I don't know if this has been posted before, but it's certainly very news worthy:
http://www.pcper.com/article.php?aid=607&type=expert&pid=1
http://anandtech.com/tradeshows/showdoc.aspx?i=3379
http://anandtech.com/video/showdoc.aspx?i=3385
http://www.techreport.com/articles.x/15367
In summary: This Hyrda chip intercepts DirectX and OpenGL calls to the GPU driver, determines which GPU's render what, then sends the balanced workloads to the GPU driver. The GPUs then render the image, then send it back to the Hydra chip for display. All the while achieving linear scaling (they demo'd Crysis at 60 FPS with two 9800 GT's). You can't mix and match different branded cards (ie HD4870 + GTX 280 is a no no), as the OS only accepts one display driver.
That's my very simple, broad summary. I recommend that everyone reads it for themselves, as this certainly sounds like the 'Holy Grail' of multi-GPU systems. Performance scales linearly from 1, 2, 3 and even 4 GPU's!
Wesker776
And so it would be possible that 2 HD4870x2 or 4 4870 in CF/ Tri GTX280 SLI to give 4x times HD4870 performance / 3x timesGTX280 performance respectively. :shock:
CBPCU
Log in to comment