The SSD Review The Worlds Dedicated SSD Education and Review Resource | 
Fixstars Solutions, Inc., a Tokyo-based innovative producer of flash storage devices, is announcing that it has commenced sales to the North American markets of its 1TB SSD-1000M and 3TB SSD-3000M solid state drives (SSDs). These drives offer enterprise-grade reliability and outstanding sequential read and write performance, and are geared toward datacenter, high-performance computing (HPC), advanced driver assistance systems (ADAS), and especially for professional content creation.
When released in Japan last November, the SSD-3000M 3TB was the world’s highest capacity 2.5” form factor SATA 6Gb/s SSD. Higher capacity drives allow for reduction of the number of drives required in professional usage scenarios, also reducing operational costs such as power, cooling and maintenance while also reducing storage infrastructure footprint. Of particular value is creation of a more reliable workflow and minimized data handling failures. The SSD-1000M and SSD-3000M both feature Fixstars’ proprietary NAND controller, which prevents latency spikes and performance deterioration, thus assuring consistently high performance. Applications such as 4K video recording and editing and encrypted film storage require fast and stable disk writes, and will benefit the most from using these new Fixstars SSDs.
According to Satoshi Miki, CEO and Co-Founder of Fixstars Corporation, “The SSD-3000M/1000M were released in Japan last November, and have been getting great feedbacks from our customers. As an innovator of storage solutions, we are focused on providing high performance and reliability SSD solutions, to accelerate our customer’s business.”
Fixstars’ new offerings feature 19nm MLC NAND and a SATA 6Gb/s interface. These are encased in a 100mm x 70mm x 9.5mm enclosure with a total weight of under 115g. Sequential read speeds are stated as (up to) 520 MB/s in both standard and boost modes. Sequential write speeds are stated as (up to) 350MB/s in standard mode, and increase to (up to) 500MB/s in boost mode. The TRIM function is supported with operating systems that also support it, as well as garbage collection and S.M.A.R.T. drive health and performance reporting attributes. Power consumption at idle is stated as less than 3W. Active power consumption is stated as less than 5W in standard mode, and less than 6.2W in boost mode. The SSD-1000M and SSD-3000M are backed by a one-year limited warranty.
The SSD-1000M and SSD-3000M are available now for order on the Fixstars website. According to the listed specifications chart, they are also planning a 5TB SSD-5000M, to be available later in Q1 of 2015. You can visit the SSD-1000M/SSD-3000M product page here; and you can view the Fixstars press release announcing these new drives in its entirety here.
>When released in Japan last November, the SSD-3000M 3TB was the world’s highest capacity 2.5” form factor SSD.
It wasn’t. Sandisk beat them with their optimus max line (maxxing at 4TB) a full 6 month earlier.
Article should have said world’s highest capacity SATA 6Gb/s SSD. The Optimus Max line are SAS drives. Article has been amended to reflect this.
Thanks, Benjamin!
The terabyte drive is $990 (on their site). That’s pretty pricey. Anything that makes it worth that premium over other 1 TB drives?
Any idea what the 3 TB is priced at?
That stuff is enterprise grade, so its bound to carry a pricepremium compared to typical consumer drives.
sorry, what’s SO enterprise grade here ? ONE YEAR LIMITED warranty ? You plan your business to crash pretty soon, do you ? Sub-par performance ?
19nm litography ? Pche. Don’t want to know the endurance, no matter what clever tricks they try to apply, the amount of re-writes to cells is limited. Point.
Sammy 850Pro and 845DC Pro will run circles around this. 5yr/10yr warranties, 10 DWPD = 14600TB written for 845DC Pro version, I will bet my ass that it will survive 10 times as much. Why ? Techreport.com report. Consumer 840Pro rated at few dozens of TB written still kicking after 2000TBW which is almost 100x more. And that’s STUPID CONSUMER drive based on 20nm litography. 845/850 are totally different league. 40nm litography which WILL survive more than 25 “enterprise” MLC you can find today, 3D without interfering so much in planar than existing drives do… it goes on and on.
