Controllers are the game-changing component in NAND flash-based storage systems. While most people are familiar with solid-state drives (SSDs) as an example, the same controller- and system-related myths also apply to USB flash disks, eMMC modules, SD cards, or any other form of managed NAND device.
1. eMMC is slow to boot
eMMC is a one chip solution that includes both the flash controller and the NAND flash. Because it’s easy to integrate, it’s often used as a cheaper alternative, even when condensing the PCB footprint isn’t important However, one of the most common myths is that it’s slow to boot. Recent developments have changed this, and some eMMC solutions can now start their booting process in less than 10 ms. This has therefore enabled eMMC to be used as both data storage and a NOR flash replacement for boot code.
2. eMMC 5.x is always faster than older eMMC 4.x devices.
Often overlooked is the fact that performance doesn’t depend solely on the interface generation of the device. The host interface generation is just as important. For example, the performance guaranteed with eMMC 5.x can’t be delivered if the host only supports eMMC 4.x. The same applies for all other interfaces, including USB and SD. The maximum interface transfer rate can only be achieved if the host supports at least the same generation of the device interface (including speed modes and other capabilities).
3. Managed NAND is only available as eMMC or UFS.
Managed NAND refers to systems encompassing both NAND flash management features and the NAND flash itself in the same chip. There are different options and levels of quality. Some managed NANDs are almost like raw NAND, but with an ECC capability. However, the flash management will have to be performed by an external processor.
Popular, more advanced implementations are eMMC or UFS. These are, however, limited to certain host systems. Advanced custom managed NAND solutions can be built easily and are able to use whatever host interface one is familiar with. Disk-on-board solutions are indeed a possibility and come with a significantly lower total cost of ownership (TCO).
4. LDPC (low-density parity check) is the most powerful ECC (error-correction code).
LDPC has become a buzzword that many associate with the highest-quality ECC. However, there are many types of LDPC implementations, and their solutions are not all equal. Some are optimized for correction quality, some for speed, some for power consumption, and some for cost. In short, don’t be falsely reassured that LDPC is the best and only solution—it’s highly dependent on the actual implementation of LDPC in the product.
5. LDPC guarantees a level of error correction.
While its aim is to guarantee a certain level of quality, LDPC depends heavily on its implementation. For example, with a block error rate below 10-12, it’s not feasible to run simulations for an LDPC implementation, as it would take years. Therefore, the error models are extrapolated. Consequently, the statistical models derived from the extrapolation aren’t accurate. This is a limitation of an LDPC implementation.
6. Your data is always 100% safe.
Flash memory has limitations and its capacity to securely house data erodes over time. The flash controller’s primary goal is to increase the perceived reliability of a flash to increase its lifetime. But some external elements can influence how well the controller can perform its task. To name a couple, Alpha particles and Power Fail Robustness may have a major adverse impact. Since these requirements are handled differently by some flash vendors, it’s invaluable to inquire as to how your flash vendor mitigates these issues.
7. SLC is always more expensive than MLC/TLC.
Of course, if you compare GB/USD, this is certainly true. But if you know your specific application and derive the respective requirements out of those, single-level cell (SLC) might be the better choice when taking into account TBW or TCO. You might be able to run the system with less memory and avoid exchange of the storage device during lifetime.
There’s a difference between industrial- and consumer-grade NAND flash-management requirements.
8. SATA has nothing left to give.
SATA may not be a new technology, as it has been on the market for a while now. But newer state-of-the-art flash controllers have started targeting specific use cases where reliability and longevity for industrial storage solutions are key. Increases in data integrity during power loss, as well as reduced TCOs through a DRAM-less design, are just a few aspects that highlight that the SATA interface has never had so much to give.
9. If you want to design a custom solution, you need to integrate source code into the firmware.
A storage solution doesn’t necessarily have to be just a basic storage system. Through modern application programming interfaces (APIs), it can be much more. In addition, the end customer can have full control of its intellectual property firmware, and be in control of its USP, without disclosing it to the flash controller vendor. This is possible on flash controllers featuring an API development toolkit.
10. CompactFlash (CF) is dead.
CompactFlash may no longer be the interface of choice for modern industrial applications. Nonetheless, it’s still used across a number of industrial applications whose host systems continue to support the mature interface. The main advantage of CF is that it’s a reliable and fast interface (>100 MB/s). New CF controllers even support 3D NAND.
11. There’s a quality difference in NAND flash.
It’s often argued that there’s a quality difference in NAND flash. However, do not be mistaken—while there are different cell densities available (SLC, MLC, TLC) with different temperature grades, the same inherently unreliable flash (P/E cycles) is embedded in consumer as well as industrial applications. At the end of the day, the flash controller determines not only the performance and the reliability, but also the longevity.
Lena Harman is Marketing Coordinator at Hyperstone.
