Solid-state drives are gaining popularity in the marketplace due to its high-speed read and high-speed interfaces and 3D flash cells that will create new data storage architectures. In the future, SSDs will have a capacity of several terabytes and can transfer several gigabytes of data per second, which will gradually replace mechanical hard disks in the future. Below, the system home will introduce these technologies in advance. SSDs load operating systems and applications two to three times faster than mechanical hard drives, so computers without SSDs are a bit out of date. However, in fact, SSDs can only fully play their performance when they are used as mass storage devices in the server. On the SATA interface of personal computers, SATA connections can only provide 300MB per second (SATA 2.0) according to different standards. Also known as the 3G interface) or 600MB per second (SATA 3.0 is also known as the 6G interface), there is no chance for the SSD to perform its full performance. In fact, the current 2.5-inch hard drive enclosures used in SSDs are designed to allow mechanical drives to be used on laptops. They are tuned to be the most suitable for mechanical hard disk read/write heads to read data on rotating metal disks. The solid state hard disk can be made very small, a memory chip is actually as small as a CPU, a plurality of memory chips can be mounted in the solid state hard disk, and the memory chip can be read or written in parallel by the control chip. Can be much faster than the SATA interface. Soon we will see new interfaces on computer motherboards and laptops. The new interface can support faster data transfer speeds and can transfer several gigabytes of data per second, which will fully realize the potential of flash technology. Second, the data density of SSDs will increase dramatically, and the price of flash storage media will drop significantly. We can get 500GB SSDs at a low price. And, thanks to the new flash cell model, we will be able to get more affordable TB-class SSDs in the future. High-speed interfaces Next-generation SSDs are poised to take off, and they will be faster than all SSDs on the market today. They will use a new interface: SATA Express (Serial ATA Express, SATAe for short), which will forward SATA data via PCI-E. So far, only the graphics card is transmitting data via PCI-E. The SATAe interface connected to the PCI-E can be connected to two SATA plugs. The transmission rate of the SATA interface can reach up to 600MB/s, while the transmission rate of the old generation PCI-E 2.0 can reach 400MB/s per channel. The new generation PCI-E The transmission rate per channel of 3.0 can be as high as 1GB/s. Since the SSD of the SATAe interface uses at least two channels to transmit data through PCI-E, the transmission rate of SATAe will be between 800MB/s and 2GB/s. But SATAe upgrade is not just an interface, but a new interface standard. For compatibility reasons, SATAe will continue to support the Advanced Host Controller Interface (AHCI) standard. The AHCI standard is an interface standard jointly developed by a number of companies under the leadership of Intel in 2004. The response time of this standard is slow. This standard that should be replaced soon is about to be replaced by a new interface standard: the Non-Volatile Memory Host Controller Interface (NVMHCI) standard is specifically for SATAe high-performance SSDs. Developed interface standard, NVMHCI's instruction set transfers data in parallel with the number of CPU cores. The CPU can directly process all storage and read instructions without queuing. In addition, NVMHCI transfers commands directly between the CPU and the SSD controller without intermediate storage, resulting in lower response times. Flash hardware optimization These are just theories or are they really feasible? In fact, SSDs that transmit data over PCI-E have been used on servers for a long time, and these SSDs can achieve transfer rates well above 4GB/s. However, due to the lack of drivers, similar devices cannot be installed on a personal user's computer. The only exception is the RevoDrive (a combination of solid state drives and a PCI-E adapter). The M.2 SSD is a transitional product designed for notebooks and intended for individual users. Currently, the M.2 SSD is still using the SATA interface, but the new SATAe interface will allow M.2 to connect via PCI-E. The Plextor M6E and the Samsung XP941 are M.2 SSDs that can be connected to PCI-E. They can be used with the new Intel 9-series chipset motherboards that are compatible with the SATAe standard. Unfortunately, the motherboards of the H97 and Z97 chipsets allow up to two PCI-E 2.0 channels, which limits the SSD's transfer rate to 800MB/s. For SSDs like the Plextor M6E, the problem is small, because the two gaps on the SSD socket (commonly known as B Key) indicate that it uses only two channels for transmission. However, for the Samsung XP941, there is only one gap in the socket (commonly known as M Key), which can use the four-channel solid-state hard disk, and the maximum speed can not be achieved by the constraints of the motherboard. For a high-speed solid-state drive such as the Samsung XP941, you will need a solid-state drive adapter card using a PCI-E slot or ASRock's Z97 Extreme6 motherboard. Currently, this motherboard can provide a 4-channel PCI-E connection for M.2. The future PCI-E 3.0 interface will enable SATAe SSDs to fully utilize their performance. The current control chip can only support PCI-E 2.0. However, the SandForce 3700 will be launched soon. In addition to supporting AHCI, the control chip can also Support for NVMHCI. Next, Toshiba's PCI-E 3.0 controller OCZ Jetstream Express will be available next year, and Intel plans to introduce a chipset that supports PCI-E 3.0. The price of a large-capacity SSD with a 3D flash memory capacity of up to 4TB is still quite affordable, but the SSDs with similar prices are only 250GB to 500GB. However, this is only the current situation. It is expected that the size and design of flash memory units will reach a new stage in the next few years, and SSDs will be able to completely reverse this situation. Currently, the flash memory cells used in SSDs are like the