Comments Locked

64 Comments

Back to Article

  • caleblloyd - Tuesday, May 19, 2015 - link

    Pagination links are broken, on mobile at least... Can't navigate to page 2 to see the factory :(
  • Kristian Vättö - Wednesday, May 20, 2015 - link

    On my end everything seems to work fine (even on mobile). What happens if you try to access the second page directly?

    http://www.anandtech.com/show/9218/ocz-fab-tour/2
  • close - Wednesday, May 20, 2015 - link

    I have to ask, as some things look surprising to me:
    1) So every new SSD already has 8 times the capacity of data already written to it? Or is it just QC and natch testing?
    2) I always imagined the FW write process as being automated. But this looks like a lot of manual work to connect each drive by hand and write the FW. Again, is this the standard process or only during the initial testing phases?
  • close - Wednesday, May 20, 2015 - link

    And on the same note, I always assumed the labeling process is automated. Either they have really low volume or labor is THAT cheap.
  • menting - Wednesday, May 20, 2015 - link

    i'm not the official answer, but it should already have 8 times the capacity of data written in, and then the firmware should zero out the counts.
  • close - Wednesday, May 20, 2015 - link

    What I'm not sure is if this happens to all drives or to selected drives, assuming that if a few drives are OK the whole batch must be. Also, the testing is done after writing the FW. Is the FW "pre-configured" to ignore the first 8 writes per LBA or do they go through connecting them to PCs all over again to reset the written data counter?
  • dreamslacker - Wednesday, May 20, 2015 - link

    They would do it for every SSD. The actual usable capacity of the modules aren't fixed or a known quantity until you actually test every cell. During this phase, you will also know which cells are 'bad' and to discard/ repair the SSD if the remaining usable cell count is lower than the set limits.
    The usable cells will then be mapped into the table so the controller knows what cells to avoid using.
    This procedure is done on mechanical disk drives too since the actual platter capacity isn't a fixed number either.

    As for the write or test process, it depends on the volume and the manufacturer. If volumes are high enough, you might not even have workers handling the F/W write or test process. A fully automated robotic arm and conveyor belt system would handle the drives and label them accordingly. Leaving the workers to package the drives.
  • MikhailT - Wednesday, May 20, 2015 - link

    1. Correct, this is what is known as the "burn-in" period. You have to write to every single NANDs or even hard drive platters a few times to make sure it is working. Many company burn in computers as well, they finish building it and then run a custom automated tool to benchmark it severely for several hours before they can ship it to you.

    Think about the electronics, 90% of the defects (my experience and others I've talked to) are usually found within the first few days of use. That's usually a sign that the company did not properly burn/test the device in before shipping it to you.

    2. It depends on the experience of the company. It cost a lot of money (machines are expensive and you have to hire people to figure these things out) to start automating the stuff and it would actually be cheaper initially to do it by hand as you have less volume to work with. As you get more money from your business revenue and volume starts to ramp, you then hire a few folks to figure out how to automate things, if it is cheaper and worth, you then invest hundred of thousands of dollars or millions to buy these equipment. That's why in the first page, it talks about this in phase 3 about committing the funds in terms of millions of dollars.
  • close - Thursday, May 21, 2015 - link

    I assumed this is done before assembling the product. So you bin the chips, check them for errors, etc. before you solder them to a PCB. This way even if you're not the manufacturer of the NAND you still get to differentiate between chips and put the better ones in better products.

    If you do the burn in and checking for defects AFTER they're soldered you're basically guaranteeing that all defects will be remedied at extra cost.
  • Kristian Vättö - Thursday, May 21, 2015 - link

    NAND binning is usually done by the NAND manufacturer or packager, but there may (or actually will) still be bad blocks. The purpose of run-in testing is to identify the bad blocks so that the controller won't use them for storage as that could potentially lead to performance issues or data loss.
  • dreamslacker - Thursday, May 21, 2015 - link

    Not quite. You can only do that if the entire process is automated and centrally controlled.

    Their SMT line uses tape reels for parts, if you wanted to pre-test NAND flash before mounting them, you will need to extract, test and reinsert into the tape reels (or trays) in a specific order.

    The mapping information from testing will also need to be entered into a centralized database. The reels or trays will then need to be sent to the SMT line in the same specific order and the circuit boards, once soldered, will also have to be sent to the firmware flashing racks in order as well. Otherwise, there is no way of telling which board contains which flash chips.

    Again, they have to extract and send for labeling in a specific order so that the label serials will match what is programmed into the drive.

    All of these basically means you need to invest into a completely automated line with conveyor belts, robotic arms, automatic labeling machines, database servers and apps to integrate.

