Back in 2015, Micron and Intel announced the development of a new type of memory. Dubbed 3D XPoint (Crosspoint), the new type of memory offered endurance levels NAND flash couldn’t match and theoretically higher performance. Intel marketed its brand as Optane while Micron intended to launch its own QuantX products. Micron never actually shipped hardware to the wider public, and Intel has recently scaled back its own Optane efforts to focus on the enterprise.
Now, Micron is announcing that it intends to leave the business altogether. It also wants to sell the fab in Lehi, Utah where it currently fabs Optane for Intel.
Micron and Intel began dissolving their joint partnership several years ago, culminating in Micron’s decision to buy Intel’s share of the same Lehi fab. Micron leaving the Optane business today doesn’t mean the company won’t continue to produce hardware for Intel, but the memory manufacturer didn’t mince words about why it was leaving the Optane market.
“Micron has now determined that there is insufficient market validation to justify the ongoing high levels of investments required to successfully commercialize 3D XPoint at scale to address the evolving memory and storage needs of its customers.”
Does Optane Have a Future?
Intel’s entire marketing push around Optane revolved around the idea that the memory standard offers features conventional NAND and DRAM can’t deliver, with higher performance and endurance than the former and lower power consumption and higher density than the latter. It has also unveiled some uses for Optane as Persistent Memory that allow for substantial performance improvements in servers if relevant applications are recoded to take advantage of the capabilities Optane brings to the table.
The problem Intel has had is twofold: First, while Optane does offer higher performance in some scenarios and better endurance overall, those gains have not been large enough to drive wider market adoption.
Intel has told the press that Crow Pass, which arrives with Sapphire Rapids, will be the first generation of Optane to offer meaningful speed boosts over its predecessor. With Sapphire Rapids expected in-market next year, this suggests Optane could get a performance boost within 12-18 months, possibly opening up a wider gap between itself and NAND.
One major problem for Intel, however, has been the collapse in NAND flash and DRAM prices over 2018 and 2019:
The original plan for Optane was for first-generation to demonstrate proof of concept, while follow-on products and growing economies of scale improved costs over time. If you look back at where DRAM and NAND prices were in 2015 – 2016, they’re quite a bit higher than today. This has put pressure on Optane and limited Intel’s ability to sell it. Micron’s comments about a smaller-than-expected market speak to this problem.
The entire point of Optane was to create a performance-enhancing bulwark that would lock vendors into Xeon sales (while, yes, also boosting x86 server performance). A number of analysts expect Storage Class Memory — which is to say, persistent memories like Samsung’s Z-NAND and Optane — to grow steadily in the next few years, as shown in the image below:
With Micron moving to exit the business, Intel has to decide where it wants to go from here. Will it support future generations of Optane using the industry-neutral CXL interface under development? Such a move could boost adoption by extending the benefits of Optane to AMD and ARM servers. Intel is likely loathe to make such a move, for precisely this reason.
Much depends on how Crow Pass is shaping up. Launching a genuinely new type of storage technology is incredibly difficult. Many of the problems Intel faces with respect to cost and software support are common to any effort to overturn an entrenched standard. With that said, Intel has sunk six years of development into Optane and might need another 2-3 years of work before that investment really starts paying off.
Speaking strictly for myself, and not on behalf of wfoojjaec: The only reason for Intel to walk away from Optane, in my personal opinion, is if it sees no long-term path that ends in commercial viability. Emphasis on “long-term.” We’re a decade past the point where improving compute efficiency was as simple as waiting for the next process node. The name of the game today is collaboration in optimizing every part of the system.
Optane is not currently a simultaneous replacement for both NAND and DRAM, but it comes closer to capturing that goal than any emerging memory of the past two decades. And Optane bit shipments, while low, are believed to be larger than all other emerging memories combined. Intel is an incredibly healthy company, with excellent balance sheets. Activist investors such as Third Point have made it clear they have their own ideas for how Intel should address its challenges, but last time I checked, Andy Grove didn’t found Intel to enrich activist investors. New CEO Pat Gelsinger was a Grove mentee and a longtime Intel employee during its heyday, so we’ll see what kind of culture changes and shifts in focus he brings to the table.
Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption
Intel's higher-end desktop CPU TDPs no longer communicate anything useful about the CPUs power consumption under load.
AMD’s New Radeon RX 6000 Series Is Optimized to Battle Ampere
AMD unveiled its RX 6000 series today. For the first time since it bought ATI in 2006, there will be some specific advantages to running AMD GPUs in AMD platforms.
Scientists Confirm the Presence of Water on the Moon
Scientists have confirmed the discovery of molecular water on the moon. Is there any of it in a form we can use? That's less clear.
Intel’s Raja Koduri to Present at Samsung Foundry’s Upcoming Conference
Intel's Raja Koduri will speak at a Samsung foundry event this week — and that's not something that would happen if Intel didn't have something to say.