Seagate and Phison get up close and SSD personal

Seagate is partnering NAND controller supplier Phison to develop enterprise NVMe SSDs.

Primarily a disk drive supplier, Seagate has been diversifying with its Lyve Cloud storage service business and SSD product line. This Phison deal is an indication that it plans to produce better enterprise-class SSDs. Seagate and Phison have collaborated on Seagate’s mainstream SATA SSD products since 2017 and have recently worked on FireCuda consumer gaming NVMe PCIe SSDs and NAS NVMe SSDs. The two companies have entered a long-term partnership to strengthen their development cycle and distribution of enterprise-class SSDs. They say their SSDs will help enterprises reduce total cost ownership (TCO) through increased storage density, lower power consumption, and higher performance.  

Jeff Fochtman, SVP of business and marketing at Seagate, said in an announcement quote: “Seagate is extremely excited to work with Phison on developing advanced SSD technology. Our selective focus allows us to serve the broad performance-driven enterprise SSD market while continuing our leadership in the specialized premium gaming segment.”

Phison CTO Sebastien Jean said: “Our leading in-house ASIC technology, coupled with engineering excellence, complements Seagate’s deep industry knowledge. In a show of commitment to this partnership, in 2020 we opened a development center in Broomfield, Colorado.”

Detailed product announcements from this partnership will be shared in the northern hemisphere summer of 2022.

We asked Jean a set of questions about the Phison/Seagate partnership

Blocks & Files: Will Seagate/Phison develop SSDs supporting PCIe 5?

Sebastien Jean: Yes. We have been developing our PCIe 5 PHY in parallel with CPU vendors and testing at PCI-Sig interoperability events for the past two years. PCI-Sig is the standards organization behind the PCIe protocol. A PHY is the low level physical interface that actually pushes the bits onto the bus. Only a few companies develop their own PHY as high-speed signalling is very complex.

Our first PCIe 5 SSD was announced at CES and will be out later this year. It is designed for high-end gaming, workstation and mid-tier enterprise applications. Our flagship enterprise solution will be announced soon. We have already started work on our PCIe 6 PHY, though we’re a few years before Gen-6 solutions are ready to be productized. 

Will Seagate/Phison develop SSDs using TLC, QLC and then PLC (penta-level cell) flash?

We already have TLC and QLC solutions on the market and are committed to give customers a broad set of options. TLC is now the baseline for SSD storage. QLC has similar read performance as TLC, but write speed is substantially slower and endurance is about half of TLC. QLC works well for read-intensive applications. We expect QLC to improve in the same way that TLC improved after a few years on the market.

The possibility of a PLC SSD is not clear at this time. The factors that make QLC challenging are compounded for PLC. This means PLC would be substantially slower on read and write while offering even less cycling endurance. To put things in perspective, TLC Enterprise SSDs offer one to three Drive Writes Per Day (DWPD). QLC is currently at 0.08 DPWD and PLC write endurance would be even lower. QLC is rapidly improving and the YoY (Year-over-Year) enterprise SSD cost has been dropping logarithmically since 2014. With the arrival of 2Tb QLC NAND die, coupled with the extra board space provided by EDSFF, we can expect to see Enterprise QLC solutions at 256TB in the next few years.  

Will Seagate/Phison develop SSDs in the M.2 format?

We have a wide range of M.2 solutions including boot drives up through high-performance client and enterprise SSDs. We also offer U.3 solutions that address both the high-performance and the high-capacity markets. As we move forward through 2024 and 2026, we can expect the form factors to shift steadily over to EDSFF E1 and E3 due to the technical requirements of supporting PCIe 6 and 7. 

Will Seagate/Phison develop SSDs in the various EDSFF formats, particularly E1.S and E3.l?

We expect to release EDSFF products early next year. The EDSFF standard is great because it is designed from the ground up for high-speed enterprise applications. The connector pins have a bit more space between them and there is additional grounding to help ensure good signal integrity for PCIe 5, 6 and 7. Looking at chassis availability, we can see that E1.S is most appealing to the general market as it is a direct replacement for M.2 22110, but adds better density, improved PCIe signaling and hot-swap capabilities. The high-end enterprise market is focused on E3.S, which came as a bit of a surprise given that E3.L is very close to 3.5″ HDD size.

Though there are some corners of the IT world that doubt the need for PCIe 6 and 7, there are fundamental technology factors that will push the adoption. Businesses like making more money, and doing things faster without substantially increasing opex is an easy decision when paired with the natural upgrade cycle. DRAM is no longer doubling at the same rate, while the CPU core count is still going up. The total MB/core ratio is decreasing, which is the exact opposite of what is needed for machine learning and data mining. CXL is already available on enterprise CPUs which allows for seamless integration of DRAM on the PCIe bus. 

PCIe 5 is roughly equivalent to DDR4, and PCIe 6 lines up with DDR5. PCI-Sig, which is the standards organization behind PCIe, has a clear line-of-sight to Gen-7 and a stated objective of releasing a new standard every one or two years. It is inevitable we will see a rapid adoption of the latest PCIe bus standards on enterprise applications.  

Will Seagate/Phison develop SSDs supporting zoned namespaces (ZNS)?

This is definitely on the horizon. ZNS is a feature that is both technically smart and has significant real-world applications. It provides a simplified and standardized way to allow IT departments to have more direct control over the NAND in the SSDs they purchase. Making NAND work as a storage solution is complex, which is why something like OpenChannel SSD did not catch on. 

Managing NAND is not a core competency for most fortune 500 companies, but ZNS changes that. In a sense it is democratizing NAND so that companies can focus their IT resources where they have the most return on investment. ZNS allows these leading companies to have the control they want in order to optimize their storage, without getting lost in the esoteric minutiae of making a reliable storage solution. I was very encouraged to see that WD and Samsung have agreed to align on their slightly divergent views of ZNS. This will be good for the industry as a whole.