No announcement yet.

Imagination Posts Original Driver Code For PowerVR Series 1 GPUs As Open-Source

  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by ArthurBorsboom View Post
    Still having a gma500 waiting for some love...
    I got one Poulsbo in my storage it was always horrible.


    • #22
      I'm still waiting a real driver for the GMA36x0 with the SGX545 gpu for a mini-itx board I'm still using and would like to see it really accelerating GUI and opengl driver working on it. Obviously I suppose it will never happen, I've read all the old discussions about it and tested this gpu with its best driver on Win 32bit as much as possible but in linux there was/is not much to test beside the gma500 module that is great still living but not accelerating much unfortunately and no OpenGL or any HD video acceleration support while the GPU could have done more.

      Some interesting things found during tests. The cpu being technically a dual core x64 one supported theorically "up to 4GB of ram" while limited to 3GB on x86 o.s. but in some case (I suppose depending on the board revision/bios) it surpassed even the specifications.
      I have a mini-itx board where two [email protected] rams can work for an incredible 8GB of single channel total ram both in Win and Linux. I suppose the x64 cpu usage and the double slot ram improved a bit the final speed.

      A PCI external video bus permit to "solve" the iGPU support problem but it add another problem too imho; on these mini-itx configs the PCI old bus I imagine might be bridged from a PCI-EX(chipset) connection and the last PCI video cards all had their PCI to PCI-EX(GPU) IC bridge too (different versions existed). So while they do work and on bench faster than the on board GPU when used in Win7 32bit, there's imho a bandwidth/latency aspect to consider that would have been solved simply using the native existing PCI-EX (single x4 or multiple x1 versions I think) connection with a PCI-EX full mainboard bus connector.
      Instead using the PCI bus (most boards were similar to this PCB layout, only few had PCI-EX mini or standard connectors) all this communication task seems to add another limiting factor to an already slow system. So the external VGA could have been a solution and it might be considered this unfortunately. I suppose the same problem exists using these PCI video cards on any modern last PCI mainboards but faster system can benefit imho from faster general speed mitigating this if only a PCI bus existed at least from my test using a Core2 much faster config.