Intel's Optane Memory Business Outlook Looks Murky (Updated)
Why make memory, if you can capitalize on logic?
Update 3/2/2022 7:20pm PT: Intel reached out and provided Tom's Hardware with the following statement:
"We continue to work closely with customers and partners on memory and storage technologies, including enabling our 3rd generation Optane Products with Sapphire Rapids. In addition we are enabling the ecosystem to be ready for key technologies such as memory tiering over CXL.” — Intel Spokesperson
Notably, the statement doesn't answer many of our outstanding questions, such as where and when Intel will begin producing Optane in volume. Additionally, mentions of Optane memory products have been notably absent in recent communications, such as the company's Investor Day, roadmaps, and other announcements.
Original story:
The lack of any announcements concerning the future of Intel's Optane products does not add confidence about their prospects in particular and 3D XPoint memory in general. But a recent comment made by Intel CEO Pat Gelsinger may shed some light on his view on the memory business. As it turns out, Gelsinger does not see it as a priority, as he does not want to compete against the influx of Chinese memory makers in a few years.
"China is pouring extraordinary capital into their semiconductor business," said the Intel chief in an interview with Stratechery. "They can't be below 10 [because they can't get EUV], and the specialty nodes are hard for them to replicate. Where does it go? It goes into memories or it goes into mature nodes. […] So, if I can get [foundry] DNA from Tower and make sure I keep it alive and healthy and prepare for the complement that I have Intel and specialty, I never want to be in memory, you see I am doing everything I can to exit our memory businesses in that regard."
While the comment was made regarding Intel's acquisition of Tower Semiconductor, it refers to China's retrofitting of logic fabs for 3D NAND memory production (see XMC – YMTC collaboration) and what it could mean for the industry.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Intel recently sold its Non-Volatile Memory Solutions Group to SK Hynix, essentially exiting the 3D NAND-based SSD business. So while Optane-branded 3D XPoint-powered SSDs and persistent memory modules are still available, we do not know anything about their future.
When Intel was founded in 1968, it originally produced memory and introduced its first microprocessor — the Intel 4004 — in 1971. Unfortunately, the company abandoned its memory business in the 1980s after it could not compete against Japanese rivals. However, it has tried to return to the memory business in one form or another over the years.
In the late 1990s, it tried to capitalize on Rambus stock options by exclusively supporting Rambus-designed RDRAM with its first-generation Pentium 4 platforms. Eventually, it became a leading supplier of enterprise-grade and 2D/3D NAND-based SSDs and even co-developed 3D XPoint storage class memory (SCM) with Micron to address markets that required higher performance as well as higher endurance.
But while Intel's storage business was performing relatively well, its 3D XPoint business has been bleeding money. By selling its 3D NAND fab in China; letting Micron sell its 3D XPoint production facility in Lehi, Utah (by not acquiring it itself); and moving 3D XPoint development and production equipment from its fab in New Mexico, Intel is clearly showing that storage and memory businesses are no longer its focus. Companies like Samsung, SK Hynix, Kioxia, and Western Digital can produce more memory, and their volumes of scale almost guarantee that their operations will be profitable in the long term, even when Chinese fabs come online.
But despite the comments made by Intel's CEO and its recent actions, we cannot say that Intel's memory saga is completely over. For example, the company's upcoming Ponte Vecchio compute GPU for supercomputers will use Intel's proprietary Rambo cache. Meanwhile, Intel's hybrid supercomputer processor Falcon Shores with both x86 and Xe cores will use a "new extreme bandwidth shared memory being developed by Intel," allegedly specifically for its high-performance computing (HPC) products.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
jeremyj_83 While Optane memory isn't a big loss, the loss of Optane SSD would be sad in a data center. Using the SSD as a write cache for VMware vSAN is quite nice and faster than regular NAND.Reply -
2Be_or_Not2Be Yeah, I especially liked the P4800X series. They were great as primary storage for virtual hosts. They allowed for higher density of virtual guests as Optane's storage performance still made the guests feel snappy enough. They also made database servers more performant as well. I'll be sorry to see them go, that's for sure.Reply -
JayNor Intel's response hints that Optane memory could move to a cxl memory controller. The leaked slides of Emerald Rapids show a CXL.mem as a feature, so perhaps they'll get the Optane DIMMs off the ddr5 memory controller and into some packages that remove the DIMM power/thermal constraints.Reply