Hi Fangjia,
I believe CUDA 8.0,although works fine up to Volta GPUs, cannot
produce binaries for Turing GPUs (e.g., RTX 2080). You should upgrade to
CUDA 10.0/10.1 if you want to run on RTX 2080.
Taisung
-----Original Message-----
From: bjyx20090941 <bjyx20090941.163.com>
Sent: Wednesday, April 3, 2019 10:24 AM
To: amber.ambermd.org
Subject: [AMBER] cudaMemcpyToSymbol: SetSim copy to cSim failed invalid
device symbol ( GeForce RTX 2080)
Dear everyone:
First, I have successfully install the pmemd.cuda with cuda-8.0 and
GeForce GTX 1080, and I run large systems of protein in MD simulation with
pmemd.cuda module. Once the same question had been happened, but I solved it
with ./update_amber --update ( pudate amber 15.0)
Today, I am install the pmemd.cuda with cuda-8.0 and GeForce RTX 2080,
but I can not solve this question:cudaMemcpyToSymbol: SetSim copy to cSim
failed invalid device symbol. So my amber cannot run with pmend.cuda, but
running with pmend and pmemd.MPI.
However, I have checked deviceQuery, such as :
a. run the deviceQuery samlpe: ./deviceQuery results = PASS.
b. nvidia-smi and nvcc: the gpu imformation is showing.
c. ./update_amber --update ( pudate amber 15.0) .
So what is the problem in my installation of pmemd.cuda.
Besides, I found that after install ./configuration -cuda gnu , make
install. there is not any error imformation shown, but make test , have
serveral error. What's wrong with my installation of pmemd.cuda
Yout sincerely,
Fangjia Fu
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 03 2019 - 08:00:02 PDT