I don’t see single thing that would drive me to buy this Japanese things. Yeah, the density is phenomenal (I have 800GB Sammies now, I could have 3000GB Fixstars which is almost 4x more dense). I don’t have two football fields of flashes yet and should I have, there would be NO SINGLE ONE with ONE YEAR LIMITED warranty because my data are worth much more.
The price ? $990 for 1TB model ? Laughable. 845DC Pro 800GB (which is the same raw capacity, just overprovisioning kicks it down) = 550eur where I live, which is about $650 max. Yes I know 800GB is not 1TB.
Performance. Don’t get me started. They’ve rather NOT specified sustained IOPS, 845DC Pro nets >90.000 random 4kB read and >50.000 random 4kB write, tested on this site. I don’t know single other SSD that can do the same in this price league or for twice as much. No, Intel 3700 can’t do 50.000 writes sustained and yes, it’s twice as expensive.
Consider everything.
Sorry, this product doesn’t stand a single chance in my datacenters no matter what the real performance will be.
One year limited warranty. How dare you ?! Duh.
>sorry, what’s SO enterprise grade here ? ONE YEAR LIMITED warranty ?
You plan your business to crash pretty soon, do you ? Sub-par
performance ?
Controller, highly binned nand, firmware…
> the amount of re-writes to cells
is limited. Point.
So is with any flash device out there.
>I don’t see single thing that would drive me to buy these Japanese things. Yeah, the density is phenomenal
Looks, a very good reason right there. High data density can be crucial sometimes. Why have 4 drives, if you can only have 1 ?
Obviously, you’re not the target group for this drive. That doesnt make it bad.
>> >sorry, what’s SO enterprise grade here ?
>> Controller, highly binned nand, firmware…
yeah, controller they buy from somebody else and have absolutely no control whatsover. They can’t say anything to development. They can’t say anything at all. I really doubt they have firmware team and heavy-duty custom firmware – Intel has been manipulating SandForce stuff, we know that. Have you seen Intel’s SSD torture procedures (anandtech) ? I somehow doubt FixMeQuick is as capable, don’t know why but I really feel that way. If I create a SSD and want to achieve “consistency”, I would just make it 1.2TB, don’t reveal, publish 1.0TB only and that’s my consistency. This has been done so many times recently.
Binned NAND. Yeah. At the moment they START MANUFACTURING NAND, then we can speak about binning. Until then, anything manufacturers throw at them is what these guys MUST accept. They can’t choose chips. They are not present at manufacture facilities. No matter what they say, they are not. Sure, they can say Micron/Intel/Toshiba/Samsung give them the best of breed, leaving only the second class for themselves 🙂
oh sorry, 19nm “binning” ? So eighties. Samsung doesn’t have to bind anything with 40nm [!!] 3D NANDs and still will be order of magnitude more reliable. Wanna bet ?
What’s the performance ? Have you seen any manufacturer not specifying ENTERPRISE-CLASS product with the uttermost details ? Once again I can’t help myself, but what I saw as the “manufacturers specification” is… crap. Trying to be polite. Unbelievable crap I haven’t seen for very very long time.
Whatever the situation is, 19nm binned will never be as reliable as 40nm 3D closed-eyes-non-binning. No matter what we say, ONE YEAR LIMITED WARRANTY was NEVER considered ENTERPRISE-GRADE. My apologies, but it really was not. This is how I see it.
Last but not least : small companies come and go. This is no Intel class, no Samsung class. Would you really bet your money on this in massive scale ? If I’m a CIO/CTO sitting on 0.2M$ or 2M$ or och 20M$ order, I would never consider these. CEO would fire me. Show me the webpage. Garage. Show me the product specification. Garage. Show me world-wide coverage. Non-existing. Show me anything real. Crap. Say you need the density. Sandisk. Say you need hundreds of petabytes of flash. Show me the price. Show me the belief in this noname company. One year warranty, still remember ?
This is not about fitting me or not. Type “intel p3700 specification” into google. First link. 44 PAGES of specification.
