Extreme long settling time (6 ms) of the DAC on STM32F4

I encounter a problem with the DAC on both STM32F4/STM32F429 Discovery boards and hope someone can give me a hint.

The DAC (using DMA with output buffer enabled) generates a FSK controlling signal sequence with only two steps, which looks like as follows

As can be seen on the screenshot, the settling time of the voltage step is about 6 ms! Since the signal sequence controls a very sensitive VCO, this long settling time results in a corresponding segment of output freeuency that is unusable for my application.

I am wondering, is this a problem of settling time?
If yes, is there a way to alleviate it?

Updated:

• The probe connects directly to the DAC pin, which is the only pin used in the measurement.
• I have calibrated my scope's probe.
• The STM32F4 discovery board is powered via the USB port.

With this setup, I tried to capture the rising edge of the step from 0 to 4000 using different time scales.
The results are as follows:

Compared to the above screenshot, the overshooting amplitude seems to be proportional to the step-voltage.

Due to my low reputation, I cannot post more than 2 links, but the zoom-in view shows a similar rising edge as @Arsenal's measurement. I think that is the real settling time.

I hope this update could lead to some new suggestions.