    It's simply more economical, for a smaller scale operation like this, to solder first and ditch the defective units as long as your components do not come with a high defect rate.
  • caleblloyd - Thursday, May 21, 2015 - link

    I think the article got posted Tuesday morning then taken down a few minutes later. But all good now!
  • dreamslacker - Wednesday, May 20, 2015 - link

    Odd that they intend to both shift towards a rack-based system and also MST at the same time.

    A decent rack type system would be running ASICs that offload the test regime from the host computer(s) and can be configured on-the-fly to run both drive tests as well as programming.
  • DanNeely - Wednesday, May 20, 2015 - link

    Even if most of the testing can be done on board the drive itself using the MST; they still need to plug the drives into something else to verify the sata connection and to read out the results of the MST. MST will just let them swap new batches of drives into the tester more frequently.
  • dreamslacker - Thursday, May 21, 2015 - link

    MST is normally used when you have reduced capability test chambers / racks. For mechanical drives that need to be tested under high temperature conditions, it allows them to dump a load of drives into what is effectively a large oven with triggered power supplies for thermal tests.

    If they are utilizing 'smarter' racks, such as the Xyratex (now owned by Seagate) modules, then the ASICs will handle everything and loading/ unloading MST firmware manually isn't needed.

    The primary difference is that a smarter ASIC based unit can automatically run the tests, dump the logs and prep the drive for sale (clear the test data, reset the counters etc) whereas MST requires the worker to extract the drives and send them to another workspace to have the MST logs dumped and the production firmware loaded in.

    The ASIC based test racks still connect to a computer, you just have the ability to use a single computer for more drives (for Seagate's Xyratex modules, it's 1 computer to 192 drives). They are, however, headless units and are fully controlled by a central server via CGI & Python scripts.
    Very little user intervention is required on their fully automated line except when they need to re-configure the racks to accept modules with different drive interfaces. This is done by a single terminal console where they telnet into the machine handling the modules and edit the configuration file by VI.
  • GTan36 - Wednesday, May 20, 2015 - link

    Bought a OCZ Arc100 240GB SSD last week for my first PC build. It's really fast and you can tell the drive is high quality and manufactured well. Boots Windows 8.1 in less than 15 seconds and was pretty affordable.

    It's nice to see OCZ becoming a quality brand.
  • r3loaded - Wednesday, May 20, 2015 - link

    Excellent article, it was like a written version of a How It's Made episode!
  • Phynaz - Wednesday, May 20, 2015 - link

    Presented by Anandtech, a division of OCZ Marketing
  • xthetenth - Wednesday, May 20, 2015 - link

    Yes, this tour of a company's factories and discussion of how they make things is different than all the others because
  • MrSpadge - Wednesday, May 20, 2015 - link

    Don't fault AT for OCZ being more open than other manufacturers.
  • Murloc - Wednesday, May 20, 2015 - link

    lol come on they visited other factories too....
    This article is neutral.
  • jihe - Wednesday, May 20, 2015 - link

    You forgot Step 0 in the process: stockpile C-grade discarded nand chips from Micron/Intel/Samsung.
  • valinor89 - Wednesday, May 20, 2015 - link

    The point of being owned by Toshiba is precisely to avoid having to do that.
  • ocz_tuff_bunny - Wednesday, May 20, 2015 - link

    Hi jihe,
    Since this is my first comment in regards to this article please allow me to identify myself as an OCZ representative. Thank you for your comment. Being apart of a Toshiba Group Company all of our drives leverage premium Toshiba NAND. In the past few years we made significant upgrades in our products and processes to improve quality and continue to make investments in this area.
  • sor - Thursday, May 21, 2015 - link

    This would surprise me. I've worked for memory manufacturers in the past, and you'll often find Samsung chips on Micron memory modules, and vice versa, simply because a company might choose to sell their die into more lucrative channels, or focus on manufacturing more lucrative configurations, or fulfill more lucrative contracts with the capacity they have. Then they supplement with product from competing manufacturers on their own products. People who pay attention to such things are occasionally surprised to see it, but in business its the bottom line that counts, and if you can make $.50 more per part by putting a configuration that you don't need internally on your wafers, and source your internal needs cheaper elsewhere, then you do it.
  • Kristian Vättö - Thursday, May 21, 2015 - link

    In the DRAM industry it's true that the original manufacturer of the chips varies more, but on the other hand nobody claims to use brand X chips in their modules (it's always just the speed and latency that are reported). In SSDs it's different as manufacturers specifically state what NAND they use, so switching from one NAND to another would be misleading marketing (and we've had some do that in the past, which lead to ugly results). Another thing is that SSD performance depends highly on NAND, so switching the chips inside would lead to varying performance unlike in DRAM.
  • jihe - Thursday, May 21, 2015 - link

