Fedora in general tends to have a more liberal update policy than Ubuntu and others when it comes to stable releases of software; new versions of the Linux kernel are shipped down to stable releases of Fedora, etc. With Fedora 21 not arriving until late in 2014, exceptions have been given to also ship new Mesa updates for Fedora 20 users to provide a more modern and updated hardware experience. For those curious how Fedora 20's performance compares to when it made its debut in December to how it performs now with all official stable updates, here's some benchmarks.
Fedora 20 shipped with the Linux 3.11 kernel and Mesa 9.2.3 but stable updates now have Fedora 20 as of this week with the Linux 3.13 kernel (no Linux 3.14 yet) and Mesa 10.1. The desktop also moved from GNOME Shell 18.104.22.168 to 3.10.4. The compiler is obviously still at GCC 4.8.2 and will remain that way until Fedora 21 when GCC 4.9 makes its debut within the Fedora camp. Most of the performance changes from upgrading F20 will be found with the open-source driver (Intel / Radeon / Nouveau) graphics performance given the big updates to the Linux kernel and Mesa. File-system performance may also change somewhat between kernel releases but the overall CPU performance is unlikely to be changed between 3.11 and 3.13~3.14 unless you were previously impacted by a regression or odd ACPI/PM clocking issue.
Between Linux 3.11 and 3.13 has been dynamic power management being enabled for some Radeon GPUs in Linux 3.13 (the newer Southern Islands hardware received DPM in Linux 3.14), some initial Intel Broadwell graphics support (Linux 3.14~3.15 and Mesa 10.2 will be ideal for the new Intel CPUs), new OpenGL extensions being enabled in Mesa 10.0~10.1, and various other new features especially for the Intel and Radeon drivers (the biggest Nouveau limitation still remains for the NVIDIA hardware with the lack of re-clocking support).
In this article are benchmarks of Fedora 20 stock against Fedora 20 with all stable release updates as of 21 April 2014. All benchmarking was handled in a fully-automated manner using the open-source Phoronix Test Suite benchmarking software.