linux - Saving GPU memory by bypassing GUI -


i have mac book pro 2 gb nvidia gpu. trying utilize gpu memory computations (python code). how saving may if bypassed gui interface , accessed machine through command line. want know if such thing save me amount of gpu memory?

the difference won't huge.

a gpu hosting console display typically have ~5-25 megabytes of memory reserved out of total memory size. on other hand, gpu hosting gui display (using nvidia gpu driver) might typically have ~50 megabytes or more reserved display use (this may vary based on size of display attached).

so can "estimate" of savings running nvidia-smi , looking @ difference between total , available memory gpu gui running. if is, example, 62mb, can "recover" around 40-50mb shutting off gui, example on linux switching runlevel 3.

i ran experiment on linux laptop quadro3000m happens have 2gb of memory. x display , nvidia gpu driver loaded, "used" memory 62mb out of 2047mb (reported nvidia-smi).

when switch runlevel 3 (x not started) memory usage dropped 4mb. means ~50mb additional available cuda.

a side benefit of shutting off gui might elimination of the display watchdog.


Comments

Popular posts from this blog

get url and add instance to a model with prefilled foreign key :django admin -

css - Make div keyboard-scrollable in jQuery Mobile? -

ruby on rails - Seeing duplicate requests handled with Unicorn -