PCIe 4.0 collides with NVMe, SSD once again challenges the speed limit

After decades of development, users have access to more and more types of hard disk products, from mechanical hard disks to SSDs in the early years. In our understanding, the speed of hard disks seems to have been constant. Growth, and the upper limit of the maximum speed we can obtain is constantly rising. The future of speed has no upper limit, but each generation of standards does have its limits. The requirements for speed push manufacturers to constantly explore the limits of hard drives. So what are the factors that restrict the speed of hard drives?

SSD, the full name of Soild State Drive, is also commonly referred to as a solid state drive, and it is one of the most common storage devices that ordinary users can touch. It is widely used in the PC industry. Almost every notebook computer and the whole machine are already equipped with a fixed storage configuration when they leave the factory. It mainly includes the following parts: controller, memory, flash memory particles, etc. These parts perform their duties as a guarantee of the essential speed of SSD.

Image for post
Image for post

From the perspective of their respective functions, the controller mainly provides an external host interface, an internal flash memory management interface, and runs the SSD firmware through the embedded chip, manages the storage address space of the SSD, the physical space of the flash memory, and takes into account garbage collection and wear leveling effect. Flash memory is an entity that stores information. Flash memory particles are distributed on the circuit board of the SSD to provide storage space for the SSD. But in addition, the corresponding protocols and standards are also an important part of determining the performance of SSDs. This article will also discuss the factors that restrict the speed of SSD from four aspects: controller, flash memory, protocol and interface.

Analysis of PCIe 4.0

In terms of definition, what we usually call PCIe is actually the English abbreviation for high-speed interconnection of peripherals, and the full English name is Peripheral Component Interconnect Express. The PCIe standard is mainly used to connect graphics cards, network cards, solid state drives, and various other non-CPU components to a unified interface standard. The PCIe standard is also the fastest link standard we can get on PCs. The speed of the PCIe interface determines the data exchange speed between all these external components and the CPU.

Image for post
Image for post

From the appearance of the PCIe standard to the present, the latest standard has come to the 4.0 period. Theoretically speaking, the maximum bandwidth (rate) of the PCIe 4.0×16 interface is 32GB/s, but the SSD does not use this interface, and the bandwidth of the PCIe 4.0×4 interface that the SSD can access can also reach 8GB/s.

Image for post
Image for post

In spite of the bad news by many media before, PCIe 4.0 has withstood the pressure. The reactions of the market and manufacturers have proved that PCIe 4.0 will not be abandoned, even if the industry giant Intel responds flatly to this, and looks extremely reluctant. Earlier, there was news that Intel seemed to “skip PCIe 4.0 and enter PCIe 5.0 directly.” From the current situation, PCIe 4.0 is still the only way for everyone to go to PCIe 5.0. In the recent period, devices using PCIe 4.0 have suddenly increased, including the newly released NVIDIA RTX 30 series graphics cards that have been out of stock.

NVMe: A good partner for PCIe interface

The full name of NVMe is Non-Volatile Memory express, which is a non-volatile memory standard, which is a protocol specifically designed for flash storage based on the M.2 interface. In fact, it has been eight years since the birth of the NVMe standard, and it has been nearly two or three years for consumer-grade NVMe products to be widely available.

In 2011, the NVMe standard was officially released. In May 2017, NVMe was upgraded to version 1.3, and the SSDs currently on the market that meet the NVMe 1.3 standard can be said to be basically high-end SSDs with the strongest performance. Solid state drives that support the 1.3 version standard have all-round improvements in write life and read and write speeds, and power consumption has also been better controlled, which is more in line with high-end players’ pursuit of hard disk life and speed.

Image for post
Image for post

As a specification for using PCIe channel SSDs, SSDs supporting the NVMe protocol have the advantages of low latency, fast speed, low power consumption, and good compatibility and parallelism. The advantage of the NVMe protocol is that the native PCIe channel and the CPU are directly connected to avoid the delay caused by the communication between the external controller and the CPU, and at the same time greatly improve the IPOS performance of the SSD.

Controller: easy to find and hard to find

Users of commonly used SSDs must have a deep understanding of how important it is to have a good controller. The main control, as the name suggests, is the hardware that controls the entire solid state. The firmware is the software stored in the main control. The main control is responsible for running this solidified software, issuing firmware algorithm operation requests to actually read, write and delete data, perform garbage collection and Tasks such as wear leveling algorithms to ensure the speed and cleanliness of the SSD.

Image for post
Image for post

According to the operating principle of the SSD, the controller connects the storage unit to the computer in the SSD, and the main controller can operate multiple flash memory particles in parallel through several channels to increase the bandwidth of the bottom layer. This means that a master that can process more queues in parallel at the same time will enable SSDs to have faster speeds. Take Samsung’s just released 980 PRO as an example. It is equipped with Samsung’s own brand new Elpis master. The official said that this master can handle 128 I/O queues at the same time, which is four times higher than the previous Phoenix master of the 970 EVO Plus.

Image for post
Image for post

A queue can contain 64,000 command sets, which means that a total of 128 queues can handle more than 8 million commands. In order to meet the needs of modern users for high performance without compromising power efficiency, it is built on an extremely fine 8nm process and Samsung uses sufficient materials. Many SSDs that use the same quality particles are often lost because they do not have sufficiently powerful controllers.

Flash memory: the basis for carrying data

The NAND flash memory particles used in solid state drives mainly include SLC, MLC, TLC, QLC, and 3D NAND (multi-layer storage). As an emerging type of flash memory, 3D NAND has stronger storage performance. It can stack more memory cells per unit area, which has an advantage in reducing the cost per bit.

Judging from the current market response, many consumers talk about “TLC” discoloration. In fact, limited by various factors, there will be more products using TLC or even QLC particles in the future. You must know that Samsung was still using MLC particles for SSDs until last year, when the industry’s TLC particles have already been rolled out. Compared with SLC and MLC, most of the criticized aspects of TLC focus on its greatly shortened reading and writing life. According to Samsung’s statistics on NVMe SSD users, 99% of users write less than 156 TBW, and 99.7% of users write less than 600 TB, which are lower than the current guaranteed durability standard of 1TB Samsung 3bit MLC SSD. The new V-NAND technology and the advancement of error correcting code (ECC) technology also make this type of NAND more durable.

Image for post
Image for post

The data transfer speed of Samsung’s sixth-generation V-NAND is faster than the previous generation. The unique “channel hole etching” technology enables the new V-NAND to realize a conductive wafer stack consisting of more than 100 layers, which increases the number of cells by about 40% compared with the previous generation 9x layer single stack structure. Similarly, the listing of more and more TLC particle solid state drives also means that MLC particles are gradually bidding farewell to the consumer-grade solid-state drive market, which is precisely the development trend of consumer-grade high-performance solid-state drives.

Controller, flash memory, protocol and interface, these four elements are the four most important factors that restrict the performance of an SSD. Consumers must also pay attention to the parameters related to these aspects when purchasing SSD products. At the same time, it must be clear that in the category of storage devices, the big names are indeed more trustworthy. Because they have their own main control chips, original factory particles and other advantages, they are more assured in terms of product quality. Consumers need to think more about the product’s main control, flash memory particles, and supported protocols and interfaces. Of course, it is also important to update the corresponding supporting facilities. Only when the protocol and interface match, can the read and write speed be maximized.

Written by

Digital Nomad

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store