OCZ’s Z-Drive 4500 PCIe sports fast specs, smart thermal control

As desktop SSDs bump into the limits of the 6Gbps Serial ATA interface, server-oriented drives are enjoying the spoils of PCI Express connectivity. OCZ’s latest Z-Drive 4500 has a eight-lane PCIe 2.0 interface that allows it to hit a claimed 2,900MB/s with sequential reads and 2,200MB/s with writes. The drive is also rated for up to 252,000 IOps with 4KB random reads and 78,000 IOps with random writes. Yowza.

Like its Z-Drive R4 predecessor, the 4500 Series uses OCZ’s Virtualized Controller Architecture (VCA) to combine eight onboard SSDs based on current-gen SandForce silicon. This VCA mojo runs on a separate chip and presents the eight-way array as a single volume. It includes a custom queuing system, configurable redundancy, provisions to extend NAND life, and support for both TRIM and the SCSI command set.

In the Z-Drive 4500, this novel controller architecture is paired with 19-nm MLC NAND from Toshiba. The Japanese flash maker purchased OCZ in December, so it’s no surprise to see Toshiba NAND anchoring the latest Z-Drive.

The drive itself comes on a full-height, half-length card. A single-slot heatsink covers the circuit board, and temperature sensors combine with onboard intelligence to keep everything from overheating. If thermals exceed a set threshold, the Z-Drive will scale back its performance until temperatures return to normal.

The thermal throttling mechanism is new, and so are the drivers, which were redesigned from the ground up. Then there’s the WXL caching software, which is OCZ’s own. The software is being introduced alongside the Z-Drive 4500, and it’s capable of using the entire drive as a cache or splitting the volume between separate storage and cache partitions.

OCZ is offering the Z-Drive 4500 in three flavors. Here’s the run-down:

Capacity Max writes

Price $/GB
800GB 680TB $2,355 $2.94
1.6TB 1300TB $3,806 $2.38
3.2TB 2500TB $6,533 $2.04

The prices are high, but so are the endurance specifications. The 3.2TB version is rated to withstand 2.5 petabytes of writes. Of course, the 800GB variant is rated for only 680TB of writes—just a little beyond the progress of our ongoing endurance experiment.

To put the Z-Drive 4500’s endurance into perspective, recall that the old Z-Drive R4 3.2TB is rated for a staggering 120 petabytes of writes. The R4 has higher write performance ratings, too, but it’s based on a full-length board, is much pricier (the 800GB version sells for $4,800), and has only three years of warranty coverage. Despite its lower endurance ratings, the Z-Drive 4500 has a longer five-year warranty.

Comments closed
    • Waco
    • 5 years ago

    So all I’m reading here is that the prior drives weren’t designed very well if they had no thermal throttling mechanism. ioDrives have had that since the very first version…

    • Buzzard44
    • 5 years ago

    Sure, the write/erase cycles may still be in the “good enough” range with these drives (at ~800 cycles), but given that the number of write/erase cycles in these offerings has dropped by a factor of *48*, I think we’re quickly approaching the limit of MLC die shrinks, and thus price reductions.*

    Where’s the memristors, PCM, or whatever is next? MOAR SPEED, LESS MONEYZ!

    * Yes I realize I’m basing this off of one SSD from one article, but I’m ballparking, yo.

      • Inverter
      • 5 years ago

      You are right, I don’t like this trend either, in particular because the page size, i.e. the minimal amount of storage that an SSD can write atomically, is growing, too!

      Imagine saving a 200 byte text file and saving it to disk, that will produce at least 3 block writes of 4 KB on the OS level (journalling, file meta-data block, file data) which will then translate to 3 page writes of 512 KB or so on the SSD level!

      That’s quite some write amplification!

        • Chrispy_
        • 5 years ago

        You’re right, but that’s a bad example with a hypotetically dumb, cacheless controller.

        Most drives and controllers have write apmplification in the region of 2.5 under a worst-case load, which I guess is why manufacturers reckon you write 1000x the drive capacity – it roughly corresponds to the 3000 writes the NAND can handle with a bit of headroom for warranty reasons.

    • Sunburn74
    • 5 years ago

    Still don’t understand the markup in price compared to conventional SSDs

      • MadManOriginal
      • 5 years ago

      Production volume? Quality control? Warranty and support?

      OH WAIT, OCZ! Well, 1 out of 3 is good if you’re a baseball player.

