SATA, mSATA, M.2, M.2 (NVMe), PCIE SSD interface details
Solid State Drive, commonly known as solid state drive, is a hard disk made from a solid state electronic memory chip array, because Taiwanese English is named after the solid capacitor is called Solid. The SSD is composed of a control unit and a storage unit (FLASH chip, DRAM chip). The specifications and definitions, functions, and usage methods of the SSD are exactly the same as those of the ordinary hard disk. The shape and size of the SSD are also identical to those of the ordinary hard disk. It is widely used in military, vehicle, industrial control, video surveillance, network monitoring, network terminals, power, medical, aviation, navigation equipment and many other fields.
The chip’s operating temperature range is very wide, commercial products (0 ~ 70 ° C) industrial products (-40 ~ 85 ° C). Although the cost is high, it is gradually spreading to the DIY market. Since SSD technology is different from traditional hard disk technology, many emerging memory vendors have emerged. Manufacturers only need to buy NAND memory, and with the appropriate control chip, they can manufacture solid state drives. The new generation of solid state drives generally use SATA-2 interface, SATA-3 interface, SAS interface, MSATA interface, PCI-E interface, NGFF interface, CFast interface, SFF-8639 interface and M.2 NVME/SATA protocol.
The main interfaces of SSDs are:
As the most widely used hard disk interface, the biggest advantage of the SATA 3.0 interface is maturity. Ordinary 2.5-inch SSD and HDD hard disk use this interface, the theoretical transmission bandwidth is 6Gbps, although the bandwidth is much worse than the new interface’s 10Gbps or even 32Gbps bandwidth, but the ordinary 2.5-inch SSD does not have such a high demand, 500MB / s read more Write speed is also sufficient.
m SATA interface
mSATA interface, full name mini version SATA interface (mini-SATA). In the early days, in order to better adapt to the use environment of such ultra-thin devices, the mSATA interface developed for portable devices came into being. Think of it as the mini version of the standard SATA interface, and the physical interface (that is, the interface type) is the same as the mini PCI-E interface.
mSATA interface is an important process of SSD miniaturization, but mSATA still does not get rid of some defects of SATA interface, such as SATA channel, the speed is still 6Gbps. For many reasons, the mSATA interface was not fired, but it was replaced by the M.2 SSD with more upgrade potential.
The M.2 interface is a new interface specification introduced by Intel to replace mSATA, which is the NGFF we have often mentioned before, namely Next Generation Form Factor.
The M.2 interface has a solid-state hard disk width of 22mm, a single-sided thickness of 2.75mm, and a double-sided flash memory layout of 3.85mm thick, but M.2 has a wide range of scalability, up to 110mm, which can increase the SSD capacity. M.2 SSD is similar to mSATA, and it does not have a metal casing. The common specifications are mainly 2242, 2260, and 2280. The width is 22mm and the length is different.
Not only the length, M.2 interface also has two different specifications, namely “socket2” and “socket3”
It seems to be all M.2 interfaces, but the protocols they support are different. The impact on their speed can be said to be very different. The M.2 interface currently supports two channel buses, one is SATA bus and the other is PCI-E bus. Of course, due to the theoretical bandwidth limitation (6Gb/s), the SATA channel can only reach 600MB/s, but the PCI-E channel is different. The bandwidth can reach 10Gb/s, so it seems to be M. 2 interface, but the “dao” is not the same, the speed will naturally be different.
The figure above shows the rate of the SATA channel on the M.2 interface.
The figure above shows the rate at which the M.2 interface takes the PCIE channel.
M.2 interface (NVMe protocol)
NVM Express (NVMe), or Non-Volatile Memory Express, is a logical device interface specification. He is a device-based logical interface-based bus transfer protocol specification (equivalent to the application layer in the communication protocol) similar to AHCI for accessing non-volatile memory media attached via the PCI-Express (PCIe) bus, although in theory The PCIe bus protocol is not required.
The purpose of this specification is to make full use of the low latency and parallelism of PCI-E channels, as well as the parallelism of contemporary processors, platforms and applications, and greatly improve the read and write performance of SSDs under controllable storage costs. Reduce the high latency caused by the AHCI interface and completely liberate the ultimate performance of the SATA era SSD.
