Split Horizon is a loop prevention method but why we need to shut it down in Frame Relay multipoint design but not in point-to-point design?

In Point-to-Point you have only one path , therefore no looping can occur

It’s because when we are creating a point to multipoint scenario via frame relay, in which all our sites has to be on same subnet, all the sites uses only one interface, no sub interfaces, to communicate with each their sites and the (different) DLCI, which helps in creating PVCs, are being used to communicate with other sites are also, dynamically or statically depending on the configuration, mapped on only one interface. And when we deploy our routing protocol in this scenario, we do not get proper routing convergence because according to the Split Horizon rule if a router receives a routing update on an interface, let’s assume that Serial0 is connected with frame Relay cloud or switch, then the router is not supposed to send that same routing information back out the same interface that it has just received, in our case its serial0. So that’s why we turn off (not recommended) the split horizon rule on the interface, so that routers can send the routing updates back on the same interface. The other way would be to use static routes, obviously, instead of routing protocol, which would work perfectly fine.            

I hope that answers your question

Assalam-o-Alaikum 

P.S You can also refer to this video, which might help also INS SHA ALLAH  

Configuration of Frame Relay in Hub & Spoke topology