    Well, will you promise to use Grade A Toshiba NANDs only then? Because OCZ had been one of the worse offenders of switching to terrible NAND chips in their SSD's, often within the same model and without any disclosure to the consumer. I would go as far as saying OCZ were using bait and switch tactics. Given your history it would take a lot to sway public sentiment. Perhaps having a NAND chip guarantee would go some way in correcting that.
  • dgingeri - Wednesday, May 20, 2015 - link

    I'm quite surprised by their firmware install process. I thought it would be more automated and part of the QA process. I also thought they would have automated the stickers long before now.
  • Wwhat - Wednesday, May 20, 2015 - link

    Many factories have such a weird final hand-done step. I think it's just so people have a job.
    And although that's fine, I think it does show the flaws in the capitalist system and the society's thinking.
  • jimjamjamie - Wednesday, May 20, 2015 - link

    "..running some sort of a Linux distro.."

    If you zoom in on the photo you can see that the applications menu button (top left) has a Fedora logo. Interesting choice!
  • fatpugsley - Wednesday, May 20, 2015 - link

    A root GUI with presumably obsolete stock Fedora doesnt seem very secure to me.
  • Stahn Aileron - Thursday, May 21, 2015 - link

    Doesn't really matter if the only network they're connected to is an internal network without internet access. The only thing those workstations should be touching and working on are SSDs. There's minimal security needs there. The best and simplest security measure there would just be an air gap.
  • dreamslacker - Thursday, May 21, 2015 - link

    It's not as uncommon as you might think. Most of the machines are effectively on an 'intranet' so they really only need a UI and kernel base that has reasonably good OoB driver support for commodity hardware and the dependencies for the specific in-house software suite.

    When I worked a short project on similar lines in Seagate back in the past, their machines ran Centos 4.4 (at the time) and they started off that line-up on RHEL before moving to Fedora Core and then to Centos 4.4. They were planning a move back to Fedora after that though.
  • der - Wednesday, May 20, 2015 - link

    They're made to they can dig deep to our pocket monies.
  • Scott_T - Wednesday, May 20, 2015 - link

    Next time make us a movie in the format of 'How its Made'
  • MrSpadge - Wednesday, May 20, 2015 - link

    I'd rather read an article at my own pace.
  • jann5s - Wednesday, May 20, 2015 - link

    Cool stuff Kristian!
  • YoloPascual - Wednesday, May 20, 2015 - link

    Wow a field trip.
  • OzzieGT - Wednesday, May 20, 2015 - link

    That poor soul who has to apply labels all day...
  • junky77 - Wednesday, May 20, 2015 - link

    this fake smile.. seems like all the people interviewed on AT and such have this smile lol
  • Murloc - Wednesday, May 20, 2015 - link

    cosmically in tact?

    Lol I guess it's the jet lag or the phone dictionary.
  • harrynsally - Wednesday, May 20, 2015 - link

    From reading actual user reviews (e.g. # of 5/4 stars vs. 1/2 starts at places like Newegg), I have not gone wrong after buying a number of Samsung 850 EVO and Micron/Crucial MX100 250GB class SSDs.

    Of course it's not just counting stars, but understanding the user experience/satisfaction.

    Although SSD street prices have really come down recently (e.g.currently 250 GB class 850 EVO = $89, ARC 100 $69 after $20 OCZ rebate etc.), will move up to 500 GB SSDs. After reading reviews in this size, my money still goes to Samsung and/or Crucial.
  • ocz_tuff_bunny - Wednesday, May 20, 2015 - link

    Hi harrynsally,
    Thank you for your comment. We totally agree when choosing a SSD it's not just counting stars, but understanding the user experience/satisfaction. OCZ is a new organization under Toshiba and have made significant changes to everything from processes to production. We hope one day we will have the opportunity to show you our commitment to user experience/satisfaction through our products.
  • RU482 - Wednesday, May 20, 2015 - link

    ummm, where did my post go?
  • Shadow7037932 - Wednesday, May 20, 2015 - link

    Wow, I'm surprised they are using normal PCs for loading up the firmware. I would have expected some custom setup for this. I guess at these volumes it makes sense to to it this way.
  • RyanVM - Thursday, May 21, 2015 - link

    Hate to say it, but OCZ as a brand is too toxic for me to touch anymore.
  • edlee - Thursday, May 21, 2015 - link

    I am not sure why they are still using ocz brand, I would much rather hear them call it Toshiba performance line, ocz will always have a negative connotation in my subconscious.