Specific benefits of NVMe include:
1 performance has several times improvement;
2 can greatly reduce the delay;
3NVMe can increase the maximum queue depth from 32 to 64000, and the IOPS capability of SSD will be greatly improved;
4 automatic power state switching and dynamic energy management functions greatly reduce power consumption;
The emergence of the 5NVMe standard addresses the issue of driver applicability between different PCIe SSDs.
Speaking of the advantages of the NVMe standard versus the AHCI standard, one of them is low latency. Because the AHCI standard itself is designed for high-latency mechanical hard disks, although SSD has been developed so far, mainstream products have begun to fail to meet the rapid development of performance, especially in terms of delay. The NVMe standard for SSD products, which reduces the high latency that occurs when storing, is one of the problems to be solved.
NVMe SSD can effectively reduce latency ( pictures from the network)
In terms of the software layer, the delay of the NVMe standard is less than half of that of AHCI. NVMe simplifies the calling mode, and does not need to read the register when executing the command; and each command of AHCI needs to read the register 4 times, which consumes 8000 times. The CPU cycles, causing a delay of approximately 2.5 microseconds.
IOPS increased greatly:
Another focus of NVMe is to improve the IOPS (read and write times per second) performance of SSDs. At present, the SATA interface SSD with good performance on the market will only test the IOPS capability with a queue depth of 32. In fact, this is the upper limit of AHCI. In fact, many flash masters can provide better queue depth. NVMe can increase the maximum queue depth from 32 to 64000, and the IOPS capability of SSD will be greatly improved.
Significant increase in queue depth (images from the network)
The advantage of low latency and good parallelism is that the random performance of the SSD can be greatly improved. This is the live running of the 950PRO series SSD. Its random performance is absolutely first-class and can be played at any queue depth. Out of great speed.
Lower power consumption:
More advanced energy management (pictures from the network)
NVMe has added automatic power state switching and dynamic energy management. The device can quickly switch to energy consumption state 1 after being idle for 50ms. After 500ms idle, it will enter state 2 with lower energy consumption. Although switching the power consumption state will cause a short delay, the power consumption in these two states can be controlled at a very low level when idle, so in terms of energy management, there is a big advantage over the mainstream SATA interface SSD. This is especially helpful for increasing the life of mobile devices such as laptops .
The mainstream operating system is gradually starting to support NVMe (pictures from the network)
The emergence of the NVMe standard solves the problem of driver applicability between different PCIe SSDs. NVMe SSD can easily match different platforms and systems, and can work normally without the need for manufacturers to provide corresponding drivers. Currently, Windows , Linux , Solaris, Unix , VMware, UEFI, etc. have added support for NVMe SSD.
In the traditional SATA hard disk, when we perform data operations, the data is first read from the hard disk to the memory, and then the data is extracted to the CPU for calculation, and then written into the memory and stored in the hard disk; PCI-E does not The same, the data is directly connected to the CPU directly through the bus, eliminating the process of memory calling the hard disk, the transmission efficiency and speed are doubled. Simply put, we can understand the two channels as two identical cars, the PCI-E channel car is like driving at high speed, and the SATA channel car is like driving on the rugged mountain road. Obviously, PCI-E SSD transfer speed is much faster than SATA SSD.
At present, PCI-E interface channels are available in PCI-E 2.0 x2 and PCI-E 3.0 x4, and the maximum speed is up to 32 Gbps, which can meet the needs of the future, and the early PCI-E hard disk can not be used as a boot disk. Most flagship SSDs will choose the PCI-E interface.
Although PCI-E SSD has many benefits, it is not suitable for everyone. PCI-E SSDs are more expensive due to flash granules and master quality issues, and are more expensive than traditional SATA SSDs. In addition, because PCI-E will occupy the bus channel, the number of CPU channels for the entry and mid-end platforms is small, and it is not suitable for adding PCI-E SSD. Only the Z170, or the top platform such as X79 and X99, can fully play PCI- E SSD performance. In general, if you are a local tycoon, then PCI-E SSD!