<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html><head><title></title>
<META http-equiv=Content-Type content="text/html; charset=utf-8">
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
body {
margin: 5px 5px 5px 5px;
background-color: #ffffff;
}
/* ========== Text Styles ========== */
hr { color: #000000}
body, table /* Normal text */
{
font-size: 9pt;
font-family: 'Courier New';
font-style: normal;
font-weight: normal;
color: #000000;
text-decoration: none;
}
span.rvts1 /* Heading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #0000ff;
}
span.rvts2 /* Subheading */
{
font-size: 10pt;
font-family: 'Arial';
font-weight: bold;
color: #000080;
}
span.rvts3 /* Keywords */
{
font-size: 10pt;
font-family: 'Arial';
font-style: italic;
color: #800000;
}
a.rvts4, span.rvts4 /* Jump 1 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
a.rvts5, span.rvts5 /* Jump 2 */
{
font-size: 10pt;
font-family: 'Arial';
color: #008000;
text-decoration: underline;
}
span.rvts6
{
font-size: 11pt;
font-family: 'segoe ui';
font-weight: bold;
color: #ffffff;
background-color: #0000ff;
}
span.rvts7
{
font-size: 11pt;
font-family: 'segoe ui';
}
a.rvts8, span.rvts8
{
font-size: 11pt;
font-family: 'segoe ui';
color: #0000ff;
text-decoration: underline;
}
span.rvts9
{
font-size: 11pt;
font-family: 'segoe ui';
color: #888888;
}
span.rvts10
{
font-size: 8pt;
font-family: 'arial';
font-style: italic;
color: #c0c0c0;
}
a.rvts11, span.rvts11
{
font-size: 8pt;
font-family: 'arial';
color: #0000ff;
text-decoration: underline;
}
/* ========== Para Styles ========== */
p,ul,ol /* Paragraph Style */
{
text-align: left;
text-indent: 0px;
padding: 0px 0px 0px 0px;
margin: 0px 0px 0px 0px;
}
.rvps1 /* Centered */
{
text-align: center;
}
--></style>
</head>
<body>
<p>I could be mixing something up, so let me double check before sharing inputs.</p>
<p><br></p>
<p>Alex</p>
<p><br></p>
<div><table border=0 cellpadding=1 cellspacing=2>
<tr valign=top>
<td width=11 style="background-color: #0000ff;">
<p><span class=rvts6>></span></p>
</td>
<td width=1215 style="background-color: #ffffff;">
<p><span class=rvts7>I'm curious what are the conditions under which you get such a exceptional speedup. Can you share your input files and/or log files?</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>--</span></p>
<p><span class=rvts7>Szilárd</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>On Thu, Jul 2, 2015 at 2:18 AM, Alex <</span><a class=rvts8 href="mailto:nedomacho@gmail.com">nedomacho@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>Yup, about 7-8 times between with and without GPU acceleration, not making this up: I had 11 ns/day and now ~80-87 ns/day, the numbers vary a bit. I've been getting a similar boost on our GPU-accelerated cluster node (dual core i7, 8 cores each) with two Tesla C2075 cards (I am directing my simulations to one of them via -gpu_id). </span></p>
<p><span class=rvts7>All runs are -ntomp 4, with or without GPU. The physics in all cases is perfectly acceptable. So far I only tested my new box on vacuum simulations, about to run the solvated version (~30K particles).</span></p>
<p><br></p>
<p><span class=rvts9>Alex</span></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>On Wed, Jul 1, 2015 at 6:09 PM, Szilárd Páll <</span><a class=rvts8 href="mailto:pall.szilard@gmail.com">pall.szilard@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>Hmmm, 8x sounds rather high, are you sure you are comparing to CPU-only runs that use proper SIMD optimized kernels?</span></p>
<p><br></p>
<p><span class=rvts7>Because of the way offload-based acceleration works, the CPU and GPU will inherently be executing concurrently only part of the runtime and as a consequence the GPU is idle part of the run-time (during integration+constraints). You can make use of this idle time by running multiple independent simulations concurrently. This can yield serious improvements in terms of _aggregate_ simulation performance especially with small inputs and many cores (see slide 51 </span><a class=rvts8 href="https://goo.gl/7DnSri)/">https://goo.gl/7DnSri)/</a></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>--</span></p>
