It seems to be a lot of noise there. I guess CH1 probe was set at 10x, while CH2 was 1x.
Put the CH2 V/div (Out) much lower to better see the noise, set the trigger to CH1 fallig edge.
Then change CH2 to the amp VCC.
The out signal is about 5mv/ °C, there shouldn't be more noise than few °C, which gets filtered in the software.
I will explain a bit how the filtering works:
The ADC bit precision (LSB) is ~0.8mV.
It makes 18 samples, discards the highest and the lowest values, and makes the average of the remaining 16.
Then the EMA filtering is done. It uses an acumulator, applying a % of the old value and a % of the new value.
This works very well at filtering noise, but can add a lot of delay if the signal changes quickly.
To overcome that, I limited the filter range for ADC differences of to 200 between new and old values.
Between 200 and 300, 1 to 100% of the new signal is applied, so over 300 the filter is completely resetted using the last value.
200ADC is 161mV, ~32°C, that's the max noise the filter will filter. If you get huge spikes, then the filter will be resetted.
I usually get 5-8°C noise.
Check my last picturesYour's seem pretty bad. I smell poor VCC filtering.