Goal: intall CUDA for nvidia qtx 750 ti
in scientific linux 6.5 (RHEL 6

Done: downloaded cuda_6.0.26_rc_linux64.run
#sh ./cuda_6.0.26_rc_linux64.run (in folder where downloaded)

Challenge: *** gzip extration failed - ensure there is enough space in temp ***


  1. temp of which disk ?
  2. how much does it need ?
  3. how to configure more space and to which program?
    a) to gzip ?
    b) to nvidia program?
  4. how to extract in empty 1TB partition?
    a) is ext4 file format ok ?
  5. how to extract in .3TB partition?
    b) is ntfs ok (ntfs-3g drivers installed)
  6. how much space does it need in system/boot drive?
    c) .32TB - space required by sl 6.5
  7. why did it try other drives or at least told how much
    it needed and how much did it think was available?

thanks in advance for any advice


  • all hard disks seem to have hundreds of gigabytes free?
  • is the install 32-bit program with max 2TB problem?
  • assumed .run download is better than .rpm, is it?
  • created tmp to all hard disks (they have temp)

temp is /tmp

you can set temp directory with (see cuda_6.0.26_rc_linux64.run --help)

-tmpdir : Use as temporary directory - useful when /tmp is noexec

and, if you wanth, can set installation(final) directory with*


  • but need more steps:

need extract first the contents of cuda_6.0.26_rc_linux64.run with

sh cuda_6.0.26_rc_linux64.run -extract=<EXTRACTION_PATH>

and then execute the .run extracted

./cuda-linux64-rel-6.0.26-17754484.run -prefix=<INSTALLATION_PATH> -tmpdir
./cuda-samples-linux-6.0.26-17754484.run -cudaprefix=<INSTALLATION_PATH> -prefix=<INSTALLATION_PATH>/samples -tmpdir

the all process need +/- 3Gb of space


giving the following parameter was the solution:

the installation program could have tried harder - there was space enough in hard disks -
maybe this was a 64-bit challenge for 32-bit program to handle > 2TB disk!