Tomohiro Tanaka

The 10th Generation 16-Core SPARC64™ Processor for Mission Critical UNIX Server

The 10th generation SPARC64™ processor named SPARC64 X contains 3-billion transistors on a 588mm2 die fabricated in an enhanced 28nm high-κ metal-gate (HKMG) CMOS process, with 13 layers of copper interconnect with low-κ dielectrics. More stress control, SiGe improvement and S/D optimization achieve about 10% higher performance than the standard 28nm high performance (28HP) process. SPARC64 X runs at 3.0GHz and consists of 16 cores, shared 24MB level 2 (L2) cache, four channels of 1.6GHz DDR3 controller, two ports of PCIe Gen3 controller, and five ports of system interface controller. ccNUMA is adopted as its memory system, and a cache coherence control unit for multi-chip systems with up to 64 processors is integrated into L2 cache control circuitry for lower latency and reduced area and power consumption.

Micromagnetic simulation of the orientation dependence of grain boundary properties on the coercivity of Nd-Fe-B sintered magnets

This paper is focused on the micromagnetic simulation study about the orientation dependence of grain boundary properties on the coercivity of polycrystalline Nd-Fe-B sintered magnets. A multigrain object with a large number of meshes is introduced to analyze such anisotropic grain boundaries and the simulation is performed by combining the finite element method and the parallel computing. When the grain boundary phase parallel to the c-plane is less ferromagnetic the process of the magnetization reversal changes and the coercivity of the multigrain object increases. The simulations with various magnetic properties of the grain boundary phases are executed to search for the way to enhance the coercivity of polycrystalline Nd-Fe-B sintered magnets.

Electrical-field and spin-transfer torque effects in CoFeB/MgO-based perpendicular magnetic tunnel junction

The electric-field (E) dependence of the magnetoresistance (RH) loops for top-pinned perpendicular CoFeB/MgO-based magnetic tunnel junctions (MTJs) in the presence of a spin-transfer torque (STT)-current was measured. The E effects were distinguished from the STT-current effects using a micromagnetic simulation. The coercive field (Hc) decreased and the RH loop shifted as both the positive and negative bias E increased owing to the STT current. Furthermore, E-assisted switching for an MTJ with a diameter of 20 nm, which exhibited a nearly coherent magnetization reversal, was demonstrated using micromagnetic simulation.

Semi-Implicit Steepest Descent Method for Energy Minimization and Its Application to Micromagnetic Simulation of Permanent Magnets

To reveal the relationship between coercivity of permanent magnets and their microstructure, micromagnetics is one of the key technologies for analyzing the magnetization reversal mechanism. Especially, microscopic reversal process such as nucleation of a reversed domain and domain wall pinning has been investigated by micromagnetic simulation [1]. However, since the mesh size in the simulation should be less than the exchange length of a magnet to calculate the domain wall structure accurately, the overall dimensions of models for such simulation is practically limited to very small length scales. Therefore, further improvement in calculation speed are demanded to simulate reversal process of more realistic microstructure.

Micromagnetic simulation of electric-field-assisted magnetization switching in perpendicular magnetic tunnel junction

The feasibility of a voltage assisted unipolar switching in perpendicular magnetic tunnel junction (MTJ) has been studied using a micromagnetic simulation. Assuming a linear modulation of anisotropy field with voltage, both parallel (P) to anti-parallel (AP) and AP to P switchings were observed by application of unipolar voltage pulse without external magnetic field assistance. In latter case, the final P state can only be achieved with an ultrashort voltage pulse which vanishes before spin transfer torque (STT) becomes dominant to restore the initial AP state. In addition, it was found that the larger change in anisotropy field is required for the MTJ with smaller diameter.

Loss Simulation by Finite-Element Magnetic Field Analysis Considering Dielectric Effect and Magnetic Hysteresis in EI-Shaped Mn-Zn Ferrite Core

Power electronic devices such as inductors and transformers are required to be driven with high frequency according to downsizing. Mn-Zn ferrite is one of the high-frequency magnetic materials. The dimensional resonance occurs in Mn-Zn cores due to the increase of the dielectric constant and significantly affects the eddy current loss [1]. The equivalent RC circuit of Mn-Zn ferrite was modeled by the grains and their boundary layers and can explain the effective dielectric property by the contribution of the capacitance [2]. The boundary layers with high-resistance suppress the eddy current in the grains at low frequencies, while as frequency increases the suppression of the eddy current decreases by charge accumulation on the surface of the grains. The calculating method of the frequency dependent dielectric property by the capacitance was proposed and the dimensional resonance was reproduced by applying the method to the magnetic field equations of linear magnetic materials by using a cylindrical approximation [3]. In order to analyze the eddy current loss of complex shaped inductors at high frequencies, we apply the dielectric effect to the A-

Magnetic Field Analysis for Dimensional Resonance in Mn–Zn Ferrite Toroidal Core and Comparison With Permeability Measurement

\varphi

Speeding Up Micromagnetic Simulation by Energy Minimization With Interpolation of Magnetostatic Field

method of the finite element magnetic field analysis. On the other hand, the magnetic hysteresis loss increases according to the increase of the magnetic flux density in the core. Therefore, we used the play model [4] to express the magnetic hysteresis for finite amplitude of magnetic flux. For confirming the calculation accuracy, core losses of an EI-shaped inductor was calculated and the frequency dependent loss was compared with experimental results. In the simulation, the core loss was divided into hysteresis loss in DC, eddy current loss and excess loss, and their contributions were analyzed. II. METHOD The current density of the grains is

Speeding up micromagnetic simulation by energy minimization with interpolation of magnetostatic field

j_{1}

ARITHMETIC PROCESSING UNIT

as follows [3].