k-Wave

1. qwen12
   

   Member
   Joined: Jun '18
   Posts: 2

   Hello!

   I knew that 2^26 grid size is the limitation of k-Wave GPU simulation due to the memory boundary, when we use 8GB or 12GB GPU such as TITAN X.
   However, I want to find the way to exceed this limitation (equal or more than 2^27). I thought using CPU is one of the way, but it seems to take a lot of time. The second way I thought was to buy 24 or 32 GB GPU (like NVIDIA quadro GV100 or m6000), but I don't know running k-Wave is possible with those workstation GPUs - quardo series. I will be really happy if someone helps me to find out the solutions :)

   Thank you

   Posted 5 years ago #

2. Jiri Jaros
   

   Developer
   Joined: Feb '12
   Posts: 115

   Hi qwen12,
   the code will work on NVIDIA quadro GV100. It might be necessary to recompile the code to support this GPU, but that should be trivial. 2^27 should fit within 24GBs.

   If you had a cluster of CPUs connected by a fast network, you could also try our distributed code which I can send to you. To beat a single GV100 you'd need at least 128 CPU cores.

   Best,
   Jiri

   Posted 5 years ago #

RSS feed for this topic

Reply

You must log in to post.