SSD SSDs are mainly composed of control units, storage units and interfaces. Among them, the control unit, which is what we call the main control chip, is the core device in the solid state drive, and is even called the CPU in the solid state drive, which assumes the role of command, operation and collaboration.
For SSDs, the quality of a master control chip directly determines the actual use experience and life of the SSD, and different architectures, the number of cores/transistors, and the level of frequency are related to the performance of the master.
Whether it is ordinary consumers or industry insiders, it is necessary to have a clearer understanding of the master control.
In recent years, domestic main control manufacturers and products have continued to emerge, and the main control products of some manufacturers have also been more mature and widely used. Under the trend and trend of domestic rise in all walks of life, more and more SSD brand owners use domestic control. So, what are the characteristics of the products of each domestic main control manufacturer? What problems need to be paid attention to in the development and production of SSD products using these domestic master control products?
Shandejun specially consulted industry professionals to sort out the text in a targeted manner to help everyone deepen their understanding of domestic master control products. Before analyzing the characteristics of each domestic master control, let's first briefly understand what the main role of SSD master control is, and its main classification.
What is the SSD master
The SSD master is essentially a processor, mainly based on the ARM architecture, and some master manufacturers use the RISC architecture. At present, the computing power of the SSD master chip is determined by the manufacturing process, the number of transistors, the number of cores, and the frequency.
The specific role of the SSD master is manifested in: first, reasonable allocation of data load on each flash memory chip, so that all flash memory particles can work normally under a certain load, coordination and maintenance of the collaboration of different block particles; second, it undertakes the entire data transfer, connects the flash memory chip and the external SATA interface; the third is responsible for the completion of various instructions of the SSD, such as Trim, CG recovery and wear balance.
On the one hand, the SSD master must communicate and collaborate with the host, receive and process the instructions sent by the host, and on the other hand, it must also deal with the flash memory particles that lack intelligence, and do a good job in the specific implementation of the underlying data access. The degree of intelligence of flash memory particles is very low, and can only follow specific flash interfaces, such as Toggle or ONFI for access, while different flash memory chips have great differences in working characteristics, which requires the master to actively adapt to the characteristics of flash memory.
We often see that the SSD standard "supports LDPC error correction", which actually contains two parts: hard judgment and soft judgment, the former is accelerated by hardware in the master control hardware, and the latter needs to be combined with the computing power of the master to strengthen the error correction effect. Because flash memory particles do not only make mistakes at the end of their life, the master error correction engine is actually always in operation, and every data written and read out must be tested and processed by the master error correction engine. Moreover, with the deterioration of the quality of the stored particles, the corresponding requirements for the SSD master control are also getting higher and higher.
SSD Master Category
At present, the master control of SSDs is mainly divided into two camps: original factory and main control manufacturer.
Original factory forms, such as Samsung, SanDisk, Armored Man, Intel or Micron's main controller, they are equipped with their own flash memory particles, and the product quality is naturally first-class. However, in addition to Samsung, many original SSDs are no longer completely using self-developed master control chips, such as some of Intel and Micron's SSD products, which use the main control of Huirong.
That is to say, the major flash memory original factories have delegated the work of master control design and firmware research and development to a special master supplier to a certain extent, and they complete the design of related master control and firmware under the technical support of the original factory. Ultimately, the original SSD product will distinguish the original SSD from the public version of the product through customized firmware.
The main control manufacturers are mostly concentrated in the United States, Taiwan and Chinese mainland. Including Marvell, Huirong, Qunlian, Dot Order, Realtek, etc., and thanks to the state's encouragement and support for chip localization in the past two years, there have been many domestic master control manufacturers, such as Guoke Micro, Deyi Micro, Lianyun Technology, Hualan Micro, Yingren, Yixin, Derui, etc.
Among them, HuiRong is somewhat special, one of its predecessors Silicon Motion was established in Silicon Valley, the United States, after the reorganization and merger in 2002, it became a Taiwan enterprise headquartered in Hsinchu County, but its business layout in the mainland is extensive, with R & D and operation teams in Shenzhen, Shanghai, Beijing and Nanjing, and the business localization is obvious.
What are the characteristics of the main control products used by domestic SSD manufacturers?
Hui Rong
In the HuiRong (SMI) master control used by domestic SSD manufacturers, the products of the SATA protocol are mainly 2258XT and 2259XT, but due to the update and upgrading of the architecture and process, the main push is now 2259XT2. The products of the NVME protocol are mainly 2263XT and 2262EN.
In the serial number of Huirong's master product, it has the "XT" suffix, indicating that it does not support DDR caching, and the "EN" suffix indicates that it supports DDR caching.
