Despite all the rhetoric it looks like Intel’s Xpoint/Optane memory isn’t selling, something that will come as a shock to no one. SemiAccurate has been tracking the price cuts for this technology and thinks Intel is at a dead end.
While we haven’t been fans of the technology because it was oversold and underdelivered, Intel ameliorated most of our concerns when they finally launched the Apache Pass DIMMs last April. The main issue we had was that endurance was not suitable for DIMM write levels over the expected lifespan of a server. At launch Intel claimed that the Xpoint DIMMs were able to be written to at full speed for the five year warranty period without a problem, something SemiAccurate was honestly not expecting them to say. If true almost all of our concerns about the technology just vanished.
The only problem remaining is the hype, something that Intel is not slowing down on. If you listen to the marketing messages, Xpoint is unique, adds value to anything you do, and will bring world peace or something. In reality the use cases look narrow and the product just doesn’t seem to be selling. Why do we say this? Two data points, consumer availability and price cuts.
On the consumer side if you walk in to just about any big box store, almost every Intel desktop had an ‘Optane’ sticker on it. Intel is forcing the silicon into just about every PC they sell likely via MDF that makes it untenable to do otherwise on a system cost level. See Ultrabooks for more on how this practice distorts the industry and hurts consumers. Lets just leave the consumer side at near ubiquity but not voluntarily so on the side of the OEMs.
Moving towards servers, the idea was to expand memory capacity cheaply and easily so customers could do things that were not possible with DRAM. This could be because of power concerns, cost of high density DIMMs, or simply that it is not possible to make or buy production 512GB DIMMs at the moment. In theory the ideas behind Xpoint are sound. Plenty of SemiAccurate’s sources report testing the technology but none seem to be deploying it in volume. Then again anyone who has finds a competitive advantage with Xpoint would not be very likely to shout about it publicly so don’t take our lack of results as definitive.
What we do know is that since the April launch there have been two price cuts on Apache Pass/Xpoint DIMMs. The 128GB, the one we have the best data on, started life around the $577 mark or just over $4.50/GB. Since then this module has had two price cuts of about $100 each ending up in the mid-$300. The last of these cuts happened just over a month ago, call it the first week of June or about two months after the official launch.
Why the cuts? Is it demand or something else? If you listen to the Intel May 8, 2019 investor meeting Q&A, they make it very clear that cratering DRAM prices are not affecting Xpoint prices. To save you a lot of time, the relevant exchange is the very last question in the Q&A. Since Intel stated that the DRAM price drop is not the reason so we will take their word for it. That said you can buy commodity registered 128GB DDR4 sticks for ~$1300 as of this writing so Xpoint pricing of mid-$300s is about the 1/4 DRAM price range Intel has been promising.
So if Xpoint pricing isn’t chasing the cratering DRAM curve, the drop must be lack of demand or something else we are completely unaware of. There is also a small chance that Xpoint could be selling like gangbusters into someplace SemiAccurate has no visibility into but that generally does not lead to declining prices. Intel’s strategy of paying OEMs to stuff Xpoint modules into consumer desktops at major retailers does not point to overwhelming demand from dark sources however. Massive price cuts also are not indicative of very high demand either. That said feel free to ascribe your own reasoning to the situation.
Then there is the most interesting part of the equation, cost, which sheds a very interesting light on Intel’s positioning. Between the current situation and the upcoming second generation Barlow Pass Xpoint DIMMs which SemiAccurate told you about last year, things are getting interesting.
Note: The following is analysis for professional level subscribers only.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- How many PCIe5 lanes does Sapphire Rapids have? - Mar 20, 2020
- What does Intel’s server platform cancellation mean? - Mar 16, 2020
- A few more bits on Intel’s server cancellation - Mar 16, 2020
- Intel kills off a server program - Mar 14, 2020
- When is Samsung’s 2nd gen zNAND due? - Mar 10, 2020