Spectrum sensing (SS) is one of the most challenging and prominent operation in the emerging cognitive radio (CR) technology. It is viewed as one of the most intelligent method for enhancing spectrum usage efficiency. Therefore, versatile SS techniques have been proposed in literature. In this paper, we consider eigenvalue-based spectrum sensing (ESS) approaches, due to their strong competitive performances. An experimental assessment of a set of ESS algorithms is then provided. Our work is motivated by the high need to evaluate the performances of SS in near real-world scenarios, as a first step for industrial integration of this CR feature in existing and future wireless communication systems. Thus, we propose a real-time testbed that relies on the software-defined radio (SDR) platform: the national instrument (NI) 2954R universal software radio peripheral (USRP), as a physical medium for analog transmissions. The different processing modules of the developed testbed have been designed within a scalable and flexible architecture. It includes data acquisition, spectrum sensing, blind estimation of the noise power and performances reporting. Indeed, both of the ESS detection potential and sensing time have been explored. We show that, by conveniently selecting the ESS technique, the detection performances of a CR device can be improved (by 35%) while reducing sensing time (by 2) at relatively low SNR value (-12 dB).