    • anotherengineer
    • 5 years ago

    “The prices are high, but so are the endurance specifications. The 3.2TB version is rated to withstand 2.5 petabytes of writes. Of course, the 800GB variant is rated for only 680TB of writes”

    Would I be correct in assuming it is because of the amount of over-provisioning?

    From your link for the 800GB pricing, it appears to have 32 chips. I would assume each chip is 32GB each, putting the total capacity at 1024GB.

    Thereby giving the 800GB model a massive 224GB of over-provisioning.

    I think I will wait for the Plextor 512GB M6e with a 5-yr warranty, supposed to come in just under $1/GB
    [url<]http://www.kitguru.net/components/ssd-drives/zardon/plextor-m6e-pci-express-512gb-ssd-review/[/url<]

      • shank15217
      • 5 years ago

      actually the endurance rating for these drives are abysmal by any enterprise standards.

    • HisDivineOrder
    • 5 years ago

    Friends don’t let friends OCZ. Nor do they let their friends go with Toshiba storage because Toshiba hard drives have a reputation for… less than excellence.

    In short, friends don’t let friends get anywhere near OCZ or Toshiba.

    Guess that makes Toshiba the perfect buyer for OCZ. Well, Seagate probably would have been first pick, but Toshiba’s a close second.

      • Chrispy_
      • 5 years ago

      OCZ had great products with poor validation and quality control.

      Toshiba hard drives seem more reliable than Seagates when you look at mass statistics, but people don’t spread FUD about Seagate any more than any other brand.

      IF OCZ is now backed by Toshiba testing and QC, and products like this come with a 5 year warranty, I’m going to give them the benefit of the doubt.

        • Crackhead Johny
        • 5 years ago

        Am I thinking a different brand? I though OCZ was the company that Tom’s Adware.. er Hardware caught buying off the shelf RAM, then filing off the maker’s logo, and gluing heat sinks on the sticks. After that they sold it at a super premium as the ultimate overclocker’s RAM?
        Am I thinking of a different company?

          • Chrispy_
          • 5 years ago

          Technically yes, legally no.

          OCZ has had a few reboots. Ryan Petersen, the founder and CEO until 2011 was a bit of a crook, but the company was generally good for PSUs after purchasing PC Power and Cooling in 2006, and their storage division was the first company to put out a decent, affordable SSD for the masses (Indilinx).

          Sandforce probably wouldn’t exist without OCZ,
          Indilinx probably wouldn’t exist without OCZ,
          OCZ offered one of the first PCI-E SSD’s for consumers
          OCZ offered the first serious attempt at SSD cacheing of large mechanical drives
          OCZ (through acquiring Indilinx) make some of the fastest consumer SSD’s in production right now.

          Basically, OCZ has been a historically divided company with some poor decisions, a shady past and some downright bad luck, but I don’t think there’s any reason to worry about them with Toshiba at the helm.

            • Haserath
            • 5 years ago

            I have seen AMD’s future…

      • Deanjo
      • 5 years ago

      Hmmmm, toshiba drives actually have an excellent reputation, better then Seagates. Toshiba’s 3.5 line are rebrands of the highest rated Hitachi deskstar line and their SSD’s are top notch as well.

      • albundy
      • 5 years ago

      their psu and ram that I still use are of good quality. it’s only been like 7 years for the ddr3 1600 platinum ram and 5 years for the 650w 80+ psu. i’d definitely let friends OCZ!

    • JdL
    • 5 years ago

    This has fantastic implications for applications that would have previously required massive data available in-memory. For sure, the CPU will be the bottleneck, as will the network stack.

    • MadManOriginal
    • 5 years ago

    Toshiba should just ditch the OCZ branding altogether, it just doesn’t have a good brand reputation.

      • derFunkenstein
      • 5 years ago

      Yeah Id’ rather buy a “Toshiba” drive than an “OCZ” one, even though they’d probably be identical.

        • Srsly_Bro
        • 5 years ago

        If the brand was “Defective SSDs” it would instill more confidence than the OCZ brand.

          • MadManOriginal
          • 5 years ago

          They could brand them as SSD and say it stands for Sucks Some D***. 😮

            • Srsly_Bro
            • 5 years ago

            LOOOOL. I bet that would be part of the call script when you ask for warranty support for an OCZ product.

