Marvell’s dual-lane SATAe SSD controller coming in Q4

SATA Express ports are popping up on new Intel motherboards, but apart from an Asus prototype, we haven’t seen any drives based on the standard. That could change at Computex, where Marvell plans to show off a new SSD controller designed specifically for the SATAe interface. The company claims this chip will enable SSD makers to build PCIe drives that are price-competitive with their SATA counterparts.

The 88SS1083 controller’s name may be an awkward mess of numbers and letters, but it’s fast, with dual lanes of PCIe Gen2 connectivity that yield 1GB/s of bidirectional bandwidth. This dual-lane link is a perfect match for the dedicated storage lanes built into Intel’s 9 Series chipset. It also supports SRIS, otherwise known as Separate Refclock with Independent SSC (Spread-Spectrum Clocking), which uses a drive-based clock generator instead of passing PCIe synchronization signals over the SATAe cable. SRIS lowers cabling costs for drive makers, and it’s supported by all the Z97 motherboards we’ve tested thus far.

Marvell hasn’t revealed the number of parallel NAND channels in the controller. However, the press release touting the chip indicates that it works with flash technologies as fine as 15 nm. The controller is compatible with the Toggle DDR 2 and ONFI 3 standards, enabling per-die transfer rates up to 400 MT/s. It also supports DevSleep and PCIe’s L1.2 low-power mode.

We’ve asked Marvell to clarify a few details about the controller and will update this story when we get a response. It could be a while before we see retail-ready drives based on the chip, though. Marvell says the controller is sampling to customers now, but mass production isn’t scheduled until the fourth quarter. The first Marvell-powered SATAe SSDs may not arrive until late this year or early next.

Comments closed
    • fhohj
    • 6 years ago

    jeez now I know they’re running out of ideas.

    really? your superpower is that you can control SSDs? really? come on.

    • Tech Savy
    • 6 years ago

    From what I understand about this it’s currently related to PCI gen 2.0 and has a bandwidth of 1Gb/sec. SataE is also supposed to be able to function with PCI gen 3.0 at 2Gb/sec.

    What I don’t understand is how is this in any way a benefit when current Sata 3.0 is capable of 6Gb/sec? Why are we going backwards with the progress of technology? I have only read one article on it and maybe there is something I am missing about features or other capabilities that make it better and different from current Sata technologies.

    If anyone can explain this I would appreciate it.

      • ddarko
      • 6 years ago

      SATA Express has a transfer rate of 1 GB/sec (gigabyte) – that’s with a captal “B.” Lower case “b” is gigabit per sec (Gb/sec). 1 GB/sec is equal to 8.59 Gb/sec so you can see that even dual lane SATA Express is faster than SATA with 6 Gb/sec.

      This is common enough mistake that is entirely the fault of scientists who came up with this nomenclature.

      • Parallax
      • 6 years ago

      Those PCIE numbers are in gigabytes (GB) not gigabits (Gb). A single PCIE 3.0 lane should provide about 1GB/sec (in each direction), so a dual-lane controller could in theory read/write 2GB/sec.

    • DPete27
    • 6 years ago

    Ugh, I don’t know why anyone would even bother making SATAe storage. Just make the transition to M.2 (NGFF) and be done with it.

      • davidbowser
      • 6 years ago

      I was wondering on that myself

      • frenchy2k1
      • 6 years ago

      It was a great idea when it was designed, 18 months ago, but now it is sandwiched between M.2 and pure PCIe cards and both are much better
      – both can be faster (2 lanes for SATAe, 4 lanes for M.2 and 16 lanes for PCIe)
      – one is smaller (M.2), llowing to fit in smaller devices
      – one is larger (PC card), allowing for bigger storage

    • SuperSpy
    • 6 years ago

    I really hope they don’t choose SATAe as the official short name. I can see people confusing it with eSATA already.

      • September
      • 6 years ago

      Too late, it’s done.

    • the
    • 6 years ago

    I still wish they’d aim higher. How about going for 4 lane of PCI-e 3.0 controller for a 4 GB/s maximum bandwidth?

    Also any word of NVMe support?

