<div dir="ltr"><div><div><div><div>Hello developers,<br></div>Have a look at the following thread:<br><br><a href="http://blenderartists.org/forum/showthread.php?254521-A-good-news-for-AMD-ATI-Graphic-cards-owners/page53">http://blenderartists.org/forum/showthread.php?254521-A-good-news-for-AMD-ATI-Graphic-cards-owners/page53</a><br>
<br></div>thanks to recent developments in the amd opencl drivers. amd gpu users are beginning to see results in compiling and rendering scenes on AMD GPUs....(such as hair )?<br></div><br></div><div>AND <br></div><div><br>
Could future developments in cycles be optimized for opencl....<br><br></div><div>What one of the blenderheads mention of opencl on amd gpus is<br><br>'Looks really good. I would expect a good speedup if we ever get the
artificially ability to limit the ammount of registers we allow the
OpenCL compiler to use. For cuda we depend on this to get decent
performance.<br>
Cycles is so big / complex that the compiler will use all registers if
we do not limit this and that decreases the amount of threads we can run
by a factor of 4-8 depending on hardware. If we ever get this
performance should be much more in line with cuda but it will need code
changes as the current code has no way of doing this for non Nvidia
hardware.'<b>JUICYFRUIT</b>..... </div><div><br></div><div>Hopefully, with this new development...<br></div><div>Everyones favourite render engine could soon see itself being multiplatform.....<br><br></div><div>Kind Regards....<br>
<br></div><div>Nazim Mer<br></div></div>