Intel to open Optane memory business

The end seems to be in sight for Intel’s troubled Optane memory business. Tucked away in the company’s brutal Q2’2022 earnings release (more on that a little later today) is a surprising revelation in a section about non-GAAP adjustments. We began phasing out our Intel Optane memory business in Q2 2022. Also, Intel’s earnings report notes that the company is taking a $559 million “optane inventory impairment” charge this quarter.

Taking these items at face value, it appears that Intel is preparing to shut down its Optane memory business and associated 3D XPoint technology. Certainly, there is a lot of disdain around the Optane name and product lines here – which is why we need clarification from Intel – as Intel has many Optane products, including “Optane memory”, “Optane persistent memory” and “Optane SSDs”. Moreover, in Intel’s previous earnings and other financial documents, the entire Optane business unit was commonly referred to as the “Optane memory business,” so it does not appear that Intel is divesting the entire Optane business unit. Optan Memory product only.

Update: 6:40pm ET

Following our inquiry, Intel sent a brief statement on Optane’s wind down. While not giving much in the way of further details on Intel’s exit, Intel does confirm that it is indeed exiting the entire Optane business.

We will continue to rationalize our portfolio to support our IDM 2.0 strategy. This includes evaluating divestment of businesses that are not sufficiently profitable or core to our strategic objectives. After careful consideration, Intel plans to cease future product development in its Optan business. We are committed to supporting Optan customers through the transition.

First unveiled by Intel in 2015, the company’s 3D XPoint memory technology is positioned as a hybrid between DRAM and solid state storage. Unique, small-addressable memory uses phase-change technology to store data instead of trapping electrons like NAND technology. As a result, 3D XPoint offers incredibly high endurance – millions of writes – as well as very high random read and write performance because the data does not have to be organized in relatively large blocks.

Intel, meanwhile, used 3D XPoint as the basis of two product lineups. For its data center customers, 3D XPoint has introduced Optane Persistent Memory, which has been developed into DIMMs as a partial replacement for traditional DRAMs. Optane DIMMs offer a slightly higher density than DRAM, and combined with its permanent non-volatile nature, it can be useful for systems that require large sets of working memory and its non-volatile nature, such as database servers. Meanwhile, Intel has used 3D XPoint as the basis for several storage products, including high-performance SSDs for the server and consumer markets and as a smaller high-speed cache for use with low-end NAND SSDs.

3D XPoint’s unique features have been a challenge for Intel since the technology’s launch. Although designed to be scalable through layer stacking, 3D XPoint manufacturing costs continue to rise above NN per bit, making the technology more expensive than high-performance SSDs. Meanwhile, Optane DIMMs, while filling dedicated space, were just as expensive and offered slower transfer rates than DRAM. So, while Intel strives to offer a product that can straddle the two product areas, for workloads that don’t take advantage of the technology’s unique capabilities, 3D XPoint isn’t as good at their respective functions as DRAM or NAND – making Optane products. Hard sell.

As a result, Intel has been losing money on its Optan business for most (if not all) of its life, including hundreds of millions of dollars in 2020. – The instances where you published those numbers, they were found to be in the red based on operating income. Also, reports from Blocks & Files say Intel is sitting on a high supply of 3D XPoint chips – on the order of two years of inventory from earlier this year. All of this highlights Intel’s problems selling Optane products and increasing write/write costs, which Intel is dealing with today with a $559M Optane impairment charge.

So, wind-down potential for Optane/3D XPoint has been in the tea leaves for some time, and Intel has been taking steps to change or limit the business. Notably, the dissolution of the Intel/Micron IMFT joint venture left Micron with the sole manufacturing plant for 3D XPoint, and Micron abandoned their own 3D XPoint plans. And after producing 3D XPoint memory into 2021, Micron eventually sold the factory to Texas Instruments for other uses. Since then, Intel hasn’t been able to get enough fabs for 3D XPoint – although if the inventory reports are true, they won’t need to produce as much memory after a while.

Meanwhile on the product side of things, the decline in the Optane business follows Intel’s earlier retreat from the client storage market. While the company has released two generations of Optane products for the data center market, it has not released a second generation of consumer products (eg Optane 905P). And, after selling their NAND business to SK Hynix (now operating as Soldigm), Intel no longer manufactures any other type of consumer storage. So while it’s sad to retire the remaining data center products, it’s a logical next step.

Intel’s previous Optane non-volatile memory roadmap: what will never happen

Overall, Intel chose to offload the Optane/3D XPoint business at a critical time for the company. Intel was previously scheduled to launch a related third generation of Optane products this year with their Sapphire Rapids Xeon CPUs. The most important of these was their “crow pass” being 3.rd Continuous generation of DIMMs, which, among other things, updates Optane’s DIMM technology to use the DDR5 interface. While Crow Pass development is complete or close to complete at this point (given Intel’s development schedule and Sapphire Rapids delays), launching and supporting the product still incurs significant up-front and long-term costs and requires Intel. Support the technology for another generation. Giving Intel a strong incentive to finally get out on a money-losing business unit.

Instead of Optan persistent memory, Intel’s official strategy is to turn to CXL memory technology (CXL.mem), which allows both volatile and non-volatile memory to be attached to the CPU on a CXL-capable PCIe bus. This achieves the same goals as Optan (non-volatile memory, large capacity) without the expense of developing an entirely separate memory technology. Sapphire Rapids will in turn be Intel’s first CPU to support CXL, and the overall technology has broad industry support.

AsteraLabs: CXL memory topology

Still, Intel’s retirement of Optane/3D XPoint marks the end of an exciting product lineup. 3D XPoint DIMMs were a new idea, even if they didn’t work, and 3D XPoint was built with ridiculously fast SSDs thanks to the huge random I/O advantage, a feature unlike any other SSD vendor. It can be fully replicated anytime soon. So for a strong in-state storage market, this marks the end of an era.

Source link

Related posts

Leave a Comment

12 − 8 =