Modern graphic cards are highly versatile because they allow the programmer to load custom code to execute onto them. This can be used to construct a structure called a per-pixel linked list, which contains all fragments composing the scene. However, with the need to render more and more complex geometry, even the most powerful hardware reaches its limit fast. To overcome this problem, the geometry is rendered on multiple systems instead of one, and finally put together for rendering. This is called remote rendering and works well for opaque scenes. The goal of this thesis is to conquer rendering of transparent objects remotely using per-pixel linked lists. Since rendering those objects requires a step called blending, standard approaches are incapable of displaying them. Three different methods are shown, compared and analyzed for their usability and performance. First, limiting the amount of depth layers is discussed. Second, identifying regions of visual change is used to reduce the amount of data to be sent. Finally, a way for reusing the previously sent fragments for the current frame is studied in detail.