      • jjj
      • 6 years ago

      At CES they had the Marvell 88SS9293 that is x4 but haven’t seen more news on it yet.

      • willmore
      • 6 years ago

      Yeah, this 6Gb/s to 10Gb/s increase isn’t even the 2x increase per generation that we’ve come to expect every few years.

      That said, BW isn’t the only reason to favor SATAe (horrible name, BTW. eSATA? eSATAp? SATAe? eSATAep?) because there’s supposed to be a latency decrease and a # of simultanious transactions in flight at any one time, right? That could make a huge difference even if there were *no* increase in BW. That said, it would have been easy to sell a BW increase to 20G/s and then let the latency improvements win the arguement when they were in the wild.

      But, to do any of that, we really need some high performance, native SATAe controllers, not just some glued together last gen parts. So, this Marvell is a good sign.

      Edit: Bit, byte? Does it really matter? Yes.

        • the
        • 6 years ago

        SATA Express with AHCI will show some latency benefits but I have a feeling they’ll be marginal compared to SATA Express using NVMe. The software stack for NVMe is actually optimized for SSD’s where as SATA Express with AHCI could [i<]in theory[/i<] be used for a mechanical hard drive. With regards to bandwidth, supporting 4 lane PCIe 3.0 is worth while. Some NVMe implementations are hitting 3.0 Gbyte/s of bandwidth already. It just feels like everytime there is a new storage bus interface, SSD's to date have been closing to saturating it right out of the gate.

          • frenchy2k1
          • 6 years ago

          I have seen devices saturating either 8x PCIe Gen2 or 4x PCIe Gen3 (so ~4GB/s both ways).
          It’s hardly a problem to saturate in sequential read/write, just have enough parallelism (lots of NAND chips, an obligation anyway as each chip storage is limited) and request big payloads.

          The real benchmark for those drives are IOPS (number of operations per second).
          Typical IOPS numbers are given at 4k payload and NVMe PCIe drive hit several 100ks IOPS. Raise payload to 4MB and you will saturate your bus…

          Coming devices will be 8x PCIe gen8 leading to ~8GB/s.

          From my point of view, this leaves SATAe in an awkward place: who needs it? M.2 will rule the portable market, even getting into laptops and some desktop while pure PCIe cards will rule the datacenter.

          SATAexpress allows for up to 2 lanes of PCIe gen3 (or ~2GB), while M.2 allows for 4 lanes and PCIe cards are only limited by the largest connector in use so far (16 lanes).

          So, again, why use SATA express drives?

            • the
            • 6 years ago

            PCIe 16x slots will be the peak performers in the datacenter as you point out they can simply scale the number of ONFI channels. Long term I simply see this niche migrate to a DIMM-like form factor per channel to increase RAS and reduce replacement costs. (Why replace all the the entire card/NAND for several bad chips on a single channel?)

            SATA Express will still have a spot in the datacenter. I sse SFF-8639 backplanes being rather popular since they can accept pretty much any mix of drives. The smaller the server the more important SFF-8639 bays will be. PCIe 16x slots are a premium on small systems used for high speed networking/GPGPU cards and not necessarily easy accessible. This also prevents M.2 usage in the same niche. Cost will also be a factor and SATA Express will have the consumer side economies of scale to benefit it.

            The one thing I don’t get is the horrid cable mess the consumer side of SATA Express will have to deal with.

        • derFunkenstein
        • 6 years ago

        [quote<]eSATA? eSATAp? SATAe? eSATAep?[/quote<] SATAxp

    • dmjifn
    • 6 years ago

    [quote=”Geoff Gasior”<]It also supports SRIS, otherwise known as Separate Refclock with Independent SSC (Spread-Spectrum Clocking)[/quote<]Inception!

      • Terra_Nocuus
      • 6 years ago

      that FLA is SRIS bsns

    • south side sammy
    • 6 years ago

    no matter how advanced the technology on the motherboards get you can bet they’ll still make sure the board gets a D-sub.

      • maxxcool
      • 6 years ago

      LOL and a ps2 port 😛

      • cmrcmk
      • 6 years ago

      Don’t forget COM headers!

Pin It on Pinterest

Share This