<p><span class=rvts7>Szilárd</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>On Wed, Jul 1, 2015 at 4:16 AM, Alex <</span><a class=rvts8 href="mailto:nedomacho@gmail.com">nedomacho@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>I am happy to say that I am getting an 8-fold increase in simulation speeds for $200.</span></p>
<p><br></p>
<p><span class=rvts7>An additional question: normally, how many simulations (separate mdruns on separate CPU cores) can be performed simultaneously on a single GPU? Say, for 20-40K particle sized simulations.</span></p>
<p><span class=rvts7>The coolers are not even spinning during a single test (mdrun -ntomp 4), and I get massive acceleration. They aren't broken, the card is just cool (small system, ~3K particles). </span></p>
<p><br></p>
<p><span class=rvts7>Thanks,</span></p>
<p><br></p>
<p><span class=rvts7>Alex</span></p>
<p><br></p>
<p><br></p>
<div><table border=0 cellpadding=1 cellspacing=2>
<tr valign=top>
<td width=15 style="background-color: #0000ff;">
<p><span class=rvts7>></span></p>
</td>
<td width=1128 style="background-color: #ffffff;">
<p><br></p>
<p><br></p>
<p><br></p>
<div><table border=0 cellpadding=1 cellspacing=2>
<tr valign=top>
<td width=16 style="background-color: #0000ff;">
<p><span class=rvts7>></span></p>
</td>
<td width=1104 style="background-color: #ffffff;">
<p><span class=rvts7>Ah, ok, so you can get a 6-pin from the PSU and another from a converted molex connector. That should be just fine, especially as the card should will not pull more than ~155W (under heavy graphics load) based on the Tomshardware review* and you are providing 225W max.</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>*</span><a class=rvts8 href="http://www.tomshardware.com/reviews/evga-super-super-clocked-gtx-960,4063-3.html">http://www.tomshardware.com/reviews/evga-super-super-clocked-gtx-960,4063-3.html</a></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>--</span></p>
<p><span class=rvts7>Szilárd</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>On Tue, Jun 30, 2015 at 7:31 PM, Alex <</span><a class=rvts8 href="mailto:nedomacho@gmail.com">nedomacho@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>Well, I don't have one like this. What I have instead is this:</span></p>
<p><br></p>
<p><span class=rvts7>1. A single 6-pin directly from the PSU.</span></p>
<p><span class=rvts7>2. A single molex to 6-pin (my PSU does provide one molex).</span></p>
<p><span class=rvts7>3. Two 6-pins to a single 8-pin converter going to the card.</span></p>
<p><br></p>
<p><span class=rvts7>In other words, I can populate both 6-pins on the 6-8 converter, just not sure about the pinouts in this case.</span></p>
<p><br></p>
<p><span class=rvts7>Not good?</span></p>
<p><br></p>
<p><span class=rvts7>Alex</span></p>
<p><br></p>
<p><br></p>
<div><table border=0 cellpadding=1 cellspacing=2>
<tr valign=top>
<td width=15 style="background-color: #0000ff;">
<p><span class=rvts7>></span></p>
</td>
<td width=1081 style="background-color: #ffffff;">
<p><span class=rvts7>What I meant is this: </span><a class=rvts8 href="http://goo.gl/8o1B5P">http://goo.gl/8o1B5P</a></p>
<p><br></p>
<p><span class=rvts7>That is 2x molex -> 8pin PCI-E. A single molex may not be enouhg.</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>--</span></p>
<p><span class=rvts7>Szilárd</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>On Tue, Jun 30, 2015 at 7:10 PM, Alex <</span><a class=rvts8 href="mailto:nedomacho@gmail.com">nedomacho@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>It is a 4-core CPU, single GPU box, so I doubt I will be running more</span></p>
<p><span class=rvts7>than one at a time. We will very likely get a different PSU, unless...</span></p>
<p><span class=rvts7>I do have a molex to 6 pin concerter sitting on this very desk. Do you</span></p>
