A 10th generation SPARC64 processor, fabricated in enhanced 28 nm CMOS, runs at 3.0 GHz and contains 16 cores with 24 MB shared L2 cache and system/DDR3/PCIe interfaces in 588 mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> die area. Using H-tree clock distribution network with shield wires, the average clock skew is minimized to 20 ps. Two-step read structure of GPR enables out-of-order execution across register windows. A large SMP system of up to 64 CPUs with ccNUMA uses a newly developed 14.5 GB/s SerDes. Column separation, alternate placement of master and slave latches and well slits are used to mitigate soft errors especially for multi-bit upsets. SER reduction is observed by neutron irradiation experiments.