Abstract: A fan module and computing device with the fan module are disclosed. The fan module includes a handle configured to actuate between an operation state and a release state. The handle in the release state allows a user to vertically remove the fan module from its respective fan module slot and away from the bottom panel.

Abstract: The present disclosure provides a battery charging system and method. The battery charging method includes: determining a degree of healthy of a battery module according to an evaluation mechanism; setting a charging standard according to the degree of healthy; by handshaking with a charger, setting a charging voltage for the charger according to the charging standard to charge the battery module; and by the charger, perform a charging operation on the battery module until a fully charged condition is satisfied.

Abstract: A semiconductor device and a method of forming the same is disclosed. The semiconductor device includes a substrate, a first well region disposed within the substrate, a second well region disposed adjacent to the first well region and within the substrate, and an array of well regions disposed within the first well region. The first well region includes a first type of dopants, the second well region includes a second type of dopants that is different from the first type of dopants, and the array of well regions include the second type of dopants. The semiconductor device further includes a metal silicide layer disposed on the array of well regions and within the substrate, a metal silicide nitride layer disposed on the metal silicide layer and within the substrate, and a contact structure disposed on the metal silicide nitride layer.

Abstract: Disclosed is a method for enhancing memory utilization and throughput of a computing platform in training a deep neural network (DNN). The critical features of the method includes: calculating a memory size for every operation in a computational graph, storing the operations in the computational graph in multiple groups with the operations in each group being executable in parallel and a total memory size less than a memory threshold of a computational device, sequentially selecting a group and updating a prefetched group buffer, and simultaneously executing the group and prefetching data for a group in the prefetched group buffer to the corresponding computational device when the prefetched group buffer is update. Because of group execution and data prefetch, the memory utilization is optimized and the throughput is significantly increased to eliminate issues of out-of-memory and thrashing.

Abstract: A method for flattening a three-dimensional shoe upper template is provided. The method includes providing a three-dimensional last model, obtaining a three-dimensional grid model, obtaining a three-dimensional thickened grid model, obtaining a two-dimensional initial-value grid model, and obtaining a two-dimensional grid model with the smallest energy value. A system and a non-transitory computer-readable medium for performing the method are also provided. The method makes it possible to precisely flatten a three-dimensional last model with a non-developable surface and thereby convert the three-dimensional last model into a two-dimensional grid model.

Abstract: An ionic compound, an absorbent and an absorption device are provided. The ionic compound has a structure represented by Formula (I): ABn, ??Formula (I) wherein A is B is R1, R2, R3, R4, R5, and R6 are independently H, C1-6 alkyl group; and n is 1 or 2.

Abstract: A package structure and a method of forming the same are provided. The package structure includes a first die, a second die, a first encapsulant, and a buffer layer. The first die and the second die are disposed side by side. The first encapsulant encapsulates the first die and the second die. The second die includes a die stack encapsulated by a second encapsulant encapsulating a die stack. The buffer layer is disposed between the first encapsulant and the second encapsulant and covers at least a sidewall of the second die and disposed between the first encapsulant and the second encapsulant. The buffer layer has a Young's modulus less than a Young's modulus of the first encapsulant and a Young's modulus of the second encapsulant.

Abstract: A multi-tap Differential Feedforward Equalizer (DFFE) configuration with both precursor and postcursor taps is provided. The DFFE has reduced noise and/or crosstalk characteristics when compared to a Feedforward Equalizer (FFE) since DFFE uses decision outputs of slicers as inputs to a finite impulse response (FIR) unlike FFE which uses actual analog signal inputs. The digital outputs of the tentative decision slicers are multiplied with tap coefficients to reduce noise. Further, since digital outputs are used as the multiplier inputs, the multipliers effectively work as adders which are less complex to implement. The decisions at the outputs of the tentative decision slicers are tentative and are used in a FIR filter to equalize the signal; the equalized signal may be provided as input to the next stage slicers. The bit-error-rate (BER) of the final stage decisions are lower or better than the BER of the previous stage tentative decisions.

Abstract: In accordance with some embodiments, a package-on-package (PoP) structure includes a first semiconductor package having a first side and a second side opposing the first side, a second semiconductor package having a first side and a second side opposing the first side, and a plurality of inter-package connector coupled between the first side of the first semiconductor package and the first side of the second semiconductor package. The PoP structure further includes a first molding material on the second side of the first semiconductor package. The second side of the second semiconductor package is substantially free of the first molding material.

Abstract: A package structure and method of forming the same are provided. The package structure includes a polymer layer, a redistribution layer, a die, and an adhesion promoter layer. The redistribution layer is disposed over the polymer layer. The die is sandwiched between the polymer layer and the redistribution layer. The adhesion promoter layer, an oxide layer, a through via, and an encapsulant are sandwiched between the polymer layer and the redistribution layer. The encapsulant is laterally encapsulates the die. The through via extends through the encapsulant. The adhesion promoter layer and the oxide layer are laterally sandwiched between the through via and the encapsulant. A bottom portion of the encapsulant is longitudinally sandwiched between the adhesion promoter layer and the polymer layer.

