back to article Inside Intel's Optanical garden

Intel has now had its brace of Optane P4800X SSD and M.2 2280 motherboard card releases and we’ve learned there is no straightforward performance comparison with equivalent flash product – Intel is eschewing that on the grounds of Optane being a significantly different storage product. We can agree with that, not having enough …

  1. future research

    Not going to succeed.

    Jim Handy writes that the expectation for Optane SSDs was that “the new memory should have performance close to that of DRAM and a price close to that of NAND. Instead, what Intel has announced is an SSD whose performance is close to that of NAND flash at a price that is close to that of DRAM.”

    This, added to the fact that it is only avalible in 16GB and 32GB sizes, is why I can't see optane being used. If it had a speed between SSD and RAM, then some poeple could justify a cost between the two. As it stands this is a dead duck. A samsung NVMe SSD is already limited by the speed of the four PCIe lanes it has the Optane disk only has 2 Lanes so I can't see how it can match performance, and is avalible up to 2TB.

    When I first came acrros the Optane term I did a google to find out what it was, but could only find marketing misinformation about it, so was sceptical, now I know it is useless.

    1. Geoff Campbell
      Boffin

      Re: Not going to succeed.

      It's early days yet.

      The most interesting thing about XPoint for me, is the evolution of machines with a few TB of memory-bus near-DRAM-speed storage that persists through power cycles. That should open up a whole bunch of new ways of structuring data on the local machines, when the software catches up with the possibilities.

      We'll see. For now, a faster SSD that's waiting on a faster interface seems OK to me.

      GJC

      1. Anonymous Coward
        Anonymous Coward

        Re: Not going to succeed.

        The most interesting thing about XPoint for me, is the evolution of machines with a few TB of memory-bus near-DRAM-speed storage that persists through power cycles. That should open up a whole bunch of new ways of structuring data on the local machines, when the software catches up with the possibilities.

        Alternatively, it heralds a new era when the stock IT method of fixing problems, namely "turn it off and on again", no longer works.

        1. Geoff Campbell
          Pirate

          Re: Not going to succeed.

          Good, that'll sort the men from the boys. Or the women from the girls, of course.

          GJC

  2. Joerg

    NAND SSD are crap! Stop with the lies against 3D XPoint !

    NAND SSD are crap! Stop with the lies against 3D XPoint !

    And I mean seriously.

    Is that just pure ignorance or is someone pushing these articles against Intel/Micron on purpose?

    Because going around telling that NAND SLC,MLC,TLC,QLC would have no endurance issues and would even last longer than 3D XPoint is just beyond pathetic!

    All NAND flash technologies are completely unreliable with a very low endurance which is a lot lower than what the claimed statistics show in the real world.

    1. Lennart Sorensen

      Re: NAND SSD are crap! Stop with the lies against 3D XPoint !

      Based on the endurance numbers intel is providing, XPoint isn't looking a whole lot better than NAND, which is rather contrary to the original claims about the technology.

      So either intel's specs are wrong, or they are having early production problems causing endurance problems.

      Certainly 30 complete writes per day is not a lot more than the 17 complete writes per day that the intel NAND SSD is rated for, especially if it is for 3 years versus 5 years. Certainly based on intel's hype I was expecting 1000 complete writes per day instead.

      The latency is rather impressive though.

      1. Joerg

        Re: NAND SSD are crap! Stop with the lies against 3D XPoint !

        Never trust anything about NAND endurance. Ever. By any manufacturer.

        If you read the SSD specs with the NAND inside it all seems good and shiny.. then go and find the tech pdf specs of the NAND flash chips and everything changes. There the stats lies are far less and the truth gets out. NAND is completely unreliable. Even the most expensive Enterprise class SLC chips.

        1. Anonymous Coward
          Anonymous Coward

          Re: NAND SSD are crap! Stop with the lies against 3D XPoint !

          ...and then you go read long term tests of SSDs and discover the specs massively understate actual tested endurance, if they survive the 1st week you'll bin the SSD for a bigger replacement before it dies.

    2. Throatwarbler Mangrove Silver badge
      WTF?

      Re: NAND SSD are crap! Stop with the lies against 3D XPoint !

      Joerg,

      I am not sure where your recurrent vitriol comes from. I can say for a fact that I've run NAND SSDs in datacenter storage arrays for close to a decade without a single failure, and I've run them in laptops, desktops, and servers with the same success. Whatever your complaints about NAND, they seems as baseless as EMC's usual scare tactics against competing storage vendors, although EMC usually has some form of research to back up their messages, whereas you have . . . nothing at all, as far as I can tell.

  3. Mikel

    Not as billed

    Disappointing.

    You know they're off their mark when they release a brand new storage and compare its performance to spinning rust. Why not tape?

  4. DougS Silver badge

    Datera's storage tiers

    They show how useless it is by the way it fits in their tiers:

    1. Intel Optane with >500,000 IOPS and <50 microseconds latency

    2. Flash with 100,000 IOPS and <500 microseconds latency

    3. Disk drives with 150 IOPS and multi-millisecond latencies

    Why have two tiers that are so close to each other? It is like when array makers tried to push two hard drive tiers on us after they added flash, where the difference between 15K SAS/FC and big slow SATA was so small. Even when you adjust it for IOPS per GB it was only 10x - simply not enough to matter when flash was several orders of magnitude faster.

    Optane is in the same boat, being at best 10x faster (but because of its low capacity, very similar on IOPS per GB) so it doesn't make sense as a third tier.

    Storage is still a two tier world, hard drives for cheap bulk storage, and flash for fast / high IOPS storage. Eventually they'll be able to make cheap flash (QLC perhaps) with pretty poor write lifetime but when it gets close enough to hard drives in $/GB that will become the bottom tier. It won't be the bottom tier because of speed (though it may be slower also) but because of the low write lifetime. It will be where you put your cold data that is written rarely and may or may not be read often.

    I think Datera is trying to hype Optane because 1) they were dumb enough to design it into their products and 2) as a startup they need all the press they can get, so they've hitched their wagon to Intel to try to grab a bit of the reflected attention on Optane.

  5. Anonymous Coward
    Anonymous Coward

    DRAM Cost vs 3D Xpoint

    Quoting how much HP charges for a DRAM upgrade is not the an accurate assessment of cost. HP pays less than $8/GB for Server class DRAM memory. They just like to make sure their customers feel good by paying a huge mark up. ... it is the HP way.

    we will see how the real performance is when we get independent benchmark data vs data that Intel told people they had to publish. When can I buy the P4800x online? next week? 2 weeks?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019