Please, show me anything like that from FixTheSky. Oh, for start, it really would be enough to show specified IOPS, sustained or not. Eh, what is the LIMITED warranty ? How much TBW ? DWPD ? Can you please find that out for me ? I can easily show that for Intel and Samsung drives, would you like document links and page numbers ? Intel even goes as far as saying every 12.5% over provisioning prolongs life XXX times. Curious ? Google it. 20 seconds max. and you have it.
I do apologize for sounding VERY negative. Really struggle to find any positives (don’t start on density please, 4U = 72 drives = 288TB = almost 3PB in rack, how much flash you need ? What’s the value of your data ? I’ve rather have three RELIABLE racks backed up with real warranty and real manufacturer than one rack of these where warranty expires before I wake up tomorrow). At the moment they would specify 5yrs warranty which has been the standard in enterprise class since-I-dont-remember-when, at that moment I would just express my doubts about stability of that company. Sure, they might not be around in 5 years and they really probably won’t be. But what we see now absolutely scares me away. Downright scares and can’t stop b**tch*n’.
They don’t believe in their own product. Saying “customers like our high reliability” backed with 1yr warranty is MEGA failure.
Would you buy new car with 5.000 miles warranty only ? Oh, really not ?
>Whatever the situation is, 19nm binned will never be as reliable as 40nm 3D closed-eyes-non-binning.
You obviously never heard of e-MLC before, hadn’t you ?
Also, all that ranting for a drive, which was just announced ?
Are you sure, you don’t work for some other competitor ? 🙂
sure I don’t work in any manufacturing or SSD/flash industry. I’m just a pure customer trying to choose the best for him. Tell me any single thing I got wrong please.
Totally incomplete data. We can’t call that a specification, I say. We’ll see. What is definitely clear is 1yr LIMITED warranty. End of the game for 99.9% REAL enterprises out there.
> Uh no. e-MLC ? Not in my house. 2D is done.
Whats wrong with e-mlc ?
It good for upwards of 40k p/e (850pro 3d nand is only good for 6k p/e).
I also highly doubt enterprise is ready to just move to 3D. Although its a better tech on paper, its still pretty much unproven in the field and nobody knows, what will come out if it. Thats why using tested 2D flash (remember 19nm flash is pretty old and tested nowdays) make sense, especially if you dont own a nand fab.
nothing wrong with e-mlc generally, that was related to “16nm e-MLC”. 16nm nothing SSD in my house or corporation 🙂
850Pro and 6k : that’s just calculation Anandtech has come up with based on SMART statistics – nothing else, nothing more. This doesn’t mean it can’t survive 50.000 P/E. We can’t judge by 850Pro (consumer class) what 3D is capable of. Remember this is 40nm, that’s unbelievably HUGE compared to anything we can meet today.
Back to real life : Samsung will never specify 100.000 P/E for 850Pro even if it is able to achieve it. That model doesn’t have customers that need it nor the price to warrant it (why would you risk it from business perspective ?). 128GB and 1TB versions have the same 150TBW specced durability – strange heh – and that’s magnitudes more than 99.99% of CONSUMERS need today and in next three years.
1TB version must be capable of 8x more, at least… just because it has 8x more cells. Unless they are all full all the time, which is not going to happen, firmware can perform better wear leveling etc, so the real difference might easily be not just 8x but 20x and see they have the same TBW. Business model we call it. Not too much competition.
3D has much less “cell interference” than planar 2D has, which also prolongs its life. 25nm 2D and 25nm 3D will have dramatic differences in real life just due to this.
The same Anandtech you quote with 6K P/E says Samsung said they have 128GB 850Pro version that already wrote 8000 TB and is still tickin’. I have not seen it myself to be honest, but I can believe it. Calculate with me : 8000 TB x 1000 / 128 GB = 62.500 P/E. 40nm ? Tick, EASILY capable of – history shows that !! 3D wearing cells less with cross-interference and all those things I don’t understand in detail as I’m not NAND engineer ? Tick again. This is not “my car can do 400mph and 98mpg” stuff, I feel there are real numbers in the background.
