Do you have an example of an input file where this is happening? How was your version of gamess compiled?

Jared, it is good to hear from you

Hope things are your fine on your side and thanks in advance for your help

Please find attached the input with a cc-pVDZ as basis set which requires 10936 ddimem. For this calculation using 32 cores (16 compute + 16 data servers) while setting the memory as 100 MW replicated and 11000 MW ddi, the actual memory consumption at the end was 270.4 GB, which should be 8*(100+ 11000/16)/1024= 6.15 GB/compute core or 98.4 GB in total! the output of GAMESS doesnt indicate that this extra memory was even used, here:

THIS CALCULATION IS RUNNING WITH MWORDS= 100, MEMDDI= 11000, AND P= 16

MEMORY USAGE PER CPU IS 8*(MWORDS + MEMDDI/P)/1024 = 6.2 GBYTES.

MINIMAL REQUIREMENT FOR THIS RUN IS MWORDS= 22, MEMDDI= 10936.

FOR P= 16, THE MEMORY USAGE PER CPU(CORE) WOULD BE 5.5 GBYTES.

DDI: Creating Array [0] - 1123600 x 9730 = 10932628000 words.

DIRECT 4-INDEX TRANSFORMATION

SCHWARZ INEQUALITY TEST SKIPPED 5760544914 INTEGRAL BLOCKS

RESULTS OF MOLLER-PLESSET 2ND ORDER CORRECTION ARE

E(SCF)= -3009.8766193423

E(2)= -7.4748675392

E(MP2)= -3017.3514868815

SPIN-COMPONENT-SCALED MP2 RESULTS ARE

E(2S)= -5.4739445056

E(2T)= -2.0009230336

E(2ST)= -7.2357077513 = 6/5 * E(2S) + 1/3 * E(2T)

SCS-MP2= -3017.1123270936

..... DONE WITH MP2 ENERGY .....

CPU 0: STEP CPU TIME= 1713.95 TOTAL CPU TIME= 2149.6 ( 35.8 MIN)

TOTAL WALL CLOCK TIME= 2183.8 SECONDS, CPU UTILIZATION IS 98.44%

18897239 WORDS OF DYNAMIC MEMORY USED

EXECUTION OF GAMESS TERMINATED NORMALLY Sun Jan 1 03:25:37 2017

DDI: 5497974600 bytes (5243.3 MB / 655 MWords) used by master data server.

***********

The problem problem gets worse when I increase the basis set, for instance cc-pVTZ with gradient which requires about 1.85 TB memory on 240 cores, the consumption reaches 4.2 TB! I wasn't able to finish that one:

THIS CALCULATION IS RUNNING WITH MWORDS= 200, MEMDDI= 220000, AND P= 240

MEMORY USAGE PER CPU IS 8*(MWORDS + MEMDDI/P)/1024 = 8.7 GBYTES.

MINIMAL REQUIREMENT FOR THIS RUN IS MWORDS= 153, MEMDDI= 214240.

FOR P= 240, THE MEMORY USAGE PER CPU(CORE) WOULD BE 8.2 GBYTES.

DDI: Creating Array [0] - 6812100 x 20100 = 136923210000 words.

DDI: Creating Array [1] - 2094 x 4020000 = 8417880000 words.

DDI: Creating Array [2] - 3407355 x 20100 = 68487835500 words.

DDI: Creating Array [3] - 20100 x 20100 = 404010000 words.

***********

The GAMESS version that I am using is the 12052014 that is compiled with GCC 4.4.5 and OpenMPI 1.4.3, this is the version that is compiled at the HPCC.

Your help is highly regarded Jared, thanks a lot

Right, if this was run on HPCC at MSU then their queue has problems with GAMESS and memory as it seems to think it needs more than it actually does. They know about it but it's not an easy fix so they haven't done anything for it, which is unfortunate and annoying.

If you're doing a geometry optimization, something like 6-31G* would probably be good enough at the MP2 level of theory and will use slightly less memory.

Thanks a lot Jared for your response, this answers my concern

You are right, I will just use a smaller basis set for the geometry optimization as you suggested in this case