From: Herald Delis (bioinfo2021great_at_gmail.com)
Date: Sun Sep 12 2021 - 00:22:07 CDT
Hello Marcelo,
Thanks, I will try different combinations for NAMD2 and ORCA.
My CPU (on lscpu command except flags)
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 44
On-line CPU(s) list: 0-43
Thread(s) per core: 1
Core(s) per socket: 22
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6152 CPU @ 2.10GHz
Stepping: 4
CPU MHz: 1000.012
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4200.00
Virtualization: VT-x
L1d cache: 32K
L1i cache: 32K
L2 cache: 1024K
L3 cache: 30976K
NUMA node0 CPU(s): 0-21
NUMA node1 CPU(s): 22-43
I have couple more question,
1.Does "extrabond" affect the speed? or the results? I have 2 Iron atoms in
the active site.
2. In QMMM-Min/QMMM-Ann/QMMM-Equi can we use SQM/HF and then tight
calculation in production QMMM?
Thanks and regards,
Herald
On Sat, Sep 11, 2021 at 2:28 PM Marcelo C. R. Melo <melomcr_at_gmail.com>
wrote:
> Hi Herald,
>
> It looks like you are requesting a relatively expensive calculation from a
> ~110 atom system, and this is taking ~35minutes to calculate each SCF (52
> hours is the time remaining to run ALL steps in your simulation). I don't
> think this is terrible, and improving this time will depend on how much
> parallelism you can achieve and what your hardware may look like.
>
> You can test the ratio NAMD/ORCA to find the optimal spot, maybe having as
> few as 2 or 3 cores for NAMD, and 37 cores for ORCA is the best balance for
> you.
>
> Do you have 40 cores or 40 threads on your system (as in 20 dual-thread
> cores)? NAMD and ORCA will likely use all the CPU, and using multithreading
> will decrease the performance.
>
> Best,
> Marcelo
>
> On Thu, 9 Sept 2021 at 21:00, Herald Delis <bioinfo2021great_at_gmail.com>
> wrote:
>
>> Hello Marcelo,
>>
>> Thank you for reply,
>>
>> Q: 1. Is 75.000 the total number of atoms in your entire system, or in
>> your QM regions? How large is the QM region itself?
>> If you have more than a few hundred atoms in your QM region, I would
>> expect ORCA to take a while to calculate it.
>>
>> A1: 75,000 is the total atoms, 113 atoms in QM region.
>>
>>
>> Q: 2. In your NAMD execution line, you could ask for fewer nodes, like 5
>> or 10, and leave the rest for ORCA.
>> For example:
>> NAMD2 -> namd2 +p10 Min.conf
>> ORCA -> qmConfigLine "% PAL NPROCS 30 END" (from Min.conf)
>>
>> A2: This works and speeds up. But it is still showing 52 hours remaining.
>>
>> QMENERGY: 16 1.0000 -3517431.7765 -3517172.7403
>>
>> (From Step 16)
>> Info: Writing QM charge output at step 16
>> Info: Writing QM position output at step 16
>> PRESSURE: 16 -3304.93 308.224 -262.003 312.544 -2742.12 -82.5491 -261.689
>> -85.8746 -2603.54
>> GPRESSURE: 16 -7758.39 303.504 -166.053 239.227 -6827.13 -132.006
>> -216.514 120.539 -6983.22
>> TIMING: 16 CPU: 4.09771, 0.11225/step Wall: 40290.3, 2258.54/step,
>> 52.6993 hours remaining, 532.187500 MB of memory in use.
