Graphs are relatively expensive front panel objects, especially if they contain a lot of points, are set to auto-scale, overlap with other objects, etc.
Even without syncronous display, you might want to put the graphs into case structures and only send data to them if necessary (e.g. max 2/sec and only if the data has changed) and not at each loop iteration. Don't autoscale (or autoscale only at longer intervals).
Use simple lines and dots instead of fancy filled circles for the points.
Some of my old LabVIEW applications are still running on P100 machines without significant load on the CPU. They key is in the programming technique. It is virtually never necessary to constantly throw newer and faster CPUs at the problem, even if it is able to temporaril
y mask poor programming.