Slashdot Mirror
Micron Plans To Buy Out Intel's Stake In Flash Memory Joint Venture For $1.5 Billion (thestreet.com)
Micron is planning to exercise a $1.5 billion option to buy Intel's 49% stake in the companies' IM Flash Technologies Joint Venture. "The option is exercisable on Jan. 1, 2019, and Micron says the deal will close six to 12 months after," reports TheStreet. From the report: In a statement, Intel suggests the timing of the deal's closing is at its discretion for up to a year after the option is exercised, while indicating it long expected Micron's decision. The companies have already made a pair of announcements this year that between them that signal the end of their age-old R&D partnership for developing non-volatile memory technologies. IMFT owns a manufacturing plant (fab) in Lehi, Utah that both produces NAND flash memory and is for now the sole manufacturer of 3D XPoint (pronounced 3D cross-point), a memory technology that Micron and Intel co-developed and announced to much fanfare in mid-2015. Intel, via its Optane product line, has a head-start on Micron in launching 3D XPoint-based products. However, Micron, via its QuantX brand, plans to launch its own 3D XPoint offerings in late 2019, using a second generation version of the technology.
What's so great about 3D XPoint? In a nutshell, it carves out a middle ground between DRAM (very fast, but not dense, relatively expensive and volatile, or unable to retain its data when power is lost) and NAND (cheap, dense and non-volatile, but relatively slow). Though more expensive than NAND -- particularly in these early days -- and not as fast as DRAM, 3D XPoint is much faster than NAND and much cheaper than DRAM, and like NAND is non-volatile. That opens up a lot of potential applications. Games can get a boost from using 3D XPoint solid-state drives (SSDs) for storage rather than conventional NAND SSDs, as could demanding workstation applications. Within data centers -- probably the largest market for the technology over the next few years -- 3D XPoint could improve the performance of demanding AI and high-performance computing (HPC) applications and enable larger deployments of high-speed, in-memory databases than what's possible using DRAM. And in both the PC/workstation and data center markets, 3D XPoint drives can work in tandem with slower types of storage to act as a high-speed cache for important or frequently-accessed data.
What's so great about 3D XPoint? In a nutshell, it carves out a middle ground between DRAM (very fast, but not dense, relatively expensive and volatile, or unable to retain its data when power is lost) and NAND (cheap, dense and non-volatile, but relatively slow). Though more expensive than NAND -- particularly in these early days -- and not as fast as DRAM, 3D XPoint is much faster than NAND and much cheaper than DRAM, and like NAND is non-volatile. That opens up a lot of potential applications. Games can get a boost from using 3D XPoint solid-state drives (SSDs) for storage rather than conventional NAND SSDs, as could demanding workstation applications. Within data centers -- probably the largest market for the technology over the next few years -- 3D XPoint could improve the performance of demanding AI and high-performance computing (HPC) applications and enable larger deployments of high-speed, in-memory databases than what's possible using DRAM. And in both the PC/workstation and data center markets, 3D XPoint drives can work in tandem with slower types of storage to act as a high-speed cache for important or frequently-accessed data.
Waiting for 4D.
for your entire family
I would really love to throw a stick of optane in my Ryzen system. I wonder if that patent falls under part of the buyout.
Even using the SSD form factor, XPoint is so much faster than flash that companies don't use it in benchmarks because it makes all of the current products look laughably slow except for a couple of synthetic metrics. The potential of it is much greater; it can be *byte addressable*, and could be fast enough to justify putting the control in the CPU, or even using the CPU's memory controller. This would require rethinking things like the file system, and the entire software stack from the OS level on. Something like this will eventually win out as SSDs turn into the HDDs of today, and HDDs turn into tape storage (archival which will exist for data centers, where people don't care about the glacial access times).
Micron believe that their solution has enough of an advantage over the competition to be worth 1.5 billion dollars. Samsung's promised competitor meanwhile hasn't made as much as a demo.
That is all Optane actually is, and they are already usable in any system from what I've read online. The issue is some of the software glue for the acceleration/disk caching capabilities. But if you just want to use them as a high speed swapspace or 'ramdrive', they can already do that today, with either an M.2 or a PCIe x4 slot.
Silly Ole me predicts:
- reduced revenues and ever slimmer margins. Clouds don't care about Intel inside.
- ceo (or new one) sez "we need to focus" (and sell off diversifying businesses like this)
- a few years go by, and cash rich Intel - after sets of rolling layoffs, sez "we need to diversify!". Let's buy a services software company (oh, what? We did that already? How'd what work?)
Too bad it wears out from being written. Using it as a substitute RAM would result in burnout in just a few seconds.