<p><span class=rvts7>think it will satisfy the card? I just don't know how much a single</span></p>
<p><span class=rvts7>molex line delivers. If you feel this should work, off to installing</span></p>
<p><span class=rvts7>everything I go.</span></p>
<p><br></p>
<p><span class=rvts7>Thanks a bunch,</span></p>
<p><span class=rvts7>Alex</span></p>
<p><br></p>
<p><span class=rvts7>SP> First of all, unless you run multiple independent simulations on the same</span></p>
<p><span class=rvts7>SP> GPU, GROMACS runs alone will never get anywhere near the peak power</span></p>
<p><span class=rvts7>SP> consumption of the GPU.</span></p>
<p><br></p>
<p><span class=rvts7>SP> The good news is that NVIDIA has gained some sanity and stopped blocking</span></p>
<p><span class=rvts7>SP> GeForce GPU info in nvidia-smi - although only for newer cars, but it does</span></p>
<p><span class=rvts7>SP> work with the 960 if you use a 352.xx driver:</span></p>
<p><span class=rvts7>SP> +------------------------------------------------------+</span></p>
<p><br></p>
<p><span class=rvts7>SP> | NVIDIA-SMI 352.21 Driver Version: 352.21 |</span></p>
<p><br></p>
<p><span class=rvts7>SP> |-------------------------------+----------------------+----------------------+</span></p>
<p><span class=rvts7>SP> | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr.</span></p>
<p><span class=rvts7>SP> ECC |</span></p>
<p><span class=rvts7>SP> | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute</span></p>
<p><span class=rvts7>SP> M. |</span></p>
<p><span class=rvts7>SP> |===============================+======================+======================|</span></p>
<p><span class=rvts7>SP> | 0 GeForce GTX 960 Off | 0000:01:00.0 On |</span></p>
<p><span class=rvts7>SP> N/A |</span></p>
<p><span class=rvts7>SP> | 8% 45C P5 15W / 130W | 1168MiB / 2044MiB | 31%</span></p>
<p><span class=rvts7>SP> Default |</span></p>
<p><span class=rvts7>SP> +-------------------------------+----------------------+----------------------+</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>SP> A single 6-pin can deliver 75W, an 8-pin 150W, so in your case, the hard</span></p>
<p><span class=rvts7>SP> limits of what your card can pull is 75W from the PCI-E slow + 150W from</span></p>
<p><span class=rvts7>SP> the cable = 225 W. With a single 6-pin cable you'll only get ~150W max.</span></p>
<p><span class=rvts7>SP> That can be OK if your card does not pull more power (e.g. the above</span></p>
<p><span class=rvts7>SP> non-overclocked card would be just fine), but as your card is overclocked,</span></p>
<p><span class=rvts7>SP> I'm not sure it won't peak above 150W.</span></p>
<p><br></p>
<p><span class=rvts7>SP> You can try to get a molex -> PCI-E power cable converter.</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>SP> --</span></p>
<p><span class=rvts7>SP> Szilárd</span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>SP> On Mon, Jun 29, 2015 at 9:56 PM, Alex <</span><a class=rvts8 href="mailto:nedomacho@gmail.com">nedomacho@gmail.com</a><span class=rvts7>> wrote:</span></p>
<p><br></p>
<p><span class=rvts7>>> Hi all,</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> I have a bit of a gromacs-unrelated question here, but I think this is a</span></p>
<p><span class=rvts7>>> better place to ask it than, say, a gaming forum. The Nvidia GTX 960 card</span></p>
<p><span class=rvts7>>> we got here came with an 8-pin AUX connector on the card side, which</span></p>
<p><span class=rvts7>>> interfaces _two_ 6-pin connectors to the PSU. It is a factory superclocked</span></p>
<p><span class=rvts7>>> card. My 525W PSU can only populate _one_ of those 6-pin connectors. The</span></p>
<p><span class=rvts7>>> EVGA website states that I need at least 400W PSU, while I have 525.</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> At the same time, I have a dedicated high-power PCI-e slot, which on the</span></p>
<p><span class=rvts7>>> motherboard says "75W PCI-e". Do I need a different PSU to populate the AUX</span></p>
<p><span class=rvts7>>> power connector completely? Are these runs equivalent to drawing max power</span></p>
