Toshiba fills out its NVMe SSD offerings with the 2.5″ XD5 series

Hey, remember way back in March of last year when we told you about Toshiba's then-upcoming datacenter SSD offerings? The last of those, the 2.5" XD5, is finally making its way to market. These drives use the same 64-layer BiCS TLC flash as the company's fancier CD5 and CM5 series, as well as the already released XD5 M.2 drives.

Toshiba emphasizes that the XD5s are optimized for low-latency operation and consistent performance, but given their specifications, we think it sounds more like "low-cost." Toshiba says the drives offer up to 2.7 GB/sec on sequential reads and 895 MB/sec on sequential writes. Random performance is even more slanted toward reads—250,000 IOPS versus just 21,000 IOPS on random writes.

The 2.5" XD5 SSDs use NVMe and hook up to U.2. The 2.5", 7-mm height U.2 drives come in sizes ranging from 1TB to 4TB, while the M.2-22110 drives come in just 2TB and 4TB flavors. Toshiba specs the drives for one full drive write per day and says that they should do their thing with no more than 7 W of power. Like the M.2 drives, the 2.5" XD5s come with a five-year warranty.

Given their intended market, you probably won't see these drives showing up on Newegg anytime soon, but they'll show up in various vendor catalogs sometime in Q2. We reckon we're likely to see a consumer version sooner or later, too.

Comments closed
    • Chrispy_
    • 8 months ago

    Gah! “up to” peak performance figures should be banned for enterprise products.

    If Toshiba are emphasizing the consistent performance, what *IS* that consistent performance?

    I don’t pay enterprise SSD prices to have hardware sit there idle for occasional peak performance; Whatever server it’s going into is getting that hardware because it NEEDS those IOPS, and it needs them all the time over a 10+ hour period a day.

    If I want bursty drives with inconsisten performance after the cache is depleted, I’ll just buy a cheaper consumer drive at 1/16th the cost, thanks.

      • nico1982
      • 8 months ago

      An enterprise professional looking for enterprise hardware for his/her enterprise should base his/her choices on something more than a meager press release anyway, don’t you think?

        • Waco
        • 8 months ago

        In general, the spec sheets for enterprise gear aren’t a whole lot better than their consumer counterparts. While they [i<]tend[/i<] to be more consistent under constant load, that's not a hard and fast rule. Sustained performance is rarely detailed.

          • Chrispy_
          • 8 months ago

          This.

          I have to usually find the equivalent consumer product/controller/NAND type and then find the steady-state performance from a consumer review site doing enterprise-level stress-testing and just make the assumption that the enterprise version may perform that poorly in the real world.

          Marketing BS is 10x worse in the enterprise world because on top of all the usual lies they’re pandering to PHB’s and using meaningless acronyms and buzzwords whilst citing improved ROI and uptime. They can’t possibly say that without knowing the cost of the system(s) it’s going into or their purpose!

          • nico1982
          • 8 months ago

          Spec sheets helps, but actual testing helps more. Who deploy hardware or software without testing it in a development/test environment first?

            • Waco
            • 8 months ago

            Nobody, hopefully, but spec sheets that lie (by omission or ignorance) waste a LOT of time.

            • nico1982
            • 8 months ago

            I totally agree with you on that.

            • Chrispy_
            • 8 months ago

            How do I know what to buy before testing?

            Are you honestly suggesting that I buy the hardware first, test it myself, and then send it back – all because the spec sheet lies by omission?

            That’s truly daft.

      • HuskerTX
      • 8 months ago

      ‘Up to’ in this context does not mean FOB, or varying performance on a single drive based on caching. In this context, it means that, as a family, the performance is ‘up to’ that number, based on capacity. Those numbers are 100% steady-state. It states that in their product brief and I’m sure it says so in more detail if you get the NDA datasheet.

      [url<]https://business.toshiba-memory.com/en-us/product/storage-products/data-center-ssd/xd5.html[/url<] This is the M.2 version and it clearly calls out sustained.

        • Chrispy_
        • 8 months ago

        I wish I could believe you, but my cynicism is born of ther lies being constantly called out in practice, resulting in me having to upspec/compensate/workaround.

        [quote<]Read and write speed may vary depending on the host devices, read and write conditions, and file size.[/quote<] I take that as a peak speed, simply because the table does differentiate between the different capacities and yet they still use the wording "up to" with the additional caveats quoted above. If that was the minimum guaranteed transfer rate, that is what they would write, not this weasel-wording table of lies.

          • Waco
          • 8 months ago

          That set of weasel-words shows up in anything stating performance. Nobody wants to have their marketing/sales department taken to court because the spec sheet didn’t have such disclaimers and someone used them in a way that caused them to not perform to spec.

Pin It on Pinterest

Share This