Home Technical Talk

ATI R9 Fury with High Bandwidth Memory - 3D work benefits ?

interpolator
Offline / Send Message
Shrike interpolator
Now reading about the new AMD R9 Fury using 4 GB of the new HBM high bandwidth memory, I was wondering if there could be exeptional benefits for viewport performance or programs like Z Brush ? (3DS / Cinema 4D viewports - GPU based renderers like octane ?)

The Fury has average game performance for the price range, but people always just talk about games, never about work.

With 512 GB / s transfer rate (at only 4 GB of ram tho) it nearly doubles other current GPUs (320 on a Titan X) but is that even from benefit ? Does the polygonal calculation benefit from the higher rate or is it mainly about textures ? Could this be a hidden gem for 3D work or is it of no benefit over a traditional card of the same price range ?


Replies

  • SonicBlue
    Options
    Offline / Send Message
    SonicBlue polycounter lvl 10
    On CineBench R15 (OpenGL) it doesn't get higher frame rates than the GTX 980, and generally (in games) the frame rate is lower.


    Article, in Russian.

    ZBrush runs on the CPU and uses the system memory, not the VRAM, I suspect It can run well on Intel HD too; don't know about 3ds Max, but considering DX11 performance are lower in AMD cards compared to Nvidia counterpart, the investment on this card doesn't look like a good idea, also, at the moment Octane runs only on CUDA technology, so :(
  • Shrike
    Options
    Offline / Send Message
    Shrike interpolator
    Ah Zbrush is on CPU ? Handles it pretty well then

    So the same power relation as in games then afterall. The fast memory could be advantageous for Mari maybe I guess with heavy texture load, but the low 4 GB limit would probably still hold it back, else I cannot really think of another scenario.
  • SonicBlue
    Options
    Offline / Send Message
    SonicBlue polycounter lvl 10
    I think you'll get a good performance boost only in software that uses double precision floating point operations, like for example simulations that uses OpenCL.

    This depends on how the OpenCL implementation is done, I think Blender's Cycles is quite slow on OpenCL, compared to what you'll get with a CUDA card.
Sign In or Register to comment.