In this paper we investigate the behaviour of Turing machines seen as dynamical systems. In this context, a Turing machine does not start from a specified initial state and tape, but from any configuration. In this context, we call a configuration immortal if the machine runs forever starting from it. The seminal article of Hooper [8] proves that we cannot decide whether a Turing machine has an immortal configuration (is immortal). However, the result does not say anything about what the immortal configurations look like. In fact, in the construction by Hooper, if the Turing machine has an immortal configuration, then it has one where the tape is almost completely empty.A few results give some structure on the set of immortal configurations: Kurka [13] asked whether there always exists a (temporally) periodic configuration, and was refuted by Blondel et al. [1]. Delvenne and Blondel [6] proved that the set of immortal configurations, if nonempty, always contains quasi-periodic configurations.In this article, we go further, and give some news results on the set of immortal configurations. We prove in particular in Theorem 2 that there exists a Turing machine for which the set of immortal configurations is nonempty and contains no computable configurations.The main ingredient is a new proof of Hooper's theorem by Kari and Ollinger [11] combined with an encoding into subshifts of an effective set with no recursive points. We first define all relevant vocabulary in the next section, then proceed to the proof.