what we're seeing here. Just musing as to whether it's an underlying limitation of the API, a hardware performance limitation or a developer skill issue. Can't think of a reason for an API to be artificially limited like that. It should be scalable up to some arbitrarily high density.
     
  15. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
  16. Silent_Buddha

    Legend

    Joined:
    Mar 13, 2007
    Messages:
    16,343
    Likes Received:
    5,319
    Now that OpenCL has been released, I'm wondering if we'll start seeing most of the current consumer CUDA apps being ported over to OpenCL for future developement.

    Regards,
    SB
     
  17. Davros

    Legend

    Joined:
    Jun 7, 2004
    Messages:
    15,035
    Likes Received:
    2,381
    I would also like to ask will my games that use havok now do physics on the gpu or will they need to be patched
     
  18. rpg.314

    Veteran

    Joined:
    Jul 21, 2008
    Messages:
    4,298
    Likes Received:
    0
    Location:
    /
    ptached, I think. you have to specefically create a gpu context in OpenCL.
     
  19. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    10,436
    Likes Received:
    443
    Location:
    New York
    Would that even be worthwhile? Are they any CPU intensive Havok games that could benefit from offloading physics calcs to the GPU?
     
  20. pcchen

    pcchen Moderator
    Moderator Veteran Subscriber

    Joined:
    Feb 6, 2002
    Messages:
    2,772
    Likes Received:
    153
    Location:
    Taiwan
    There may be none currently (I'm not sure about this), but if GPU accelerated physics become more common, game developers will use it more.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...