    • kamikaziechameleon
    • 5 years ago

    I’ll take 5… Wait I didn’t say I’d pay for them.

    • UnfriendlyFire
    • 5 years ago

    For a second, I thought OCZ was still alive.

    I think Toshiba took a risk with using OCZ’s brand…

      • internetsandman
      • 5 years ago

      Completely forgot about this. Most of the comments (mine included) seem to be questioning whether this is going to be as unreliable as typical OCZ products. How much of OCZ’s operations have been taken over by Toshiba?

        • MadManOriginal
        • 5 years ago

        All of them AFAIK.

      • slowriot
      • 5 years ago

      I thought I came across an older news post for a second. I can’t believe Toshibia is using the OCZ brand either.

      • Srsly_Bro
      • 5 years ago

      I’m still terrified by the name OCZ, regardless of current owner.

    • internetsandman
    • 5 years ago

    Does OCZ have the same reliability problems and reputation in the server/workstation industries as it does in the consumer space? If so, that’s a lot of coin to drop on corrupted data

    • jdaven
    • 5 years ago

    That’s almost as fast as DDR400 single channel memory.

    Good job, OCZ, I mean, Toshiba.

    • dmjifn
    • 5 years ago

    Man, I already own a Toshiba SSD and have to say that I’m happy with it… am I a jerk if I snub this because they got some OCZ on it? Couldn’t they have named it the new “Toshiba Hindenburg 4500” or something more reassuring instead?

      • derFunkenstein
      • 5 years ago

      Speaking of the Hindenburg, spinning drive manufacturers are missing out on not naming helium hard drives names like “Blimp” or “Zeppelin”.

    • drfish
    • 5 years ago

    Hmm… I’d be willing to give OCZ/Toshiba another shot but there would have to be NO compromises in performance, none of that, “well you can see here that the controller is clearly limiting performance due to the number of transactions” crap. The other no-deal compromise would be price (for a consumer variant), seems like a PCIe 4x 512GB version at $600-700 would be interesting… Existing offerings still fail to meet the no performance compromise criteria…

    • ClickClick5
    • 5 years ago

    I can’t fathom what a read/write speed of this caliber would consist of. A Windows system will boot fully in 7 seconds with a good SATA SSD already. The reset of the system would have to be a bottleneck with this. I mean it could load the ENTIRE Windows 7 install in ~6.8 seconds.

    By the time I finally go SSD, it will be worth it all.

      • Dashak
      • 5 years ago

      Considering how much faster [i<]any[/i<] three or four year old SSD is over [i<]any[/i<] HDD, it's been worth it for quite a while. I think you know this already, but I thought I'd just emphasize it for others that may have been living under rocks.

      • Wildchild
      • 5 years ago

      Your computer would do things before you even made any decisions.

        • Haserath
        • 5 years ago

        So… Pr0n already loaded right as we sit down then?

      • Krogoth
      • 5 years ago

      Windows boot times are CPU-bound with SSD media.

        • derFunkenstein
        • 5 years ago

        I believe that – just watching the CPU meters in VMware it seems that at least with Windows 7, booting is done on a single thread.

          • ClickClick5
          • 5 years ago

          I can +1 this too. My main system at work is Linux, with Windows 7 running in a VM.

          • Ninjitsu
          • 5 years ago

          Probably, they were touting a multi-core startup in windows 8…

            • derFunkenstein
            • 5 years ago

            Oh, guess I didn’t realize that. Maybe that’s why it feels like 8 starts faster on my PC than 7 did.

            • cygnus1
            • 5 years ago

            It’s really been multi-core for a while. What they did in win8 is eliminate many start up process inter-dependencies so that they could then run concurrently

            • Haserath
            • 5 years ago

            It looks like they should at least switch out the Windows 7 boot duration for Windows 8 or <insert linux choice here>

            *They as in Tech report in their SSD reviews

      • MadManOriginal
      • 5 years ago

      1) SSD is already worth it, and has been for a few years ever since they hit ~$0.50/GB.

      2) Enterprise use, not single user use. That means things like databases where high random IOps matter.

    • TwoEars
    • 5 years ago

    I’m sorry but the OCZ name is dead to me.

    Too many missteps.

    • Neutronbeam
    • 5 years ago

    Wow. I am standing here beside myself with awe.

      • chuckula
      • 5 years ago

      Excuse me. I just bumped into you next to you.

Pin It on Pinterest

Share This