2016

Best Graphics Card- GPU And CPU

Recently, Lisa Su, CEO of AMD, has notified how the lack of Radeon best graphics card on the market is not due to a lack of production of Global Foundries, as to a limited availability of memories for video cards ( In addition to Miners, of course ). The various manufacturers of DRAM, in fact, seem to prefer mainly the smartphone houses (LPDDRe3, LPDDR4) and console (GDDR5), leaving aside the PC users ( Even the prices of DDR4 are sky-high! ).

Things may change in the near future, as the three big sisters (Samsung, SK Hynix and Micron) are starting the large-scale production of GDDR6 memories. These memories will replace the GDDR5, used since 2008 (The first video card to mount these memories was the much appreciated Radeon HD4850 ​​AMD). SK Hynix has already announced that these memories have been available for their partners, and soon both Samsung and Micron will follow. The next generation of AMD and NVIDIA GPUs, therefore, should at least initially benefit from good availability.

Revolution of CPU and GPU

TSMC announced a new wafer stacking technology that could allow CPU and GPU manufacturers to create more powerful chips without increasing the die size. As? Overlapping them! During the Technology Symposium in progress in Santa Clara, California, TSMC has officially unveiled the new stacking technology that has been called Wafer on Wafer, or WoW, which allows two chips to be connected to each other on the same interposer , but vertically, overlapping them.

A technology similar to that used in the latest 3D NAND memories that we find in most of the SSDs announced in this period. The individual wafers are connected directly to each other using the TSVs (through-silicon vias) mode and will not only increase the number of cores in the space of a single chip, but also drastically increase the communication speed between the two components.

Obviously there is a downside, and even this technology can have weaknesses. One of these is in the solidity of the wafers. By stacking them, the possibility that one of the two breaks is not so remote and this would render the second component useless. Precisely for this reason it could be too expensive to use the new stacking technology on cheap products, and therefore low yield.

Probably therefore, at least initially we will see this WoW applied only to high-end products in particularly reliable and advanced production chains. At this point there are countless scenarios for the production of GPUs and CPUs , whose architectures could become even more long-lived thanks to updated versions using this technology to double the number of cores and increase computing power without having to distort the design.

This is a very interesting future, even if at the moment TSMC has limited itself to claiming to have the technology ready , without providing further details about the first applications of the same. We therefore have to wait for new developments.

TSMC begins mass production for 7nm chips

The TSMC giant has announced it has started mass production of the first generation 7nm chips . TSMC has over a dozen customers who will use this new production technology. Unlike the 10nm intended for mobile phones, the 7nm process is designed to be used for high-power products such as CPU, GPU, FPGA and cryptocurrency mining ASIC. This distinction makes the 7nm a substitute for the FinFET + 16nm process, the latter is currently used to create most of the chips on the market including the Nvidia Pascal best graphics card, but also the chips of the latest Playstation 4 Pro and Xbox One X consoles .

With the 7nm chips it is possible to achieve an area reduction of 70% compared to 16 nm while maintaining the same number and complexity of the transistors, also with lower consumption up to 60%. All these interesting features will lead to an evolution of current technological products.