At least two retailers from the US on Friday have started to take pre-orders on Intel’s Optane DC Persistent Memory modules and therefore had to reveal their prices. Today they only offered 128 GB and 256 GB modules, possibly indicating that the range-topping 512 GB Optane is set to ship a bit later.

The main purpose of Intel’s Optane DC Persistent Memory modules is to bring higher memory capacities closer to the CPU at a fraction of DRAM cost. Even with recent drops of DDR4 prices, high-capacity RDIMMs and LRDIMMs (i.e., 128 GB or 256 GB) cost thousands of US Dollars. Meanwhile, Intel’s 3D XPoint memory used for Optane-branded products is cheaper to make than DRAM, but offers a high performance and a strong endurance, so in many cases using Optane DC Persistent Memory instead of DDR4 makes sense for memory-intensive applications.

Without any surprises, Intel’s Optane DC modules are priced in accordance with their market positioning: well above regular SSDs, but well below DDR4 RDIMMs/LRDIMMs of the same capacity. Depending on the retailer, the Optane DC 128 GB costs $842 - $893, whereas the Intel Optane DC 256 GB is priced at $2668 - $2850. Keep in mind that these are not official MSRPs of Intel (which will likely keep its RCPs confidential as they will depend on multiple factors), but prices at which certain retailers may sell them.

Pre-Order Prices of Optane DC Persistent Memory Modules
Optane DC Persistent Memory 128 GB $893
$6.97 per GB
$6.57 per GB
Optane DC Persistent Memory 256 GB $2,850
$11.13 per GB
$10.42 per GB
Optane DC Persistent Memory 512 GB ? ?

Unfortunately, neither of the retailers that started to take pre-orders on Intel’s Optane DC Persistent Memory revealed their ETA. Intel’s official position is that actual systems that use the modules will be available in June, so it looks like the chip giant is about to start shipments of its Optane DIMMs to interested parties.

Related Reading:

Source: (1, 2), (1, 2) (via Momomo_US/Twitter)

Comments Locked


View All Comments

  • duploxxx - Monday, April 8, 2019 - link

    add a ram drive, marginal cost.
  • Toadster - Monday, April 8, 2019 - link

    great example of using persistent memory as cache in a 2-tier storage HCI cluster
  • name99 - Sunday, April 28, 2019 - link

    It's not clear quite what that proves. The Optane is used as "cache", which to me, suggests the persistence is not important, all that's important is that a large direct address space is provided. Which in turn suggests that some sort of alternative like I suggested (lots of DRAM behind a controller that re-interpreted the JEDEC commands) would work just as well IF connected to a memory controller that understood this.

    I keep raising thins point because Optane seems like a very striking case of a technology that's being pushed by a company in the face of alternatives that are just as feasible. We see this a LOT in the software space, especially when it comes to standards --- everyone wants THEIR way of solving a problem to be in the spec, but the Huawei solution, the Qualcomm solution, and the Ericsson solution are all basically just as good; it's politics that determines which one goes in.

    We've seen it also at the consumer level, with things like Blu-Ray vs HD-DVD.

    But it's rare that we've seen this at such a fundamental eco-system level. In the past a particular tech choice (DDR-n, or flash, or using a GPU) was so obviously superior to the alternatives that the basic idea was agreed upon by everyone; it was only minor details that differed.
    Optane feels, however, very much more like Beta vs VHS, there ARE tech alternatives to Optane that could be pushed if someone wanted, and the supposed superiorities of Optane to such alternatives have melted away (rather than increasing!) every year as we keep getting more details of the tech.
  • IntelUser2000 - Saturday, April 6, 2019 - link


    It doesn't matter right now. The Optane PMM(Persistent Memory Module) devices can act like RAM, but persistent. You can do lots of cool things with persistent memory that's really fast.

    However, nothing in the consumer space supports it. It just arrived in the server space, so its a start, but how many years away is it for consumers? Then you need Windows that supports it, and more time for applications as well.

    Again, very cool stuff, but nothing you'll see benefits from yet.
  • Diogene7 - Saturday, April 6, 2019 - link

    @IntelUser2000: Thanks for the clarification :). I know that, as of 2019, there is very little support (if any) for the consumer space, which is a quite frustrating for me.

    When I went from using a computer using an HDD to a computer using a SSD, I noticed tremendous improvements in terms of user experience due to lower latency data access.

    On a personal basis, my interest would be to completely replace a NVMe SSD by Storage Class Memory (SCM) on the memory bus to importantly lower data latency.

    At first, I only need that a consumer version of the Operating System (OS) like Microsoft Windows or Apple iOS support to vew SCM on memory bus as a « traditionnal » SSD but as it is on the memory bus, it would have much lower data access latency (hopefully to get at least a 10x improvement).

    Then, at a later stage, sure the applications could be optimized, but as each individual application would need to be optimized by each developper this will take a much longer time.

    From my point of view, it is typically the kind of innovation that smartphone manufacturers should be investing and focusing on, instead as of 2019 of smartphone with flexible displays...

    In 2019, the first smartphones with flexible displays will cost 500€ to 1000€ more than a premium smartphone, and I would be ready to invest 500€ more in a smartphone that would use SCM on memory bus instead to replace UFS Nand flash storage.

    I think it is typically what Apple should be investing in and working on before others as it is something that brings real visible and tangible value to the end user: the sooner, the better !!!
  • IntelUser2000 - Saturday, April 6, 2019 - link

    You can get Optane NVMe SSDs to get 7-10x latency advantage over the best, non-Optane NVMe SSDs, today.

    But you won't get anywhere near 7-10x advantage, because your OS and software isn't built to take advantage of it. In loading, Optane SSDs are maybe 40% best in super heavy games. Most of the time, it'll be 10-20% at best.

    You got big gain from going to SSD from HDD because HDDs were that much slower and software was far away from being a bottleneck.

    Even SSDs are maybe 2-4x fast at loading compared to HDDs despite the latency part being 100x better.

    Another reason you don't get benefit is because for *decades* software has been built to minimize using HDDs.

    Optane as a storage device on the memory bus will get you even less benefit because of that bottleneck. Optane as RAM has potential to revolutionize things but it'll take 3-5 years to see real changes(once its available) and 10-15 years after that to fully take advantage of it, maybe more.
  • IntelUser2000 - Saturday, April 6, 2019 - link

    So to clarify, there isn't "little to no support" for Optane on the memory bus for consumers like you and me. There's nothing. Zero. Zip. Nada. There's "little support" in servers, and that's what the launch enables.

    If you are a subscriber to Anandtech forums I can explain it to you in detail in PM.
  • twtech - Sunday, April 7, 2019 - link

    I'm guessing the basic/simple implementation will be to have the Optane memory show up as a hard drive in Windows. That would be sufficient for our needs for the time being.

    In the future, it should be possible to more or less eliminate the loading step for some data, and access it directly by address from where it resides - that's where the real benefits will come in.

    But for the time being, it sounds like it can be effective as an alternative to running a RAM drive.
  • tuxRoller - Monday, April 8, 2019 - link

    Err, other than Linux? And anything that uses dax filesystems?
    Definitely not optimized but still a good deal better than anything else that's remotely consumer oriented.
  • ilt24 - Saturday, April 6, 2019 - link

    @Diogene7..."Time to boot full windows"

    What I think would be very interesting would be to see recovery times (from a crash) for several types of servers (DB, VM Host, Web...) with and without the Persistent Memory.

Log in

Don't have an account? Sign up now