Overall, SMI master control has good stability and compatibility, plug disk response time is fast, no matter what kind of particles, new or old particles, good quality or poor, their FW (firmware) can support, and overall customer satisfaction. The FW of the SMI master is relatively more understandable, some new small functions are updated quickly, the ECC settings are also simple, and after the level setting is done according to the requirements, the RDT scans the bad blocks according to the settings.
Due to the wide range of SMI's master control services, compared with the master control of other manufacturers, SMI is less concerned about the support of some inkdie particles.
Get a little
To get a micro-master, the most used SATA protocol master controller is 9082HC and 9082HP. Among them, the 9082HP master control defect rate will be higher than the SMI, but the FW is simpler than the SMI, as long as it can be supported, there is almost no need to do any Diving setting. No need to debug what driver, use a micro-master to do K1 to open the card once, and do not need to open K1 again after running the RTT. When continuing RTT card opening, you only need to set ECC to open the card as needed, for example, first use 20 + 50 to open the card, then FW will open the card according to the good blocks required by 20 + 50, if the defect rate is high, you can change to the 20 + 60 ECC setting to open the card, which greatly saves the RTT time.
For some inkdie particles, the support of a micro relative to SMI is better, and it can be seen that the technical debugging of a micro is still some level.
Especially for problems that arise in production, the engineers of a micro will solve them at the first time. In this regard, it may also be due to the fact that the micro company is in Shenzhen, so the problem will be handled very quickly and well-served. Therefore, for enterprises entering the SSD industry for the first time (especially Shenzhen enterprises), it will be more convenient to adopt the main control of a micro.
Or for reasons of strengthening the control of their own master control, many aspects of the FW of Deyi micro do not support changes, which means that any need to open or increase the function needs to be modified by micro-own people.
In addition, perhaps because the FW architecture is different, a micro-upgrade of FW needs to start with K2, and SMI only needs to do K3 card opening, which means more complicated processes. Moreover, SSD products equipped with a micro-master control are not as fast as SMIs in H2test and HD Tune performance tests.
Realtek
Realtek's SATA protocol master control mainly has 5732, 5733, has slowly transitioned to 1135, NVME protocol master control has 5765, 5766.
Realtek enters the market at a low price, which is relatively cheap. Among them, 5732, 5733 use qFN package, and is 2 rows of interleaved pins, which brings high requirements to the production process, SMT (Surface Mounting Technology, surface mount technology) process is not good enough, will cause a lot of SMT patch problems. This problem, after transitioning to the 1135, a single-row QFN package, is relatively better.
Under Realtek's solution, each product capacity is inconsistent and requires Realtek to debug config files, and its FW is relatively messy.
In addition, the use of Realtek master control also requires a lot of engineering verification, even if the FW is provided, it also needs to be engineering verification, or every time the FW is updated, it also needs engineering verification, and even sometimes, the engineering verification problem is solved, and the other problems are missed.
Realtek FW has a lot of card opening packages, it can be said that each customer uses a different card opening package, Realtek will update FW for each batch of particles sent by customers.
Realtek's master control test also needs to test the current, Realtek product current will be too large, so that including running RDT need to insert the disk first and then power on, if it is inserted in the powered state, it will cause some disks not to run. However, although the current will be large when Realtek first runs the R DT, it is better than SMI in terms of support for HY3DV5 and HY3DV6.
Realtek's main control is a headache, for example, some functions in its FW seem to exist, but in fact can not be realized, or some functions whether set on or off, the default is open.
However, Realtek master control also has another advantage is that it uses a 2K ECC error correction length, which is equivalent to other 1K error correction length settingSLC ECC = 20 and TLC ECC = 50 in the case of setting SLC ECC = 40 and TLC ECC = 100.
In contrast, Realtek's main control will be slower than SMI, and the compatibility is not as good as SMI, so each product must be verified in various ways to ensure that there is no problem in reproduction. Doing realtek solutions is basically very good once smooth, and it needs to be debugged and optimized.
It can be said that the only advantage of Realtek's main control is that the price is cheap and the supply is stable. Therefore, the use of Realtek master control needs to be familiar to do. The reason may be because Realtek is a long-established semiconductor company with deep roots, but it has not long entered the SSD industry and has little experience.
Lianyun
At present, the main controller of SATA protocol has 0901-C, 0902-C, 1102-C, and the main controller of NVMe protocol has 1002-C, 1202-C, 1602-C, all of which use self-developed Agile ECC error correction technology.
At present, lianyun master controller mainly with gooddie particles to do more, support less inkdie. Lianyun's FW is all In English, and the setting is more troublesome. A domestic master control manufacturer, the software developed is only English and no Chinese, which is more puzzling.
In addition, Guoke Micro mainly targets the security encryption market segment and enterprise-level products, which are rarely seen in the open market, and will not be elaborated here.