After doing some rough calculations, you’ll need to average 24 of these chips to equal the accelerometer noise performance of a single ISM330DHCX.
Here are calculations on which that is based:
From ADXL345 Datasheet:
Typical Sensitivity (@2 g, 100Hz): 10-bit resolution 256 LSB/g
Accel Noise: 0.75 LSB rms
So each bit is equal to 1/256 G = 3.9mG. And the X/Y axis noise level is .75 of a bit @100Hz, so the noise is 3.9*.75mG = 2.95mG = 2950uG. Dividing that by square root of 100Hz yields 295 uG/sqrt(hz). That’s on par with the MPU-9250, which had specs of 300uG/sqrt(hz).
The ISM330DHCX datasheet indicates it’s typical “Acceleration noise density in high-performance mode” is 60ug/sqrt(hz)
There’s another thing to consider:
ADXL345 is a 10-bit sensor (since the noise level spec’d above is .75/bit it’s resolution is less than the noise), ISM330DHCX is 16-bit. 6 bits of resolution is really a huge difference, although of course some of that’s going to be noise.
I encourage you to review the accelerometer noise density specs carefully if you plan to estimate displacement with them. Getting familiar with the datasheets is a very valuable skill.