Getting complete results when processing aggregate queries on public SPARQL endpoints is a challenge, mainly due to the application of quotas. Although Web preemption allows processing aggregate queries online, on preemptable SPARQL servers, data transfer is still very large when processing count-distinct aggregate queries. In this paper, it is shown that countdistinct aggregate queries can be approximated with low data transfer by extending the partial aggregation operator with Hyper-LogLog++ sketches. Experimental results demonstrate that the proposed approach outperforms existing approaches by orders of magnitude in terms of the amount of data transferred.