Open GoogleCodeExporter opened 9 years ago
Hi Ghassan,
The GeForce GTX 760 is a good choice. I am currently using GTX 780's that are
very similar.
You will have to compile the code for CUDA Compute Capability 3.0 for optimum
performance.
I am still using CUDA version 5.0 but possibly the code will work fine with the
latest version of CUDA.
MC-GPU doesn't have any special hardware requirement, as long as you can
compile and run the sample codes distributed with the CUDA SDK (openGL is not
necessary for MC-GPU).
You will also need enough RAM memory in the CPU and the GPU to allocate the
data for the simulations, which is typically determined by the size of the
voxelized geometry file (note that for each voxel the program allocates memory
for the material number, density, deposited energy and deposited energy
squared). 2048MB of GPU memory should be enough unless you have a huge
super-high resolution phantom.
It is also a good idea to have a simple GPU for graphics while you use the good
GPU for computations, because CUDA has limitations on the computing time for
GPUs connected to a monitor (you can also turn off the Linux X windows and work
from the command line).
Best regards!
Andreu
Original comment by andre...@gmail.com
on 20 Nov 2013 at 7:44
Hi Andreu,
I have to choose between two graphic cards GeForce and their system
requirements.
1. the GTX 770 with 2GB compute capabilty 3.0 and system CPU 16 GB
2. the GTX 780 with 3GB compute capabilty 3.5 and CPU system 8 GB.
Is it necessary to choose the GTX 780 with 3GB compute capabilty 3.5 for
running the MCPGU?
Are there any problems to work with CUDA version 5.5?
Best regards
Ghassan
Original comment by ghassan....@gmail.com
on 27 Nov 2013 at 3:22
Original issue reported on code.google.com by
ghassan....@gmail.com
on 20 Nov 2013 at 1:08