Well... What you know about microprocessors and how microprocessor works?

https://towardsdatascience.com/your-processor-is-trying-to-predict-the-future-long-before-you-start-with-ai-892beeaf6a48 https://www.quora.com/CPUs-How-is-branch-prediction-implemented-in-microprocessors

Whether branch prediction is a good idea depends on the microarchitecture of the processor. For example, if we assume a single cycle processor, we do not need branch prediction whatsoever. This is because the instruction will be fetched, decoded, and executed in the same clock cycle. This allows us to check the condition of the branch, and determine whether or not the branch takes or fails, all before we decide what the next address of the next instruction is. In a single cycle design we have 100% confidence that whatever our next address of the next instruction is, and therefore no prediction is needed. Of course, no one builds single cycle machines anymore. Recall that the clock period of the single-cycle processor is limited by the longest instruction. This is a fairly severe limitation, and will absolutely inhibit processor performance.

The problem is therefore much,much lower performance of microprocessor.

And thing is big,as with branch prediction you have for example 4-5 operations in one cycle,while one waiting without branch prediction with only one branch could cost 10-20 cycles !