<p><span class=rvts7>>> during gaming?</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> Thanks!</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> Alex</span></p>
<p><span class=rvts7>>> --</span></p>
<p><span class=rvts7>>> Gromacs Users mailing list</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> * Please search the archive at</span></p>
<p><span class=rvts7>>> </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List">http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List</a><span class=rvts7> before</span></p>
<p><span class=rvts7>>> posting!</span></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> * Can't post? Read </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists">http://www.gromacs.org/Support/Mailing_Lists</a></p>
<p><span class=rvts7>>></span></p>
<p><span class=rvts7>>> * For (un)subscribe requests visit</span></p>
<p><span class=rvts7>>> </span><a class=rvts8 href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users</a><span class=rvts7> or</span></p>
<p><span class=rvts7>>> send a mail to </span><a class=rvts8 href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a><span class=rvts7>.</span></p>
<p><span class=rvts7>>></span></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>--</span></p>
<p><span class=rvts7>Gromacs Users mailing list</span></p>
<p><br></p>
<p><span class=rvts7>* Please search the archive at </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List">http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List</a><span class=rvts7> before posting!</span></p>
<p><br></p>
<p><span class=rvts7>* Can't post? Read </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists">http://www.gromacs.org/Support/Mailing_Lists</a></p>
<p><br></p>
<p><span class=rvts7>* For (un)subscribe requests visit</span></p>
<p><a class=rvts8 href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users</a><span class=rvts7> or send a mail to </span><a class=rvts8 href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a><span class=rvts7>.</span></p>
</td>
</tr>
</table>
</div>
<p><br></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>-- </span></p>
<p><span class=rvts7>Best regards,</span></p>
<p><span class=rvts7> Alex </span><a class=rvts8 href="mailto:nedomacho@gmail.com">mailto:nedomacho@gmail.com</a></p>
</td>
</tr>
</table>
</div>
<p><br></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts7>-- </span></p>
<p><span class=rvts7>Best regards,</span></p>
<p><span class=rvts7> Alex </span><a class=rvts8 href="mailto:nedomacho@gmail.com">mailto:nedomacho@gmail.com</a></p>
</td>
</tr>
</table>
</div>
<p><br></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts9>-- </span></p>
<p><span class=rvts9>Best regards,</span></p>
<p><span class=rvts9> Alex </span><a class=rvts8 href="mailto:nedomacho@gmail.com">mailto:nedomacho@gmail.com</a></p>
<p><br></p>
<p><span class=rvts9>--</span></p>
<p><span class=rvts9>Gromacs Users mailing list</span></p>
<p><br></p>
<p><span class=rvts9>* Please search the archive at </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List">http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List</a><span class=rvts9> before posting!</span></p>
<p><br></p>
<p><span class=rvts9>* Can't post? Read </span><a class=rvts8 href="http://www.gromacs.org/Support/Mailing_Lists">http://www.gromacs.org/Support/Mailing_Lists</a></p>
<p><br></p>
<p><span class=rvts9>* For (un)subscribe requests visit</span></p>
<p><a class=rvts8 href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users</a><span class=rvts9> or send a mail to </span><a class=rvts8 href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a><span class=rvts9>.</span></p>
</td>
</tr>
</table>
</div>
<p><br></p>
<p><br></p>
<p><br></p>
<p><br></p>
<p><span class=rvts10>-- </span></p>
<p><span class=rvts10>Best regards,</span></p>
<p><span class=rvts10> Alex </span><a class=rvts11 href="mailto:nedomacho@gmail.com">mailto:nedomacho@gmail.com</a></p>
</body></html>