I was introduced to NVIDIA's Tesla GPU Accelerator modules/cards recently. After some basic experimentation and reading, I'm just not convinced they can/do make much of a difference in displaying multiple MP camera displays like a VMS or NVR typically is tasked to do.
Unlike many GPUs, these accelerators cannot be used for mining, so supply is mostly exempt from crazy GPU demand. This video introduces the Tesla M40:
********* *************:
**'* ********** ** ***** ********** ** real-world ************ ** *-***, ***** **** saving ********* ** ** **%
...*** ** ***, ** *****'* **** like ****** *** ** * *** server **** ****** ********** *** ******* does **** ** ******* *** ***********. Lags *** ***** ***, ********* **** views *** ***** ****, ** ****** trying ** ******* **** **** **-** thumbnails, ***.
** *'* *** **** ************* ***** GPU ************ ** ******** ******** ** using *** *****. **** ***** ****** explain *** ******, *** ** ***** it **** *** *** ************ *** VMSes ** ******* *********** ** * reach:
** *** *** ***** *** ************? Do **** ****** ******* ***/******* ******** performance **** **** *****?
*** '***** **** **** *****' ** based ** * *** ** ********** that **** **** ****** ** ** with * ***.
** ******* ** ****** *** ** looking ** *** * *** *** able ** **** ********* ** ********* hardware *********. * *** ****** **** at **** ;-) *** ** * GPU *** *** ** ***** *******.
***...**** *** ** * **** *** piece ** **** *** *** *** give *** * **** ***** ** capability *** *** *** *** ** using. *** *** ******** ***** ***** are *** ***** **** **** *** past ****.
* **** ***** ******* ** **** in **** *** ****** ***** ** located ****:***** *******
* ******* *** **** **** * 'recording' ******** *** * '*******' ********. Both ** ***** ********* ****** ********* on *** ***** ********** ********* ** the ******. *** ***** ********* ****** is *** **** *** ****** *** can **** *** *** **** ** needs.
********* **** * **** *** **** their ********* ****** *** * ** card ********* *** **** *******. * did *** ******** ******* *** **** was **** ** ******* * *** on *** ********* ******.
** *** ***** ****, **** ****** VMSs **** *** *** ***** *** the '*******' ********* ** ******* ******* of *******, *** ** **** **** improve **** *** ** **** ** a ***** ******.
* ********* *** ****** **** **** VMSs **** ****** ** *** *** memory. ** ***** ***** **** *** building ** *** ****** *** ******* them **** *** **** *****.
** *******, * *** **** *** depend ** **** **********, *** ******* good ***** ***********. * ********* ** NVDEC *********** *** ** ***** ****:***** ***********
**** **** *** *********** ** ** individual ***** ***** **** *** ** up **** **** ** *** ** up *** **** *****. *** * VMS, *** ******* *** ***** *** very ******* *********** ************ *** *** cost ************* *** *** **** *** bigger ****.
* ***** ** **** *** ** that ***** **** *** **** ********* of ******** ***** ********* ** ******** cards (********* *** ******* *** **** examples). **** ****** **** *** ***** that **** ******** ***** ***** ** (ie *** ******** ******* *** ***), will ***** ***********.
* **** *** ** ********** *** many *****/***** ****** ***** ** * card ** ****:*****-***** *** ****
** *** ** ****, ***** **** will *** *** ** *** ** card ******** *** **********, *** ***** is ***** *** **** ****.
** * ******* ******, *** *** of ******* **** *** *** *** what ** ***** **** ** * NV **** ******* *** ******* **** PM-counters ********* ** *******.
**** ** **** *** ***** *****, so *** *** ** *** *** 'nvidia-smi' ******* ** * ***** **** to ****** *** **** ** *** GPUs ** *** ******.
*** **** ****** ***** **** ******* have '**** ** **********' ** ** long **** *** ****** ***** ***. They *** **** ****** ** *** AI/ML *****.
*** *** **** **** ** ** is ******** **** *** *** ** AI *********.