Arguably the best gaming mouse has full linux support, in that all settings can be directly set on the mouse itself, not through external software, and that's Zowie. There is also full linux open source support for Logitech peripherals, from the G5 to the G400s, so both for IR-laser-focused and visible-light-through-lens-focused sensor mice for people that like the classic Logitech ergonomics.
Most mice will allow you to directly edit the settings on the mouse or keyboard controller though, except for instance Microsoft and Apple hardware products, of which the internals are obfuscated in every sense. Corsair and Razer are also mostly obfuscated, and so are other typical "gaming" brands. They are still class compliant though, and open source software offers flexibility that doesn't require the hardware to be as flexible as on Windows. For instance, most open source games have an unlimited config range for sensitivity, and acceleration is not usual in open source applications, and raw mouse input is pretty much default X11 behavior... a lot of the typical gaming peripherals come with features that solve problems that are specific to MS-Windows and MS-Windows-only applications. Just about any 10 USD optical mouse will probably give you the same degree of performance you would get from an expensive gaming mouse in MS-Windows, if you can live with the build quality and the ergonomics.
In general, if you want software flexibility on open source, it's useful to run hardware that's a few years old. If you buy hardware that was released in the last two years, well, then you will run into problems with things like Debian Stable and Ubuntu LTS. New hardware is not fun, because most of the time, the performance increase is not worth the price for mainstream open source users, because the mainstream applications just aren't very demanding. Commercial applications might be nerfed, like CS:GO for linux for instance, but open source applications that do the same, display huge performance (far beyond the performance on even the heaviest Windows systems) even on very modest systems. AMD systems have the benefit of having a slower release cycle and the benefit of supporting all of the most important features throughout the range. All AMD systems will offer full hardware passthrough virtualization and bug-free/unspiked encryption modules, etc... things that are important to open source users.
There is very few hardware that doesn't work in linux. Most hardware is just supported by the linux kernel. There are a few exceptions of hardware that is only supported through hacks, but for which there is no native open source support. These are mainly only some older Broadcom Wi-Fi chips and of course nVidia GPU's.
Through crowdfunding, an open source driver for nVidia GPU's was developed through reverse engineering nVidia chips with an electron microscope. These open source nVidia drivers do not offer the performance of the closed source nVidia proprietary drivers, and they will never reach the same level of performance, because nVidia just won't give crucial information to the Linux Foundation about their hardware. Until 2012, nVidia had a pretty good Linux driver (proprietary of course), but since the summer of 2012, they've pretty much stopped developing that, and the linux drivers by nVidia are pretty awful. AMD used to do the same as nVidia, but even when they're more influential now than nVidia in the GPU market (which is dominated by Intel by the way, more than half of the world's GPU's are made by Intel), AMD has made it a priority to migrate towards open source friendly drivers. AMD does that in two ways: 1. they have a specialist on payroll that works on the open source driver development, and 2. they have provided open source kernel modules that work with the proprietary Catalyst driver (which can not be open sourced because it contains 2D acceleration licenses that don't belong to AMD), so that there is no more need for binary blobs in the kernel, which eliminates problems when upgrading the kernel, and eliminates security and privacy risks. Intel only has open source drivers, but - despite of Intel having by far the largest team of developers - they are having huge problems getting their shit to keep up with AMD for instance. For instance, Beignet, the OpenCL acceleration component for Intel GPU's, hardly works after over two years of development. So for basic use, Intel is a pretty good solution, but AMD also, because a modern kernel will work with AMD GPU's just fine, with 80+ % of the performance of Catalyst, and even the gamers that want more, can still use Catalyst in a pretty safe manner on bleeding edge systems.
