feasibility of execution of IP algos in GPU

Hi sir!
this is vaishali. we are interested in purchasing the nvidia tegra seriesTX1 evaluation board.The application we are involved in is detailly explained below with requirements.
The front end of our application is a mipi based CCD/CMOS sensor snapping out 60fps and capable of upto 4k resolution. The input stream rate is about 2.5GBps running into TK1 processor and followed by the display part through HDMI interface.
we would like to understand whether there is a feasibility of dropping in ImageProcessing(IP) algorithms in the GPU processor integrated in tk1 kit and also executing those within 7-10 ms, analogous to executing the IP algorithms in matlab forum supported by i7 processor. On the whole, for one frame to be acquired, processed and displayed the time limit should be less than 15 ms.
please let us let us know asap. your early response would be appreciated. Thank you in advance!

Hi vaishali,

Are you asking about TK1/TX1?

The input stream rate is about 2.5GBps running into TK1 processor and followed by the display part through HDMI interface.
Why is it 2.5GBps?

we would like to understand whether there is a feasibility of dropping in ImageProcessing(IP) algorithms in the GPU processor integrated in tk1 kit and also executing those within 7-10 ms, analogous to executing the IP algorithms in matlab forum supported by i7 processor.
I cannot tell the time because it depends on your ImageProcessing. You can port and try your IP on DevKit and check it.

Hey ! thanks for that.
will we be supported with the technical assistance from your side team in importing the IP algos into the development kit?

Hi vaisha,

We will be lovely to support your development issue on Jetson platform/SDK.

Thanks