transistors of a CPU. The only difference is that the flash cells can permanently store charge in the floating gate. Currently, manufacturers have successfully upgraded the flash cell fabrication process to 20nm and are rapidly moving toward another, smaller process. Every improvement in technology means a reduction in costs and production costs, so manufacturers are happy to switch to another type of flash unit: 3D flash. The capacity of SSDs is another issue that affects the price. At present, the largest SSD SanDisk Optimus Max has a capacity of 4TB, but basically only enterprise users will spend tens of thousands of dollars to buy. For the SSDs used by individual users, this flash technology is unrealistic. Samsung and Toshiba are currently taking another path. They achieve the purpose of storing 3 bits of data (8 different charges representing a binary value between 000 and 111) for each flash cell by storing TLC flash cells of 8 different charge levels in the floating gate, by increasing the storage density. Increase the capacity of SSDs. However, this flash unit can only support 1 000 write or delete operations. Normally, the MLC flash memory unit used by the SSD for individual users can record 2 bits of data per unit and can withstand 10,000 write or delete operations. The server solid state hard disk PM853T using TLC unit, Samsung's Device Writes Per Day (DWPD, which indicates the amount of data that the SSD can write every day) is 0.3~1.6 times for the whole disk. It can guarantee a life of 5 years. Obviously, the PM853T is only suitable for data that is rarely refreshed. In contrast, other common server SSDs, DWPD are usually between 10 and 30. The limitations of 2D flash memory If you want to achieve high cost performance while maintaining high capacity and high quality, the only way is to reduce the size of the flash memory unit. However, if the size of the flash memory unit is further compressed, the flash memory unit may have problems if the size is less than 20 nm. First, to solve the limitation of laser burning at 193 nm wavelength in the production process, the ultra-ultraviolet exposure tool that solves this problem is still too expensive for the production of flash memory cells. Moreover, manufacturers are also faced with physical limitations. When the size of the unit casing is less than 20 nm, some components will have only a few atomic layers and it is impossible to continue to shrink. In addition, an inter-poly dielectric layer (IPD) between the control gate and the floating gate may also cause problems. The control gate discharges or fills the charge of the floating gate by applying a voltage, and the floating gate must maintain its charge, so that its surroundings are surrounded by IPD, preventing the charge from escaping through the control gate. The IPD layer cannot be too thin. When the thickness of the IPD is less than 10 nm, the charge of the floating gate will decrease with time. At present, manufacturers have upgraded the flash cell process to 16nm~19nm. The control gate originally made of silicon has been replaced by metal and silicon oxide, and IPD has been replaced by high dielectric material, such as oxidation. hafnium. Therefore, it is possible to construct a more effective control gate, and it is no longer necessary to surround the floating gate from three sides. Instead, the space between the control gates of the cells is filled with bubbles to reduce interference. However, these measures can not solve the problem completely to solve the problem of IPD, and can only play a certain delay. The limitation of 3D flash 2D flash memory that breaks through the limitation can be solved by different methods. For example, the structure of the flash memory unit has only one layer adjacent to each other, and the 3D structure that can be stacked is used. The more layers are stacked, the more layers are stacked. The higher the storage density. Samsung first proposed the concept of 3D flash memory in 2013. The first generation of V-Nand contains 24 layers of flash memory cells, thus achieving the same storage density of the current best 2D flash memory. Although the superimposed cells require a space of 80 nm in diameter, the large stacked structure still has its advantages. V-Nand flash memory operates at low voltages and supports 35,000 write or delete cycles, more than three times the MLC unit. And according to Samsung's introduction, the writing speed will also increase exponentially. Samsung has a large production plan, and it is expected that the V-Nand chip will increase from 128GB to 1TB in 2018. In fact, this only needs to stack more layers, and the 192 layer can reach 1TB. Based on this data density, the cost of SSDs between 4TB and 8TB will be affordable for individual users. The industry believes that as long as 3D flash memory is stacked over 40 layers, the price per GB will be cheaper than 2D flash. Technically, Toshiba's Bit Cost Scalable (BICS) flash memory is almost the same as V-Nand flash. BICS uses a memory layer made of silicon nitride, which surrounds the charge in the memory layer by two layers of oxide. The control gates adjacent to the cell can remove charge from the memory layer or can be written. V-NAND flash memory works the same way, but the storage layer uses other materials, and its storage layer is not wrapped by silicon-based oxide, but wrapped by high dielectric material made of aluminum. According to Samsung's introduction, the control grid uses tantalum nitride to ensure faster removal of the unit, and the new material also ensures the charge level, which further increases the service life. At present, Toshiba is in the process of rebuilding the old chip production equipment manufacturing BICS, the sample can be launched as early as March 2015. Although the current BICS prototype only plans to stack 16 layers, it can be expected that more than 30 layers of officially sold chips will be enough to compete with V-NAND flash. IHS market experts believe that 3D flash memory will be everywhere in 2016, data density will be able to compete with mechanical hard drives, and if the price is loose, then mechanical hard drives are likely to become history soon. This article comes from [System Home] www.xp85.com