Flam4 - Nvidia CUDA Accelerated Render
Fractal Architect supports multiple plugin rendering engines.
- OpenCL running on CPU - all Intel Macs - Snow Leopard & Lion
- OpenCL running on GPU - many Macs - Lion only
- Flam4CUDA running on Nvidia GPU - Macs with Nvidia GPUs - Snow Leopard only
- Flam3 CPU - User installable - not bundled with app - Snow Leopard only
As distributed on the Mac App Store, it has a disabled Flam4CUDA renderer and instead uses the OpenCL CPU based renderer. The flam4 internal renderer is a NVidia CUDA based GPU accelerated renderer (which is not enabled in the App Store version of Fractal Architect). The App Stores imposes strict guidelines on what can be included with an application and does not allow inclusion of custom device drivers, such as the CUDA driver for Mac OS.
Flam4's Big Advantage over OpenCL CPU: Speed
Note: since OpenCL GPU is supported for Nvidia on Lion, Flam4CUDA is only useful on Snow Leopard.
Depending on the GPU and CPU found in the Mac, Flam4 can render from 1.5 to 20 times faster than the CPU based OpenCL renderer
The bigger the image the bigger the improvement. Small thumbnails' rendering performance is about the same.
Flam4's Disadvantages
- Images may not render the same on flam4 as compared to flam3.
If you find a discrepancy, please email the affected fractal document (not the rendered image) to: support@fractalarchitect.net - Flam4 needs to have lots of free GPU memory to work with. It is a good idea to close other applications and/or reboot before using the Flam4 renderer.
- Its all about the CUDA driver. Its a moving target and is under rapid development by Nvidia. The latest 4.0 drivers have compatibility issues with running the 2 GPUs in the 2008/2009 Mac Book Pro's. Only the very latest release of CUDA works on Mac OS X Lion.
Hardware Prerequisite - Mac with Nvidia GPU
The internal CUDA renderer can be only activated on Macs that have a NVidia GPU. The actual GPU incorporated into your Mac can be found in the About This Mac (found under the Apple menu - on the far left of any application's menu). The video card(s) is listed on the Graphics/Display page of that utility. It needs to be a CUDA capable NVidia GPU.
Obtaining CUDA
If you have a CUDA capable Mac, you will need to download the CUDA developer driver from NVidia. Nvidia Cuda Downloads Page
Enabling the Flam4 renderer
Finally after installing the correct version of CUDA, under Fractal Architect's Rendering Option Preference you must enable the CUDA GPU accelerated renderer. You can switch back to the OpenCL CPU renderer at any time.
Flam4 CUDA accelerated renderer and Snow Leopard
The Flam4 renderer is an optional feature available for use on Snow Leopard only.
- CUDA is supported only on Macs with Nvidia GPU's. The Flam4 rendering preference is not available on other Macs.
- Customers must download and install a Nvidia development CUDA driver. See: http://developer.nvidia.com/cuda-downloads Customers can ignore the CUDA Toolkit and CUDA SDK download options.
- CUDA driver versions 5.0.36 or greater are supported for CUDA GPU acceleration. (Mac Snow Leopard only)
- Flam4 renderer has a 35MB free GPU memory requirement. If you get a warning of low GPU memory, you may recover GPU memory by:
- Closing other applications that consume GPU memory. (Safari in particular. Games and other graphic editors.)
- If that is not enough, Log out/ log back into your Mac. Restarting the Mac is another option.
- The amount of free GPU memory (at the time of last fractal render) is shown at top right corner of Fractal Architect Log window.
Solving Problems
Find the answers to common problems with the Flam4 renderer.