TeraChem User Forum

Please login or register.

Login with username, password and session length
Advanced search  

News:

Kepler Support out in v1.50K!

Pages: [1] 2 3 ... 10
 1 
 on: May 31, 2017, 09:28:59 PM 
Started by Bokhimi - Last post by Bokhimi
I am configuring a new workstation.

In the present time, I am running TeraChem in a Workstation with 4 K80.

Is it possible to run TeraChem in a node or Workstation with 8 K80? If this is the case, which is your experience with these systems.

I would like to know if you are planning to run soon TeraChem in the new Nvidia Volta architecture.

This information is important for configuring the workstation.


 2 
 on: November 02, 2016, 11:51:03 AM 
Started by Bokhimi - Last post by ufimtsev
Comparing gradients may be tricky sometimes if you want many digits after the decimal to be the same. First thing -- the SCF convergence threshold should be as tight as possible, since the energy is quadratic in delta C and gradients are linear. In DFT, you really want to run the calculations on exactly same grids. This is often impossible, because different packages use different grids, and it's recommended to use ultra fine grids to minimize numerical noise. I heard that some programs (not TeraChem, of course) don't compute the grid derivatives contribution to the gradient, i.e. their gradients are just incorrect, and you don't want to use them in your tests.
-Ivan

 3 
 on: November 01, 2016, 02:55:52 AM 
Started by Bokhimi - Last post by Bokhimi
Thanks Ivan

In the comparison, for TeraChem I used the basis 6-31g and not 6-31g*.

Doing this change in the basis the comparison gave the same gradient values.

 4 
 on: October 29, 2016, 11:53:17 AM 
Started by Bokhimi - Last post by ufimtsev
Please check your start file, something may be wrong with the method or basis set. Here is a result for

coordinates   water.xyz
charge        0
method        rb3lyp
run           gradient
basis         6-31g*
end


FINAL ENERGY: -458.5384933282 a.u.
CENTER OF MASS: {0.000077, 0.000006, 0.000004} ANGS
DIPOLE MOMENT: {-0.003041, 0.000195, -0.002096} (|D| = 0.003698) DEBYE

Running Mulliken population analysis...

GPU Memory: 9411 Mb
Writing out molden info

Gradient units are Hartree/Bohr
---------------------------------------------------
        dE/dX            dE/dY            dE/dZ
    0.0023710677     0.0051839449     0.0054077283
    0.0074956937    -0.0105356040     0.0035103656
    0.0077306173     0.0004218266    -0.0108471130
   -0.0064421340     0.0004360239     0.0042808221
    0.0083020774     0.0071885933     0.0077152096
   -0.0043217808     0.0045016851    -0.0045524299
   -0.0077132761    -0.0004261841     0.0108439615
    0.0043001102    -0.0045131765     0.0045531296
   -0.0083017346    -0.0071968118    -0.0077279848
    0.0064403922    -0.0004466109    -0.0042795274
   -0.0074943199     0.0105399984    -0.0035002380
   -0.0023718638    -0.0051780596    -0.0054114713
   -0.0056043331     0.0043749254    -0.0029214078
   -0.0032229892    -0.0034493254    -0.0060460568
    0.0052150610    -0.0016924325    -0.0053162523
   -0.0051983044     0.0017065812     0.0053155498
    0.0032208366     0.0034650309     0.0060489377
    0.0055948798    -0.0043804047     0.0029267768
---------------------------------------------------

 5 
 on: October 24, 2016, 09:35:12 PM 
Started by Bokhimi - Last post by Bokhimi
Because I have troubles using the gradient values given by TeraChem for my system. I compared the gradient values obtained using TeraChem with the corresponding ones obtained with Gaussian (Results are attached). For the comparison I obtained these values for a water cluster made of 6 water molecules.

Why I obtained so different results?

Bokhimi

 6 
 on: September 08, 2016, 01:01:29 PM 
Started by ufimtsev - Last post by ufimtsev
TeraChem 1.91 supports floating licenses and can run on any node that has Internet access (direct or via proxy server) without need for a license file bound to the node. This means one can run it on publicly available GPU clouds like Amazon EC2 in evaluation mode, or production mode if the floating license file is placed in TeraChem root directory. Login to Amazon EC2 console -> IMAGES -> AMIs, then select [Public Images] in the search bar dropdown and search for 'TeraChem' if you would like to try it out. Make sure the owner account is 380438227580.
PetaChem Team

 7 
 on: August 16, 2016, 09:05:39 PM 
Started by markus - Last post by admin
When you do an unrestricted CIS calculation, you will AUTOMATICALLY get different spin multiplicities in the excited states.  The output provides the expectation value of S^2 for each state in the summary (under the column labeled <S^2>).  In general, you will see that these show that the excited states are not spin-pure, i.e. they are not eigenfunctions of S^2.  This is expected with UHF and UCIS (but also makes these methods somewhat problematic and restricted methods are often preferred).  The bottom line here is that cismult is not a valid keyword when doing unrestricted CIS calculations (because the multiplicity of both ground and excited states is not well-defined since neither ground nor excited states will usually be S^2 eigenfunctions in an unrestricted calculation).
-Todd

 8 
 on: August 16, 2016, 07:36:12 PM 
Started by ufimtsev - Last post by admin
Eventually, but that will need to wait until the next release.  Please note that there is no released CUDA toolkit which supports Pascal at present.
-PetaChem Team

 9 
 on: August 16, 2016, 08:51:11 AM 
Started by markus - Last post by markus
I'm trying to calculate triplet excitations using TDDFT. Setting cismult 3 (together with a singlet ground state and an unrestricted calculation) gives the
error message:  Incorrect multiplicity

Is there anything extra I need to specify?

Thanks!
Markus

 10 
 on: August 07, 2016, 07:26:09 AM 
Started by ufimtsev - Last post by f4ding
Since the Pascal cards look to be so much faster than the Maxwell class card, will there be any support for the Pascal cards with v1.9?

Pages: [1] 2 3 ... 10