Comments Locked

23 Comments

Back to Article

  • peevee - Friday, August 2, 2019 - link

    Sustained random write IOPS correspond to the performance of 1 controller only. Weird.
  • extide - Friday, August 2, 2019 - link

    It's sustained is the peak of one controller -- not so weird.
  • peevee - Friday, August 2, 2019 - link

    It is sustained for 1 controller. See that 4M peak write IOPS? 1M peak per controller even assuming perfect scaling at peak.
  • eek2121 - Friday, August 2, 2019 - link

    Even with multiple controllers, chances are you won't get perfect scaling. Random I/O is hard, even for spinning discs.
  • FunBunny2 - Sunday, August 4, 2019 - link

    "Random I/O is hard, even for spinning discs."

    wow. given that spinning rust is, be definition, sequential and SSD is, by definition, non-sequential (one might say random, but that's not quite right). either way, random and sequential access are essentially the same on SSD. the more channels, the better, of course.
  • Scott_T - Friday, August 2, 2019 - link

    with a power consumption of up to 65 watts you'd think there'd be a fan on there.
  • FunBunny2 - Friday, August 2, 2019 - link

    "with a power consumption of up to 65 watts you'd think there'd be a fan on there. "

    "a minimum of 400 LFM of air flow"

    ergo, that has to come from the chassis. let Mikey do it.
  • eek2121 - Friday, August 2, 2019 - link

    My old 100 watt light bulbs didn't have a fan on those!
  • hoyabembe - Friday, August 2, 2019 - link

    To be fair, the filament is operating at thousands of degrees C though :D
  • PeachNCream - Friday, August 2, 2019 - link

    It would be more awesome if my NAND flash chips were operating at thousands of degrees. LED case lighting would be unnecessary as the glow changes from red to white hot. Someone get on this right away!
  • Dragonstongue - Saturday, August 3, 2019 - link

    that would be nifty as oh hell.
    seeing as NAND operates far better at higher temps (though need to keep everything else away from this level of heat)

    You could even put on some small vac tubes(icognito heat exchange to a vapor chamber to help it work better -more so if do a small TEC unit for 50w or so (just enough to keep the controllers happy) and if vac tube, would not be all that hard for an audio card maker to "sister up" with one of their designs, so you get wicked drive performance, a nice "working glow" and crazy nice "on board" sound.

    DAMN wish I had the coin I would make it happen .. 1 x16 length card with solid state/nvme not that hard, sound card, not that hard, put vacumm tube on either, really not that hard, marry them without bunch of cables or crap bloatware, that takes skill ( and $$$$$$)

    The glow from it working and/or vac tube instead of nasty RGB, I could get me some of that.
  • olde94 - Sunday, August 4, 2019 - link

    i really want to see this happen!
  • Samus - Saturday, August 3, 2019 - link

    Needs some LEDs too.
  • 5080 - Friday, August 2, 2019 - link

    NVMe 1.2.1, interesting. Why did they choose that over NVMe 1.3 or 1.4?
  • prisonerX - Friday, August 2, 2019 - link

    Probably because there is a lead time for their own development and testing, and you generally don't want to rely on the latest and most buggy, least supported standard for your new product, if you're a small company at least.
  • 5080 - Friday, August 2, 2019 - link

    NVMe 1.3 is a ratified standard since 2016
  • 5080 - Friday, August 2, 2019 - link

    It might be a mistake in the data sheet since the Phison PS5016-E16 does support NVMe 1.3
  • ksec - Friday, August 2, 2019 - link

    For Reference, last Generation, DDR3 1600 Mhz Memory is 12.8GB/S, this SSD has the speed roughly double or Dual Channel of DDR3.
  • Santoval - Saturday, August 3, 2019 - link

    Not quite. DDR3 RAM has more or less the same speed at both sequential and random read and write modes. SSDs and NAND flash work quite differently. They are fastest at sequential read/write and their performance drops fast when switching to random read/write, particularly of very small files.
    So it's really not such a big deal if SSDs require 6+ TB of NAND, 16 PCIe 4.0 lanes and no less than *four* PCIe 4.0 controllers to match dual channel DDR3 1600 only in sequential read & write speed.
  • p1esk - Friday, August 2, 2019 - link

    3 years warranty? I thought the standard for enterprise is 5 years.
  • mdrejhon - Friday, August 2, 2019 - link

    Great scott! Who needs such nuclear bandwidth? That thing must be flux capacitor powered!

    I'll take it.
  • FunBunny2 - Sunday, August 4, 2019 - link

    "If you're reading a multiple of a flash page at a time, then you can get random bandwidth to match sequential. Under the hood, decent SSDs actually round robin sequential reads across all of the underlying hardware, so they can run as fast as whatever their bottleneck resource is. The bottleneck could be error correction, the SATA link, or the channels (copper traces) between the NAND and the controller, for example. Random reads with enough parallelism give the SSD enough scheduling freedom to have the same throughput as sequential."
    -- russell_sears
    here: https://news.ycombinator.com/item?id=12162665
  • FastCarsLike - Sunday, May 31, 2020 - link

    This is already outdated folks, PS5 has the "fastest drive in the world". Get gud liqid!

Log in

Don't have an account? Sign up now