From owner-chemistry@ccl.net Fri Feb 8 13:03:00 2019 From: "Michael Morgan michaelmorgan937(0)gmail.com" To: CCL Subject: CCL:G: cpu and memory benchmark performance calculation Message-Id: <-53611-190208122529-7726-U212J891jQQ4rfvMa3KbGQ-,-server.ccl.net> X-Original-From: "Michael Morgan" Content-Language: en-us Content-Type: multipart/alternative; boundary="----=_NextPart_000_0023_01D4BFA0.FB4C6F30" Date: Fri, 8 Feb 2019 11:25:22 -0600 MIME-Version: 1.0 Sent to CCL by: "Michael Morgan" [michaelmorgan937]-[gmail.com] This is a multipart message in MIME format. ------=_NextPart_000_0023_01D4BFA0.FB4C6F30 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Dear all, If I want to test cpu or memory performance on a quantum calculation program (such as gaussian), usually what types of calculation are most suitable for respectively? For example, for parallel cpu performance, what types of calculation are mainly related to number of cpu cores but barely dependent on memory? On the other side, what types of calculation depend mainly on memory amount but not cpu? Thank you very much. Michael ------=_NextPart_000_0023_01D4BFA0.FB4C6F30 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Dear all,

 

If I want to = test cpu or memory performance on a quantum calculation program (such as = gaussian), usually what types of calculation are most suitable for = respectively?

 

For example, = for parallel cpu performance, what types of calculation are mainly = related to number of cpu cores but barely dependent on memory? On the = other side, what types of calculation depend mainly on memory amount but = not cpu?

 

Thank you very much.

 

Michael

------=_NextPart_000_0023_01D4BFA0.FB4C6F30-- From owner-chemistry@ccl.net Fri Feb 8 14:54:00 2019 From: "Close, David M. CLOSED=mail.etsu.edu" To: CCL Subject: CCL:G: [EXTERNAL] CCL:G: cpu and memory benchmark performance calculation Message-Id: <-53612-190208145208-15322-RAeQ5K5KPDT45dg1CQyHuA..server.ccl.net> X-Original-From: "Close, David M." Content-Language: en-US Content-Type: multipart/alternative; boundary="_000_DM6PR07MB4394139D0B8CCC0E8E9AE4A082690DM6PR07MB4394namp_" Date: Fri, 8 Feb 2019 19:51:59 +0000 MIME-Version: 1.0 Sent to CCL by: "Close, David M." [CLOSED#,#mail.etsu.edu] --_000_DM6PR07MB4394139D0B8CCC0E8E9AE4A082690DM6PR07MB4394namp_ Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Michael: There are several answers to your question. You do not say what types of= calculations you are interested in, and what computers you have access to.= So I would suggest some simple tests. Let's say that you have 8 nodes wi= th 5 Gb per node. So you request 8 nodes and 40 Gb of memory. So then dou= ble this by requesting 16 nodes and 80 Gb of memory. You can see in the li= terature that Gaussian doesn't scale very well, but you have to test this w= ith calculations you are interested in. So try this, and then write CCL ag= ain with specific details about what basis sets you use, and what hardware = you have available. Regards, Dave Close. > From: owner-chemistry+closed=3D=3Detsu.edu[a]ccl.net [mailto:owner-chemistry+= closed=3D=3Detsu.edu[a]ccl.net] On Behalf Of Michael Morgan michaelmorgan937(= 0)gmail.com Sent: Friday, February 08, 2019 12:25 PM To: Close, David M. Subject: [EXTERNAL] CCL:G: cpu and memory benchmark performance calculation Dear all, If I want to test cpu or memory performance on a quantum calculation progra= m (such as gaussian), usually what types of calculation are most suitable f= or respectively? For example, for parallel cpu performance, what types of calculation are ma= inly related to number of cpu cores but barely dependent on memory? On the = other side, what types of calculation depend mainly on memory amount but no= t cpu? Thank you very much. Michael The [EXTERNAL] tag in the subject line identifies emails that do NOT origin= ate from an ETSU person or service. Please exercise caution when handling e= mails from external sources. Any email that is unsolicited and requires you= to take immediate action, appears to be forged or is PHISHING for informat= ion can be verified by emailing the ITS Help Desk. --_000_DM6PR07MB4394139D0B8CCC0E8E9AE4A082690DM6PR07MB4394namp_ Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Michael:

  There are sever= al answers to your question.  You do not say what types of calculation= s you are interested in, and what computers you have access to.  So I = would suggest some simple tests.  Let’s say that you have 8 nodes with 5 Gb per node.  So you request 8 nodes and 40 Gb of= memory.  So then double this by requesting 16 nodes and 80 Gb of memo= ry.  You can see in the literature that Gaussian doesn’t scale v= ery well, but you have to test this with calculations you are interested in.  So try this, and then write CCL again with sp= ecific details about what basis sets you use, and what hardware you have av= ailable.

  Regards, Dave C= lose.

 

From: owner-chemistry+closed=3D=3Detsu.ed= u[a]ccl.net [mailto:owner-chemistry+closed=3D=3Detsu.edu[a]ccl.net] On Behalf Of Michael Morgan michaelmorgan937(0)gmail.com
Sent: Friday, February 08, 2019 12:25 PM
To: Close, David M. <CLOSED[a]mail.etsu.edu>
Subject: [EXTERNAL] CCL:G: cpu and memory benchmark performance calc= ulation

 

Dear all,

 

If I want to test cpu or memory performance on a qua= ntum calculation program (such as gaussian), usually what types of calculat= ion are most suitable for respectively?

 

For example, for parallel cpu performance, what type= s of calculation are mainly related to number of cpu cores but barely depen= dent on memory? On the other side, what types of calculation depend mainly = on memory amount but not cpu?

 

Thank you very much.

 

Michael

The [EXTERNAL] tag in the subject line ide= ntifies emails that do NOT originate from an ETSU person or service. Please= exercise caution when handling emails from external sources. Any email tha= t is unsolicited and requires you to take immediate action, appears to be forged or is PHISHING for informat= ion can be verified by emailing the ITS Help Desk.

--_000_DM6PR07MB4394139D0B8CCC0E8E9AE4A082690DM6PR07MB4394namp_--