Reproducing Kernel Hilbert Spaces (RKHS) and frames provide a more general setting for classical, i.e. \(L^2 (I) \) sampling theorems, [\textit{N. Aronszajn}, Trans. Am. Math. Soc. 68, 337--404 (1950; Zbl 0037.20701); \textit{S. Saitoh} and \textit{Y. Sawano}, Theory of reproducing kernels and applications. Singapore: Springer (2016; Zbl 1358.46004)]. Extensions of RKHS and frames to the setting of Banach spaces are considered by several authors (\textit{D. T. Stoeva} [Int. J. Pure Appl. Math. 52, No. 1, 1--14 (2009; Zbl 1173.42322)]; \textit{H. Zhang} et al. [J. Mach. Learn. Res. 10, 2741--2775 (2009; Zbl 1235.68217)]; \textit{H. Zhang} and \textit{J. Zhang} [Appl. Comput. Harmon. Anal. 31, No. 1, 1--25 (2011; Zbl 1221.42067)]). A direct sampling theorem in such context is presented in [\textit{A. G. García} and \textit{A. Portal}, Mediterr. J. Math. 10, No. 3, 1401--1417 (2013; Zbl 1279.42035)], see also [Zhang und Zhang, loc. cit.]. This paper is devoted to a converse of Kramer-type sampling theorem over semi-inner product reproducing kernel Banach spaces. The authors prove an analogue to Theorem 1 of [\textit{A. G. García} and \textit{F. H. Szafraniec}, Sampl. Theory Signal Image Process. 1, No. 1, 53--61 (2002; Zbl 1067.94008)], where RKHS are treated instead. An illustrative example in the context of Lebesgue spaces on an interval is given in the paper.