Hi dear all,
I just carrying out a computational run of a single point energy at MP2/6-31g++(2p,2d)//B3LYP/6-31g++(d,p), but when I tried to do the refinement at MP2/6-31g++(2p,2d) the next error appear in my log file:
Is it a problem with the integration of method or memory of my CPU? Thanks a lot. Julián.
"Disk-based method using OVN memory for 31 occupieds at a time.
Estimated scratch disk usage= 2478105962 words.
Actual scratch disk usage= 2477844842 words.
JobTyp=2 Pass 1: I= 1 to 31 NPSUse= 5 ParTrn=T ParDer=T DoDerP=F.
Erroneous write. Write -1 instead of 1470464.
fd = 4
orig len = 1875968 left = 1470464
g_write"
This error usually happens when your scratch space is not sufficient to carry over your calculations (pretty common using MP2). In most cases adding your available scratch space in a maxdisk keyword is enough to overcome this problem, something like maxdisk=80gb, but remember, the error might persist if your scratch is not large enough to run your desired calculation.
More info: http://gaussian.com/maxdisk/
You need other disk or more disk space. Maybe deleting or moving other scractch file you have the space needed to run your job.
Hey Julian! I was solving a similar problem by adding more memory in my com-file, i.e. %mem=50GB and so on (depending on how much you need). Then it is good to empty the space where you store those files produced by Gaussian, or if you run on a super-computer to use the memory on the node itself rather than in your submission directory. Finally, one more solution is to use less processors but reserve the whole node. If you node has 20 cores you shall use, let's say 4 out of them. It worked for me as well.
- Badges
- Science topic