>> ENERGY: 16 8544.0808 5371.1756 4389.7518
>> 47.8456 -255440.1271 20369.6414 1.7922 -3517172.7403
>> 0.0000 -3733888.5801 0.0000 -3733888.5801
>> -3733888.5801 0.0000 -2883.5306 -7189.5819
>> 806859.3967 -2883.5306 -7189.5819
>>
>> WRITING EXTENDED SYSTEM TO RESTART FILE AT STEP 16
>> LINE MINIMIZER BRACKET: DX 0.000456178 0.000121109 DU -1281.69 90.8142
>> DUDX -5.59903e+06 2423.13 1.49124e+06
>> MINIMIZER RESTARTING CONJUGATE GRADIENT ALGORITHM
>> LINE MINIMIZER REDUCING GRADIENT FROM 1.29277e+07 TO 12927.7
>> WRITING COORDINATES TO DCD FILE QMMM-MinX.dcd AT STEP 16
>> WRITING COORDINATES TO RESTART FILE AT STEP 16
>> FINISHED WRITING RESTART COORDINATES
>> WRITING VELOCITIES TO RESTART FILE AT STEP 16
>> FINISHED WRITING RESTART VELOCITIES
>>
>> Q:3. If you only have one QM region in your system, you only need to ask
>> for 1 QM simulations per node. This won't change your calculation.
>>
>> A: This works too. But please check the above answer.
>>
>> 4. What kind of QM calculation are you running? semi empirical? HF?
>>
>> A: QM Calculation:
>>
>> ! uks BP86 def2-TZVP def2/J EnGrad TightSCF
>> %maxcore 3000
>> % PAL NPROCS 30 END
>> % basis
>> newgto C "DEF2-SVP" end
>> newgto H "DEF2-SVP" end
>> end
>> % scf
>> MaxIter 500
>> end
>> %pointcharges "qmmm_exec/QMMM-Min/0/qmmm_0.input.pntchrg"
>>
>>
>> %coords
>> CTyp xyz
>> Charge 3.000000
>> Mult 1.000000
>> Units Angs
>> coords
>>
>>
>> Additionationlly, I am using extrabonds for keeping my metal ion in
>> place. Will it affect the speed or the calculation in any way?
>>
>> Thanks and regards,
>> Herald
>>
>> On Fri, Sep 10, 2021 at 2:09 AM Marcelo C. R. Melo <melomcr_at_gmail.com>
>> wrote:
>>
>>> Hi Herald,
>>>
>>> I may have a couple of suggestions and questions for you.
>>>
>>> 1. Is 75.000 the total number of atoms in your entire system, or in your
>>> QM regions? How large is the QM region itself?
>>> If you have more than a few hundred atoms in your QM region, I would
>>> expect ORCA to take a while to calculate it.
>>>
>>> 2. In your NAMD execution line, you could ask for fewer nodes, like 5 or
>>> 10, and leave the rest for ORCA.
>>> For example:
>>> NAMD2 -> namd2 +p10 Min.conf
>>> ORCA -> qmConfigLine "% PAL NPROCS 30 END" (from Min.conf)
>>>
>>> 3. If you only have one QM region in your system, you only need to ask
>>> for 1 QM simulations per node. This won't change your calculation.
>>>
>>> 4. What kind of QM calculation are you running? semi empirical? HF?
>>>
>>> Best,
>>> Marcelo
>>>
>>> On Thu, 9 Sept 2021 at 06:53, Herald Delis <bioinfo2021great_at_gmail.com>
>>> wrote:
>>>
>>>> [image: image.gif]Dear All,
>>>>
>>>> I am trying to run QMMM-MD using orca-namd2.
>>>> But it happens to be too slow, I wanted to know if I am doing something
>>>> wrong or is there a better way.
>>>> I have a CPU node with 40 cores (and 4 unused GPU), my calculations of
>>>> 75,000 atoms QMMM-Minimization takes 4 days.
>>>>
>>>> NAMD2 -> namd2 +p22 Min.conf
>>>> ORCA -> qmConfigLine "% PAL NPROCS 20 END" (from Min.conf)
>>>>
>>>> # Number of simultaneous QM simulations per node
>>>> QMSimsPerNode 20
>>>>
>>>> Thanks and regards,
>>>> Herald
>>>>
>>>>
>>>>
>>>>
This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST