I've simulated inverting amplifier with ADA4817 in LTSpice and noticed that if gain is -1V/V (both resistors the same) and the resistors are 1Mohms then bandwidth is much lower than with lower resistances of the resistors. I'd expect that if the gain remains the same -1V/V then I should get at least 1MHz of bandwidth using either 1Mohm of few ohms resistors because this amplifier has high bandwidth and high input impedance but I didn't. I've noticed that the reason is that input impedance in this simulation is actually quite low, about hundreds of ohms at 1MHz.
I opened the model of ADA4817 and noticed that there is capacitor (Ccap 45 2 0.8p) connected directly between output and inverting input. Can I assume that low input impedance in simulation is caused by this capacitor? Can I assume that real amplifier will have much higher impedance (so also the bandwidth), caused mainly by input common mode and differential capacitances or there is really the capacitor put inside the silicon between input and output?