Abstract: A die includes a substrate, a conductive pad, a connector and a protection layer. The conductive pad is disposed over the substrate. The connector is disposed on the conductive pad. The connector includes a seed layer and a conductive post. The protection layer laterally covers the connector. Topmost surfaces of the seed layer and the conductive post and a top surface of the protection layer are level with each other.

Abstract: Disclosed is a method for enhancing memory utilization and throughput of a computing platform in training a deep neural network (DNN). The critical features of the method includes: calculating a memory size for every operation in a computational graph, storing the operations in the computational graph in multiple groups with the operations in each group being executable in parallel and a total memory size less than a memory threshold of a computational device, sequentially selecting a group and updating a prefetched group buffer, and simultaneously executing the group and prefetching data for a group in the prefetched group buffer to the corresponding computational device when the prefetched group buffer is update. Because of group execution and data prefetch, the memory utilization is optimized and the throughput is significantly increased to eliminate issues of out-of-memory and thrashing.

Abstract: A temperature-sensing and humidity-controlling fiber includes a hydrophilic material and a temperature-sensing material. The temperature-sensing material has a lower critical solution temperature (LCST) between 31.2° C. and 32.5° C. when a light transmittance thereof is in a range from 3% to 80%, in which a wavelength of the light is between 450 nm and 550 nm.

Abstract: A method of manufacturing a semiconductor device includes forming a first layer of a first planarizing material over a patterned surface of a substrate, forming a second layer of a second planarizing material over the first planarizing layer, crosslinking a portion of the first planarizing material and a portion of the second planarizing material, and removing a portion of the second planarizing material that is not crosslinked. In an embodiment, the method further includes forming a third layer of a third planarizing material over the second planarizing material after removing the portion of the second planarizing material that is not crosslinked. The third planarizing material can include a bottom anti-reflective coating or a spin-on carbon, and an acid or an acid generator. The first planarizing material can include a spin-on carbon, and an acid, a thermal acid generator or a photoacid generator.

Abstract: An n-type metal oxide semiconductor transistor includes a gate structure, two source/drain regions, two amorphous portions and a silicide. The gate structure is disposed on a substrate. The two source/drain regions are disposed in the substrate and respectively located at two sides of the gate structure, wherein at least one of the source/drain regions is formed with a dislocation. The two amorphous portions are respectively disposed in the two source/drain regions. The silicide is disposed on the two source/drain regions, wherein at least one portion of the silicide overlaps the two amorphous portions.

Abstract: A modular headphone system includes a headband holding an earphone module at each end of two ends thereof, a circuit module formed of an active noise cancelling module, a Bluetooth module, high-frequency module and/or a multi-channel audio controller and detachably mounted in one earphone module, and a power module detachably mounted in the other earphone module. Each earphone module includes an earphone body pivotally connected to one end of the headband, a speaker holder mounted at the earphone body and electrically connected to the earphone body, a speaker mounted in the speaker holder and electrically connected to the speaker holder and an ear cushion covered on the speaker holder over the speaker.

Abstract: A modular headphone system includes a headband holding an earphone module at each end of two ends thereof, a circuit module formed of an active noise cancelling module, a Bluetooth module, high-frequency module and/or a multi-channel audio controller and detachably mounted in one earphone module, and a power module detachably mounted in the other earphone module. Each earphone module includes an earphone body pivotally connected to one end of the headband, a speaker holder mounted at the earphone body and electrically connected to the earphone body, a speaker mounted in the speaker holder and electrically connected to the speaker holder and an ear cushion covered on the speaker holder over the speaker.

Abstract: An extended translation look-aside buffers (eTLB) for converting virtual addresses into physical addresses is described, which includes both a set associative memory structure (SAM) and a content addressable memory (CAM) structure. An improved approach for operating an eTLB is described in which the same instruction is issued to perform the same task regardless of the exact underlying memory structure within the eTLB being accessed. For flush operations, the same instruction to perform a TLB flush can be provided to the eTLB that operates upon both the CAM and the SAM, which is then handled differently by the underlying implementation system of the eTLB depending upon whether the CAM and/or SAM is being accessed.

Abstract: An architecture is described for performing memory management in a virtualization environment. Multiple levels of caches are provided to perform address translations, where at least one of the caches contains a mapping between a guest virtual address and a host physical address. This type of caching implementation serves to minimize the need to perform costly multi-stage translations in a virtualization environment.

Abstract: An architecture and method are described for performing memory management. A page walker cache is provided to cache data used during the page walk process. This cache structure speeds up the page walk process, which significantly reduces the expense of performing a page walk. The page walker cache also reduces the cost associated with usage of memory access bandwidths.