    I understand its run by Toshiba, and completely different management style but still, I feel safer with a big brand name behind my ssd, like crucial and Samsung. I would accept Toshiba as a quality band supplier , just dump the ocz brand
  • Murloc - Thursday, May 21, 2015 - link

    I don't feel safe with samsung after the 840 evo fiasco
  • Oxford Guy - Friday, May 22, 2015 - link

    The EVO sucks but it's not even close to the OCZ fiasco with the Vertex 2. That involved pure bait and switch fraud and worse. Then they followed it up with the 3 line which was also buggy.
  • Oxford Guy - Friday, May 22, 2015 - link

    "I am not sure why they are still using ocz brand"

    For the same reason Nvidia's website lists the 970 as having 4 GB of 224 GB/s VRAM. They know consumers are stupid, generally — with attention spans rivaling the common gnat.
  • Pessimism - Thursday, May 21, 2015 - link

    Disappointing that marketing weighs so heavily into their design process.
  • Refuge - Thursday, May 21, 2015 - link

    Thats how it is everywhere.
  • ryanmt - Thursday, May 21, 2015 - link

    "and has been achieved through science, but there's also trial and error in the mix. "

    What exactly to you think 'science' is?
  • ArmedandDangerous - Thursday, May 21, 2015 - link

    Very informative, but I have to say, the writing style and quality feels like a regular blogger putting up posts on a blog. It didn't have the usual insights and technological knowledge of a Anand writer. Waiting for more works from you Kristian, and keep on improving :)
  • Slappi - Friday, May 22, 2015 - link

    Ralph Schmitt should be thrown in jail along with OCZ's former CEO who robbed shareholders of hundreds of millions of dollars. I wouldn't buy an OCZ drive if it was the fastest drive on the planet.
  • Oxford Guy - Friday, May 22, 2015 - link

    Who doesn't like a little bait and switch? Send out Vertex 2 drives with 32-bit NAND then sell drives with half the NAND chips to consumers. Whoops!

    Then come up with a "voluntary replacement plan" that doesn't apply to the expensive (at the time) 240 GB model. And, top it all off with spectacularly buggy drives — drives that are basically ticking time bombs waiting to brick when the user does something weird like put their laptop to sleep.
  • Ethos Evoss - Sunday, May 24, 2015 - link

    Jeeezus picked up worst company ..
  • Richard Buyme - Sunday, May 24, 2015 - link

    Good ole' ocz
    Same as the original ocz.safeshopper.com that conned and scammed end users when they were a reseller courtesy of Ryan Petersen
    Sadly, time heals all wounds it would seem, even on the internet. Who knows how many thousands of dollars this corporation reamed out of customers' wallets back in the day (circa 2000-2001), orders taken but not shipped.
    This article details some of what OCZ's founder was like before he abruptly resigned without comment in 2012
    http://seekingalpha.com/instablog/890075-copperfie...
  • hirschma - Monday, May 25, 2015 - link

    The article isn't quite correct in some areas.

    The first part of the process is called "stencil printing". A thin, stainless stencil is used in a stencil printer (more or less a robotic squeegee) to apply solder paste to the board. There are no sockets - the paste is applied directly to the pads.

    Solder paste does not spoil after two hours; while it is temperature sensitive, it isn't that sensitive. It probably takes over two hours for a jar that size (.5 kg) to come up to room temps.

    Although it isn't mentioned, I'd expect an outfit like this to do a paste inspection before the pick and place machine places the components.

    Lastly, the choice of using through-hole, hand soldered SATA connectors is odd. There's no reason why they couldn't use surface mount just as easily (although through-hole is going to withstand more abuse).
  • hirschma - Monday, May 25, 2015 - link

    Heh, took another look.

    If they're doing things "according to hoyle", there should be an X-Ray inspection for BTCs - bottom terminated chips. Just because a board works, doesn't mean that it'll continue to do so if there are certain defects that can only be caught that way.

    Also - strangely - they do appear to be using surface mount connectors, but why they put them on in a separate process (if I read the article correctly) is very strange. Might have something to do with the testing - they may do an initial test before singulation (removing the individual boards), and then only put the connects on if it passes (the connectors are probably more expensive than the labor).

    Last comment - not sure why they're using such small panels (i.e., only 4up). The standard in China is a square meter - likely due to the the contract manufacturer having equipment that can only handle smaller panels.
  • dreamslacker - Tuesday, May 26, 2015 - link

    Seems like they have to de-panelize the boards before mounting the SATA connector because of the orientation during the panel design (the SATA connector pads are facing each other).
  • JoyTech - Monday, June 1, 2015 - link

    Kristian: just curious roughly how many people are employed at the OCZ SSD factory? Maybe it is confidential, but curious as to the number of employees, because you mentioned OCZ looking at automating parts of their process; where the process as of now looks pretty automated, especially for the critical steps.
  • belawrence - Monday, June 1, 2015 - link

    For those mentioning "How It's Made", they did an episode showing the manufacturing process of SSDs already - I've seen it at least ten times by now.
  • boe - Monday, June 15, 2015 - link

    bring on some 10TB and 32TB 3d NAND drives!

Log in

Don't have an account? Sign up now