Morrowind est le rpg pc le plus proche des rpg de plateaux dans ce domaine, tu peux vraiment orienter ton perso comme tu le souhaites, et il y a une vraie liberté, pas une fausse liberté comme dans plein d'autres rpg

"XML is like violence: if it doesn't work, use more."

Et puis qu'est-ce qu'il y a d'original dans arcanum ? J'aimerais bien le savoir.

et avec les mods si on  a le malheur d'en installer un qui est mal balancé on se fait encore plus chier.

Speed Test
Votre Bande Passante  11756.314 Kbps (1469.539 Ko/sec)

Stiff systems
 
A system may include both fast-moving and slow-moving dynamics, which are represented by the time constants of the system model. We characterize the stiffness of a system as the ratio of its longest time constant to its shortest time constant. We consider a system to be stiff if this ratio is roughly three orders of magnitude or higher. When simulating stiff systems, we must use care in the choice of integration method to ensure that the solution remains numerically stable and produces accurate results while executing with reasonable efficiency.
 
Consider the open-loop dynamic system represented by the Simulink model shown in Figure 4. This system consists of a first order actuator driving a first order plant. In this example, the actuator has a time constant of 1ms and the plant has a time constant of 1s. The system is driven by a unit step input.
 
If we simulate the system of Figure 4 using the forward Euler integration method and a step time of 1.9ms for a period of 10s, we find that the peak magnitude of the integration error is about 0.001. If we increase the step time slightly to 2.01ms the simulation becomes numerically unstable and the truncation error grows without bound. The point of this example is that even in this case of mild stiffness, there is only a little degradation of simulation accuracy before instability occurs.
 
Many real world dynamic systems have several orders of magnitude more stiffness than the example of Figure 3. To simulate these very stiff systems with an arbitrarily selected integration algorithm, it may be necessary to take extremely small time steps to accommodate the shortest system time constant. This is inefficient because the system components with longer time constants do not require the same small integration steps needed to simulate the faster components. In fact, if a very small time step is used for the slow-moving components, the accuracy of the overall simulation may suffer due to roundoff error that accumulates over the large number of integration steps required.
 
Two approaches are commonly used in the simulation of stiff systems. The first is to select an integration algorithm that works well in the presence of stiffness. For non-real-time simulation applications, many integration algorithms are available that perform well with stiff systems. The implicit algorithms introduced above, for example, have good performance in the presence of stiffness. The Gear family of implicit algorithms is intended primarily for the simulation of stiff systems.3 By using integration algorithms with good performance in the presence of stiffness, it is possible to maintain simulation accuracy and stability while taking much larger time steps than would be possible with other algorithm types such as Runge-Kutta.
 
A second approach to the simulation of stiff systems involves the use of multiframing. Multiframe simulations break the simulation into multiple parts that execute at different frame rates. In the example of Figure 4, we might run the actuator simulation with a 0.1ms step time and the plant simulation with a 0.1s step time.
 
The multiframe method adds complexity because we must extrapolate or interpolate any values from the slower frame rate subsystem that are used as inputs by the fast frame rate subsystem. There may also be some difficulty in performing I/O at fixed intervals if the various frame rates are running on a single processor.
 
Of these two approaches to simulating stiff systems, only the multiframing method is applicable for real-time applications. This is due to the unsuitability of implicit integration algorithms in situations where real time I/O is required. In non-real-time simulations, it is often simplest to employ an integration algorithm with good performance in the presence of stiffness, which avoids the difficulties associated with multiframing.