PowerVR VS 3.0 Demo!!!

Discussion in 'Architecture and Products' started by radiATIon, May 23, 2004.

  1. 991060

    Regular

    Joined:
    Jul 29, 2003
    Messages:
    640
    Likes Received:
    2
    Location:
    Beijing
    Interesting, thanks Ante P
     
  2. 991060

    Regular

    Joined:
    Jul 29, 2003
    Messages:
    640
    Likes Received:
    2
    Location:
    Beijing
    I just got this reply from nVIDIA concerning the errors in the demos:
     
  3. hstewarth

    Newcomer

    Joined:
    Apr 13, 2004
    Messages:
    99
    Likes Received:
    0

    To me this sounds like its not running on Direct X 9.0c.. We may have to wait to DX9.0c before it can this topic is a real issue or not. Also if the demo detects the ID of video card for special effect - then they maybe never the same.
     
  4. Druga Runda

    Druga Runda Sleepy Substitute
    Regular

    Joined:
    Sep 30, 2003
    Messages:
    647
    Likes Received:
    43
    and shapeshifting in front of you...

    man those people do love their jobs :)
     
  5. ActionNews

    Newcomer

    Joined:
    May 14, 2002
    Messages:
    59
    Likes Received:
    0
    Location:
    Germany
    Do you mean PowerVR hardware running DX9 Demos :shock: :?: Well you don't need DX9 for SEGAs Arcarde :twisted: , or am i completly wrong ;)?

    CU ActionNews
     
  6. Tim Murray

    Tim Murray the Windom Earle of mobile SOCs
    Veteran

    Joined:
    May 25, 2003
    Messages:
    3,278
    Likes Received:
    66
    Location:
    Mountain View, CA
    refrast, I assume, is whathe means by reference device.
     
  7. Mintmaster

    Veteran

    Joined:
    Mar 31, 2002
    Messages:
    3,897
    Likes Received:
    87
    Kristof, can you confirm what NVidia was saying above? Is the refrast giving you incorrect images with your updated voxel demo?
     
  8. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    At least both give different images on the refrast than on the nv40 / 61.32(which has sometimes a known performance bug and the latest HLSL isn't also free of bugs, so maybe it's a little bit early to do some real benchmarks). The Cloth demo is nearly working, after some small mods(only minor stuff, driver-bug/prob. releated), VSDisplayPhaseB.vsh seems to be like the problem shader - so no cloth in the final image :(

    3D-Analyse isn't 3_0 ready yet, but I'am working on it.

    Thomas
     
  9. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    We assume that NVIDIA modified an old version of Voxel themselves to work on the new runtime and as part of those modifications broke the demo. The conversion we did (just replace texld with texldl) works just fine and looks exactly as before as far as I know. I definitely know that cloth should be working fine.

    Basically if you download the new voxel version from our pvrdev website it should be working correctly on refrast and any hardware which supports SM3.0. If anyone thinks a result obtained on the refrast is wrong send a screenshot to devrel and we'll look into it.

    K-
     
  10. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    Any changes other than the texture format change to R16G16F ?

    VSDisplayPhaseB contains the actual vertex texturing instructions so that is pretty much the base functionality that is falling over then ? Could be driver issues with multiple samplers, I use 3 samplers (textures) in that demo. The only other thing I thought of is that I am not setting the actual filtering mode myself, D3D default is point sampling though so it should not be an issue.

    K-
     
  11. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    I changed every clear(D3DCLEAR_ZBUFFER) / clear(D3DCLEAR_TARGET) to clear (D3DCLEAR_ZBUFFER|D3DCLEAR_TARGET) and D3DFMT_R16F to D3DFMT_G16R16F.

    Thomas
     
  12. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    Hmmm... <shrug>
     
  13. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    the other views (Depth/Position/Normal) seems to be okay now....
     
  14. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    hstewarth we provide source code for that demo so you can see exactly what is being done, and there definitely is no video card ID detection code in there... AFAIK all PowerVR demos released so far are written for standard APIs and will work on any card that meets the minimum caps spec required.

    K-
     
  15. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    setteing the render tagets do D3DFMT_A32B32G32R32F helps a little bit, the cloth is now displayed, but a little bit wrong....

    Thomas
     
  16. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    Thats odd... a waste of storage space and bandwidth then ? Or are you redistributing the data ? Anyway definitely sounds like driver issues.

    Do you have any screenshots to illustrate the "little bit wrong" ?

    K-
     
  17. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    Just changed the format and the clear's, nothing else.

    [​IMG]

    Thomas
     
  18. Kristof

    Regular Alpha

    Joined:
    Jan 30, 2002
    Messages:
    733
    Likes Received:
    1
    Location:
    Abbots Langley
    Interesting effect. Definitely looks like the normals are wrong, the position is hard to say since all kinds of accuracy effects could come into play especially since we have been fiddling with the texture formats.

    Actually out of curiosity could you check using caps viewer under adapter formats, any of them, and then subcategory texture formats and then check for which formats they expose D3DUSAGE_DMAP. I wonder if NV40 can only do vertex texturing from a very small subset of textures, possibly they must have 4 components for the vertex texturing to work ?

    Bit more work would be to use the following :

    Code:
    An application can query if a format is supported as a vertex texture by calling IDirect3D9::CheckDeviceFormat with D3DUSAGE_QUERY_VERTEXTEXTURE. 
    Still if you'd try to use an unsupported format why is the call not failing ? Unless I am not checking the return code but even then it should still show up in the debug stream when you enable full dx debug level. Anything showing up there ?

    K-
     
  19. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    D3DUSAGE_DMAP isn't supported on any format. D3DUSAGE_QUERY_VERTEXTEXTURE is only supported for D3DFMT_R32F and D3DFMT_A32B32G32R32F.

    Thomas
     
  20. tb

    tb
    Newcomer

    Joined:
    Feb 7, 2002
    Messages:
    241
    Likes Received:
    0
    Location:
    Germany / Thuringia
    new drivers helped ;)

    [​IMG]
    [​IMG]

    Thomas
     
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...