<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class="">Dear Filippo,<br class=""><br class="">Thank you very much for the quick and informative reply.<br class=""><br class="">It seems that I’m out of luck then. However I would still be interested in testing the double precision capabilities of the two cards with QE-GPU if possible. Is there a workaround for the error I’m getting? Also do you know if there is a plan to support the Maxwell architecture in the future? I’m thinking whether the performance might be improved by “switching the GPU to TCC mode” as explained in the link below. It seems like TCC mode is supported for QUADRO M5000.<div class=""><br class=""></div><div class=""><a href="http://arrayfire.com/explaining-fp64-performance-on-gpus/" class="">http://arrayfire.com/explaining-fp64-performance-on-gpus/</a></div><div class=""><br class=""></div><div class="">Best regards,<br class="">Gunnar<br class=""><br class=""><blockquote type="cite" class="">On 12 May 2016, at 03:36, Dae Kwang Jun <<a href="mailto:jdaekwang@gmail.com" class="">jdaekwang@gmail.com</a>> wrote:<br class=""><br class="">Hello Gunnar,<br class=""><br class="">On May 11, 2016, at 4:15 PM, Gunnar Palsson <<a href="mailto:gunnar.karl@gmail.com" class="">gunnar.karl@gmail.com</a>> wrote:<br class=""><blockquote type="cite" class="">My question is: Is there a way to compile QE-GPU with the Maxwell architecture and if so how? I read on the forum that unfortunately the Maxwell architecture does not do double precision very well.<br class=""></blockquote><br class="">Maxell is not supported, you can force the compilation but as you pointed already in your email double precision is going to be bad.<br class=""><br class=""><br class=""><blockquote type="cite" class="">Is it a prohibitive loss of precision if one restricts the calculations to single precision?<br class=""></blockquote><br class="">Well... for the GPU implementation you simply cannot switch precision on-demand. QE-GPU reflects the QE implementation, if the original code si double precision than the GPU code is double precision. QE is all double precision so the switch cannot be done.<br class=""><br class="">I cannot comment on what it is going to happen if you switch everywhere from double to single (for some part it may work, for other may not), domain experts in the physics can give a proper answer to this. from an implementation point of you, again, cannot be done.<br class=""><br class="">HTH<br class=""><br class="">Cheers<br class=""><br class="">--<br class="">Mr. Filippo SPIGA, M.Sc.<br class="">Quantum ESPRESSO Foundation<br class=""><a href="http://www.quantum-espresso.org" class="">http://www.quantum-espresso.org</a> ~ skype: filippo.spiga<br class=""><br class="">*****<br class="">Disclaimer: "Please note this message and any attachments are CONFIDENTIAL and may be privileged or otherwise protected from disclosure. The contents are not to be disclosed to anyone other than the addressee. Unauthorized recipients are requested to preserve this confidentiality and to advise the sender immediately of any error in transmission."<br class=""><br class=""><br class="">_______________________________________________<br class="">Pw_forum mailing list<br class=""><a href="mailto:Pw_forum@pwscf.org" class="">Pw_forum@pwscf.org</a><br class="">http://pwscf.org/mailman/listinfo/pw_forum<br class=""></blockquote><br class=""></div></body></html>