Intel is no longer producing Optane 3D XPoint, but the storage class memory product is leaving behind a legacy that could continue to influence products yet to come to market.
Optane's demise came with a whimper during Intel's July 22 earnings call, just seven years after coming to market. Intel developed the SCM product, a storage medium with non-volatile memory expansion, to fill a gap between the expensive, high-performance DRAM and the more affordable but lower-performant flash memory.
Optane, which can also be used as a storage accelerator to relieve performance issues between flash and memory, as well as expand and persist memory, failed to make a dent in the market. Jeff Janukowicz, analyst at IDC, said Intel couldn't make Optane profitable enough to continue operations.
"For Optane to be successful, it needed to bring the right performance, have the right economics and an ecosystem to take advantage of the technology," Janukowicz said.
Intel made progress in all these areas, but the current environment and Intel's changing strategies made the success of Optane too difficult, he said. The result was that only a handful of customers found strong use cases for Optane.
A small crowd with a large need
Optane had a significant effect on the market and helped define some storage vendors like Vast Data and MemVerge.
Vast Data is an all-flash scale-out file and object data storage vendor. Vast uses quad-level cell (QLC) storage for its primary storage, bringing higher levels of density at a lower cost. However, QLC doesn't have the level of endurance that triple-level cell (TLC) NAND offers, so Vast used Optane SSDs as a write buffer.
Optane isn't the only type of storage that provides high endurance. This can also be done with single-level cell (SLC) NAND, which also provides high performance and endurance compared with multi-level cell (MLC) or TLC NAND, according to Marc Staimer, president of Dragon Slayer Consulting.
Vast Data has switched over to Kioxia's FL6, an SLC-based SSD that still acts as a write buffer for QLC endurance, for future builds of its Universal Storage.
"The difference between the Optane SSD and an SLC SSD was not that big," Staimer said. "The biggest difference was in the PMem and the typical SSD."
PMem, or persistent memory module, became significant to MemVerge and Oracle Exadata. Both focus on in-memory processing, where a larger memory pool is an advantage. The Intel Optane DC PMem was the first PMem product to market, used in DIMMs that sit with DRAM on the memory bus near the CPU. While not as performant as DRAM, Intel's PMem was higher in capacity and lower in cost, and acted similar to DRAM, allowing for much larger memory pools, Staimer said.
What went wrong
Optane SSDs and PMem gave a select few vendors an advantage, but not one that relied on the product.
Joseph Unsworth, research vice president at Gartner, concluded that Intel attempted to address storage and memory bottlenecks through high-performance caching and persistent memory, but struggled with adoption. The lack of profitability, drain on resources and turbulent road -- such as losing Micron as a partner and supplier -- weighed heavy on invested partners, he said.
"OEM partners' reticence to adopt the technology and inability to clearly articulate the value proposition limited its appeal," Unsworth wrote in the report "Intel Concludes Its Memory Journey With Optane Business Exit."
While Intel struggled with several issues around profitability and adoption, use cases were still lacking.
Intel was attempting a Sisyphean feat in some ways with Optane, according to Don Jeanette, vice president of Trendfocus. Vendors keep costs in mind as it relates to performance that fits all their customers' needs.
"Intel was trying to tackle a market that is either not willing to pay or does not need that technology," Jeanette said.
The storage and memory market is full of products that are mature and have good price points. Optane brought better performance, but came from a single supplier, bringing up the question of "supplier B," Jeanette said. If a customer is using Optane and it becomes integral, what happens when Intel can no longer supply the technology?
Looking solely at Optane's performance or where it sat in the storage hierarchy isn't enough, Jeanette said. While it brought more performance than NAND, it came at a higher cost, and the cost differential wasn't worth it in the end.
"You're already in a very cost-sensitive market," he said. "Pennies make a difference."
Costs don't just affect end users. Intel managed its Non-volatile Memory Solutions Group before selling it to SK Hynix, according to Jim Handy, general director and semiconductor analyst at Objective Analysis. The group saw a loss at a time when its competition saw a profit for most quarters.
The estimated loss was in the billions. According to a forthcoming "Emerging Memories" report from Objective Analysis and Coughlin Associates, Intel was estimated to have lost nearly $7 billion from 2016 to 2020.
Joseph UnsworthResearch vice president, Gartner
What was learned
Optane was hard-pressed to be a success from its inception, Jeanette said. The market has to need a new technology for it to be successful. The older technology should be at a point where it either can't keep up or doesn't have the features needed for certain use cases. Instead, Intel introduced a new product and then sought out problems the technology could solve.
For Unsworth, Optane came out with promise, but took too long to become established, which will leave a mark long after Optane's exit.
"Intel's failure to scale the technology profitably will raise the bar even higher for future emerging memory technologies," he wrote in the Gartner report.
While Optane didn't solve a direct need, it did offer a way to lower costs in terms of total memory, and others have already begun to advance that mission. Several alternatives to Optane, and potentially alternatives to RAM and NAND, have been in development for some time with commercially available SCM in the form of Kioxia's FL6. FL6 brings higher performance than NAND with better economics than Optane, IDC's Janukowicz said.
Other SCM technologies in development now include resistive RAM, magnetoresistive RAM and spin-transfer torque MRAM, to name a few, according to Staimer. "The only storage class memory that made a dent was Optane," he said.
The in-development technologies are faster than Optane and are closer to DRAM in terms of performance, but they don't yet have the production yields to make them effective, Staimer said.
Regardless, the current market might not require an alternative now, Jeanette said. A roadmap exists for improving on the metrics of existing technologies like NAND and DRAM.
"If it is not broke, don't fix it," Jeanette said.
Adam Armstrong is a TechTarget news writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.com.