weird GPU package insufficient memory error

I was running a eam/alloy simulation fine with the GPU package yesterday. After I ran it, I ran an update for Fedora 20, which included an update to the kernel and gpu driver. Since then, when I go to run my simulation I receive this message,

LAMMPS (30 Jan 2014)
Reading data file …
orthogonal box = (-0.426382 -0.426382 -0.426382) to (49.2064 49.2064 49.2064)
1 by 1 by 1 MPI processor grid
reading atoms …
6750 atoms
reading velocities …
6750 velocities

I was running a eam/alloy simulation fine with the GPU package yesterday.
After I ran it, I ran an update for Fedora 20, which included an update to
the kernel and gpu driver. Since then, when I go to run my simulation I
receive this message,

LAMMPS (30 Jan 2014)
Reading data file ...
  orthogonal box = (-0.426382 -0.426382 -0.426382) to (49.2064 49.2064
49.2064)
  1 by 1 by 1 MPI processor grid
  reading atoms ...
  6750 atoms
  reading velocities ...
  6750 velocities

--------------------------------------------------------------------------
- Using GPGPU acceleration for eam:
- with 1 proc(s) per device.
--------------------------------------------------------------------------
GPU 0: GeForce GTX 660 Ti, 1344 cores, 2.7/3 GB, 1.1 GHZ (Double Precision)
--------------------------------------------------------------------------

Initializing GPU and compiling on process 0...Done.
Initializing GPU 0 on core 0...Done.

Setting up run ...
ERROR on proc 0: Insufficient memory on accelerator
(../pair_eam_gpu.cpp:133)
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
Cuda driver error 4 in call at file 'geryon/nvd_timer.h' in line 98.
application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0
Cuda driver error 4 in call at file 'geryon/nvd_device.h' in line 124.
application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0

I noticed it says it is using 2.7-2.8 GB out of 3GB available on the GPU but
when I check this through the system it is only using a few hundred MB at
any given time. I have NOT changed anything in my input script from when it
was working fine til now. Would the updated driver cause an issue like this
and should I try to rollback to a previous version? I did do a clean build

difficult to say. you should carefully check your system log. i assume
you did do a reboot. depending on how you installed the nvidia drivers
(which ones do you use exactly?) your configuration might have gotten
out of whack. i have this on my F20 box:

[[email protected]... src]$ rpm -qa | grep -i nvid
xorg-x11-drv-nvidia-331.38-1.fc20.x86_64
akmod-nvidia-331.38-5.fc20.x86_64
xorg-x11-drv-nvidia-devel-331.38-1.fc20.x86_64
xorg-x11-drv-nvidia-libs-331.38-1.fc20.x86_64
kmod-nvidia-3.12.6-300.fc20.x86_64-331.38-5.fc20.x86_64

of the GPU library as well as a clean build of LAMMPS after the update and I
received the message the first time.

rebuild should not make a difference. since you didn't change the
toolkit or compiler.

i have not tested the GPU package in a while, so this is a good reason
to make a check.

more later,
      axel.