If you deploy any released patches, there will be some performance hit, but it may not be as bad as you're reading elsewhere. This article by ARS Technica does a good job of explaining why the problem exists and current mitigation efforts. In my little cosmos (my home PC running an Intel i5-8600K processor) the available BIOS update from DELL hasn't slowed down much. A before/after test in Adobe Photoshop CS6 and Adobe Premiere Pro CS6 showed a 5.1% impact (Photoshop; Gaussian blur on a 14,400 x 21,600 RGB image) and 7.3% impact (Premiere; rendered a 15-minute HD compilation to MP4). In my SQL 2016 VM running on that same host, a query against almost 48,000,000 rows of data utilizing joins against 2 tables and a view showed an increase in 28 seconds (+1.2%) to return the 12,588 rows I was interested in (or, rather, the query I created was interested in). When I forced a page count of a 320MB PowerPoint print job, it took 3 extra seconds (+0.08%) to complete.
Conversely, when I used PCMark 10 just to establish some type of baseline, it gave me more dire news than I actually experienced. But benchmarking applications are intended to abuse the machine, so if your infrastructure is already stressed when under moderate load, be prepared to hear complaints (is my guess) from end users. The awesome thing about both Uniprint and Blueprint (and the MobilePrint accessory) is that a lot of the activity that would make people upset about waiting actually happens when they're doing other stuff (like walking to a printer, finishing an email or assignment, etc.), which minimizes much of the impact that any performance hit may propagate. We rely on other applications (Microsoft SQL Server) and the operating systems to provide the majority of I/O management, as do most applications due to the separation between the executive layer and the kernel, so the true target of questions is going to wind up being Microsoft. We're just along for the ride.
My thinking is that virtualized environments (both zero-desktop and virtual servers) will actually suffer a little less than direct-hardware implementations because virtualization creates, really, another abstraction layer. But that's just my opinion and limited experience "post patch."