Honestly I think we can say at least a 1% death with quite a bit of confidence at this stage.
Korea, with large scale testing, contact tracing and new cases very low and dropping without lockdown, is today at 1.95% (John Hopkins expert morons were saying it proved 0.6% when it was 0.6%). We can assume at most 100% undersampling there imo, or it would be spreading more/cases would have untraceable contacts, and the death rate has quite a bit to climb yet.
Bergamo has these deaths for March, 0.5% population-level death rate in March alone (assuminng 100% of the population infected and all deaths played out):
You can add another 50% onto that by the time it's done, deaths only peaked at the end of March which is the halfway point for total deaths. Halfway is 100% but I'm being conservative and only adding 50%. And if you assume these tests are correct, 38% infected, you end up with a 1.97% population death rate, conservatively.