This paper shows experimental results on columnstore performance in some settings. This paper focuses on utilization efficiency of CPU. The authors analyze effects of some techniques related to CPU such as Prediction*, SIMD, Prefetch, Turbo boost and Hyper Threading. This paper shows the effect when performing multi-thread processing is quite smaller than one when performing single-thread processing due to memory bandwidth limit and cpu data cache miss. The author of this paper presented the detail in VLDB2020 “Micro-architectures & Flash Storage” session.
* Prediction: exploit bitwise operation and eliminate branchesto reduce cpu stalling time from branch mispredictions
 U. Sirin, A. Ailamaki, Microarchitectural Analysis of OLAP: Limitations and Opportunities, PVLDB2020.