Optane Memory – Has Intel Created a Universal Memory Technology?
On a typical day I will read all manner of articles and press releases. I will comment on them and post these to the client side of ned. This is part of ned’s subscription service. It helps ned’s clients sort through the stream of news.
I am writing here in response to an IEEE Spectrum article about Intel’s recently released Optane memory products. In short, Intel’s Optane products are based on 3D XPoint memory, which Intel developed with Micron. Optane is currently marketed as an intermediary between storage and memory. It sounds arcane but it will be a big deal if Optane can someday become the Universal Memory i.e. replace storage and DRAM. As I have written before, a Universal Memory will be difficult to realize, but if it is realized it will be very valuable.
where am I going?
I want to focus on one point. In the sixth paragraph the IEEE article reads “Intel … won’t say what the technology really is, but this doesn’t seem to bother researchers or analysts.” It bothers me. I want to know what it is.
At the heart of any memory are unit cells i.e. where data is stored. Generally, there is one bit per unit cell. Digging deeper, each unit cell has a storage element. In current memories the storage element is silicon-based. We know that the 3D XPoint storage element is based on a new material i.e. it is not silicon-based. There are many in the community that believe 3D XPoint is based on PCRAM. I have written about PCRAM at length over the years (here and here, for example) and was not surprised by its early demise. If Optane is actually based on a PCRAM storage element it would certainly be different from the one that is no longer with us. Some level of research in PCRAM would have continued over the years. This, in turn, may have lead to an evolution in the technology and it became viable. It will still be “phase change” but it would be a new material or structure.
From a marketing standpoint Intel may be interested in leaving PCRAM behind them. The first iteration failed and people may not differentiate a second iteration if the name is the same. Intel has lofty goals for Optane. This is quite likely why they wanted to rebrand. Get away from the past and move on towards a Universal Memory. The IEEE article provides many quotes from Intel execs. Unfortunately, it appears to take them at face value. Does the author believe the hype? Should we believe Intel? Can we form a rational opinion on our own? I think we can.
We can form a rational opinion on our own. For this, we need information about the storage element. With knowledge of the material and structure of the storage element we can look at the research and patents to assess the current performance. From here we can look forward and opine, based on current evidence and understanding, on whether their lofty Universal Memory goals are achievable. If we do not go after this information we rely on the hype coming from Intel’s marketing department, making it difficult to look forward.
Institutional investors want to know ahead of time if a product’s goals are achievable i.e. is the target market achievable. It is when I have independent knowledge of the structure that I can look forward. Luckily, there are others out there that want to know the same details around the unit cell’s structure. I have no doubt the RE houses are cross-sectioning a 3D XPoint memory array and determining the materials and structure of the storage element right now. Once I know the details of the memory cell and its storage element I can look forward for ned’s clients.