On linux, to have flexibility in using older kernels, it's usually a good idea to keep one hardware generation behind on hardware products with a fast release cycle. The performance difference mostly isn't worth the extra cost anyway, and early adopters are basically just people that do the testing for the others. It's necessary to test stuff in open source, but not necessary to pay extra for maybe, the manufacturers should (but most often don't) provide new hardware to open source communities like they do to closed source companies, and the users should not have to pay for doing the dirty work. If you're part of a dev community, it might be a good choice to go for blazing hot new hardware, to help development, and the benefit of doing that in open source is, that you can get credit for the solutions, and that you get feedback on the state of things, and that everything just works in terms of the existing features and performance for that category of hardware, even though new features might take a few weeks to get implemented by the open source community. On closed source though, customers that buy blazing hot new hardware, are simply exploited by manufacturers for debugging, and are not informed correctly about what does and what does not work. For instance, on Windows, nVidia's Maxwell based GPU's have not been delivering working CUDA acceleration. In fact, using the proprietary drivers, the amount of open source applications in linux that have gotten CUDA to work for Maxwell, is higher than the amount of MS-Windows applications that have gotten CUDA to work on Maxwell, which is quite ludicrous...
On open source linux, AMD's offering is quite compelling for mainstream users at the moment: for little money, it's possible to have a well matured and full-featured high performance linux system based upon AMD's FX range or APU range, with a high performance AMD GP-GPU. Of the latest AMD GPU series, I particularly like the R9 285, because it's lower power, which again saves money, while still offering ample graphical and floating point performance, and a good audio experience, using only open source software. I would go for a 970 or 990FX based mobo, a good AiO cooler for the CPU, and an R9 285, which is pretty silent and doesn't require a very expensive PSU. RAM is actually more important on linux than it is for most users on Windows. The linux operating system and applications use a lot less RAM than closed source applications, but linux and open source applications can actually get more benefit from more RAM than most Windows systems. Most Windows games are still 32-bit games, even the Steam Client (also in Linux) is 32-bit, which means that it will never use more than 3.6 GB of RAM. That shows that Windows and Windows software is really holding the systems back, when people are spending ever more money on ever more powerful CPU's and GPU's, to run applications that require just about the same amount of RAM as they did 15 years ago. In linux, RAM usage is much less than in Windows, but extra RAM will lead to much better performance of common applications like Firefox (has been 64-bit for ages in linux, still 32-bit in Windows), LibreOffice Calc (huge spreadsheets will not impact performance, also because LibreOffice Calc uses OpenCL acceleration, so whereas for instance a humongous spreadsheet on Excel in Windows will become a dog to work with, it will still be snappy in Calc on linux, even on a much more modest system, and even for the same amount of RAM, and if the system starts to struggle, on the linux system all you need is a RAM upgrade to be able to load a larger spreadsheet in RAM, but on the Windows system, a lot more hardware upgrade will be needed, in particular a stronger CPU, because there is no OpenCL acceleration, and thus a new mobo, etc...). So I would go with 16 GB of RAM as a minimum, so that there is some space for some virtual machines, etc... just more functional flexibility.
Another thing with linux, is that the desktop PC isn't the central piece of the linux environment any more. A lot of devices run linux these days, and they all work together out of the box in a linux environment. No more pesky proprietary protocols that make it impossible to use advanced features out of the box, etc... no more need for spyware-enabled proprietary software to connect interestingly priced appliances like media servers or NAS boxes from the WD consumer line for instance. For 100 USD, you can actually buy a WD MyCloud with 2 TB of space, and WD delivers a free DNS service that is pretty well thought out, based upon the hardware identifier and full strong encryption. If you use that with the WD software, it's of course not that interesting from a privacy point of view, but these devices (and also devices like the WD-TV Live Hub for instance), run linux inside, and are just out of the box compatible with any linux machine, without any downloaded application for Windows, without any specific software that spies upon you, etc... it just works very well right out of the box, and you don't even have to pay for a DNS service to roll out your own cloud if all you need is ftp and dlna. So buying hardware as an open source user, is a completely different thing than buying hardware for closed source. An open source user sees things completely differently, selects devices for other reasons, because the software is much more important than the hardware, and there is no need to buy ever more hardware to compensate for the failure of the software. Open source users don't buy new hardware because they need more performance very often, they mostly buy new hardware because they want more features and functionality, like NASes, 3D printers, dev kits, larger and more monitors for more clutter free vertical desk space that makes life simpler and work more efficient, etc...