http://www.k-wave.org/forum/topic/can-k-wave-support-gpus-connected-with-nvlink Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 22:32:51 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/can-k-wave-support-gpus-connected-with-nvlink#post-8403 Tue, 04 Jan 2022 08:04:53 +0000 luni.jimmy 8403@http://www.k-wave.org/forum/ <p>Hi Brad,</p> <p>Thanks for your reply and the valuable advice!<br /> We haven't decided to use NVLink yet.<br /> If anyone tried please share the experience, thanks again!</p> <p>Jimmy </p> http://www.k-wave.org/forum/topic/can-k-wave-support-gpus-connected-with-nvlink#post-8365 Thu, 16 Dec 2021 10:26:35 +0000 Bradley Treeby 8365@http://www.k-wave.org/forum/ <p>Hi luni,</p> <p>In general, I'd advise using <code>kspaceFirstOrder3DG</code> to run the native CUDA code - it will be faster and more memory efficient than using the MATLAB code with the <code>gpuArray</code> option.</p> <p>Coming back to your question, I'm not sure - I don't have a multi-GPU system to hand that has NVLink. If you do try, please post back and let us know!</p> <p>Brad. </p> http://www.k-wave.org/forum/topic/can-k-wave-support-gpus-connected-with-nvlink#post-8351 Wed, 24 Nov 2021 18:34:48 +0000 luni.jimmy 8351@http://www.k-wave.org/forum/ <p>Hi, all,</p> <p>When running simulations on the GPU using the MATLAB parallel computing toolbox ("gpuArray"), the maximal number of grid points is limited by GPU memory.</p> <p>Since there are 2 RTX 2080 Ti GPUs are available in our lab and they support NVLink, we're considering purchasing a compatible bridge to connect them.</p> <p>However, whether or not they can be used as a single GPU (with nearly doubled resources) also depends on software support.</p> <p>MATLAB's documentation suggests that its gpuArray variables supports NVLink:<br /> <a href="https://www.mathworks.com/help/parallel-computing/gop.html" rel="nofollow">https://www.mathworks.com/help/parallel-computing/gop.html</a><br /> "If x is a gpuArray, data transfer between multiple GPUs in a parallel pool uses fast peer-to-peer communication, including NVLink, if available."<br /> But we're not sure if the whole simulation process will be supported.</p> <p>Has anyone tried this before? Thanks in advance! </p>