<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Dear Reinaldo<br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
QE computations are double precision (fp64). Hence the card has to have excellent fp64 performance. Titan X pascal version has a relatively poor fp64 performance. The best card nowadays that offer best balance between cost and fp64 performance is the new Titan
V (volta) with a price around 3000 USD.<br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Bottom Line, my advice is that you put the gpu budget on a better CPU with more threads. That's much more effective than buying even the best gpu based on my limited experience.
<br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Yasser Al Wahedi <br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Assistant Professor <br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Khalifa University <br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Get <a href="https://aka.ms/ghei36">Outlook for Android</a></div>
<br>
<br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
From: Reinaldo Pis Diez<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Sent: Thursday, 15 February, 22:45<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Subject: [Pw_forum] best/better gpu cards for using with QE<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
To: PWSCF Forum<br>
<br>
<br>
</div>
<div dir="auto" style="direction: ltr; margin: 0; padding: 0; font-family: sans-serif; font-size: 11pt; color: black; ">
Dear folks I've managed to compile the gpu-version of QE gently provided by Filippo Spiga. However, the gpu card I've got (Titan X, Pascal architecture) has 11 gb of global memory so it goes easily out of memory when I try to simulate a medium system (say ~80
atoms) or at least these are the messages I obtain when running those jobs. Thus, I've got two questions concerning the use of gpu cards with QE: 1) What gpu cards are you using for serious runs/research? How much global memory do they have? I've recently
learnt that the Quatro P6000 model has about 24 GB of global memory but it costs about 5000 euros. 2) Does it make sense to combine more than one gpu card on one mother board? That is, is global memory additive in such cases? Sorry if this question is non-sense,
I'm a newbie in gpu stuff. Thanks in advance. Regards Reinaldo Pis Diez Center of Inorganic Chemistry Natl Univ of La Plata, Argentina _______________________________________________ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
<br>
<br>
</div>
</body>
</html>