You know Techreport.com durability articles. 840Pro, 256GB, crossed 2000TB – not saying it was in the best condition or whatever, but it was still working and that alone already is 8000 P/E. Agree, not ALL cells survived it. We all know 840Pro has nowhere near 40nm and ain’t 3D. Once again 60.000 P/E is not so unrealistic. Right ? If I check out how P/E was shrinking with litography shrinks… my heart bleeds. I’m just putting numbers behind my belief. These are the reasons why they don’t seem unrealistic to me.
I honestly believe Intel 3D, Micron/Crucial 3D and Samsung 3D will have guaranteed 50.000 P/E or maybe even 100.000 P/E models this year. That’s 27 DWPD (50 DWPD) / 5yrs, not that extraordinary and it is achieved TODAY WITH CURRENT 2D <=25nm technology ! 3D will be much better, and it will be much larger. Yes, I strongly believe 50.000 P/E is nothing, absolutely nothing for 40nm 3D.
Enterprise not being ready for 3D ? that's just a corporate bullshhhh*t, inflexibility, time necessary for testing to provide guarantees etc. Maybe HP 3Par doesn't talk to Samsung about 3D usage because they have still 3 years remaining in contract with Sandisk (why they don't use Intel ? Toshiba ? Crucial ? Samsung ? Somebody got bribed years ago to sign GOOD contract ? Hell do I know ?). So they maybe just can't, even if they would like to. Maybe Pure, Violin or some of these guys will be first to open the box with 3D. And you will see it explode with extreme durability blablabla, and suddenly everyone will start wondering how we could ever survive without 3D. Maybe they are all testing it now. I don't know.
Yes, this is how business world works. What is the current competition in 3D ? None.
Uh, nice talking to you btw.
Nobody is denying 3D isn’t good (endurance is surely better than 2D nand).
But 2D has some life in it. It may be inferior in endurance, but thats not the whole story to the whole reliability.
On paper, 3D might sound better in every way, but that still doesnt make it better suitable for enterprise to just jump on the bandwagon.
Given how 3D is still an immature tech, there are plety of things that need to be sorted out for maximum reliability; like tested firmware, proper controllers….
I’m sure this SSD will find its way to certain nieches, where highy capacity drives on sata are must. It just happens, you’re not that nieche.
Also bear in mind that such prices on their website are “single lot” prices. If a datacenter, say, was purchasing multiple drives they would be priced differently based on purchasing a larger quantity.
SSD-3000M is made up by the FPGA and eMMC.
(Link destination is an article in Japanese.)
https://www.nikkei.com/article/DGXMZO75503950R10C14A8000000/
Wow, thats pretty much the worst way you can do an ssd.
Slow emmc chips and an a damn FPGA on a ENTERPRISE PRODUCT ???
Now i get it, @disqus_T33SOCFBTe:disqus
Did you read the linked article?
Merely to at least just deny there is no fairness.
This product, and to have been equipped with a lot of eMMC chip than the actual capacity, FPGA is used to integrate the management.
And, it is also easy to customize in response to customer demand.
Its still an FPGA. You used that to prototype or manufacture at small scale, not a finished product, where reliability is key.
For 3TB or more, it would make more sense to just use two phison s10 controllers (this seems to be one of the first 3rd party controllers to support more than 1TB) and raid0 them.
This would reduce complexity, since you dont need a big fpga to control everything but you just slap two of those controllers connected with a decent raid0 controller.
And since you can get 128GB packages nowdays, you would only need 24 of them to hit 3TB capacity (provided, 3rd parties dont have access to 16 die packages, otherwise this is down to 12).
Perfectly doable on a single pcb..
Ok. And that you have not read the article, you is I know that do not understand this product.
yeah, i have a hard time understanding brokem english, that comes from google translate.
It is certainly troublesome, but browsing of information should be.
It has been almost written in such a feeling.
I.The SSD of easily customizable new structure the speed and capacity
II.First, put the target for video devices, and medical devices,etc.
III.Was disassembled the Fixstars SSD-3000M. It will be 52 pieces equipped with eMMC chip of 64G bytes
IV. eMMC is equipped with a controller to manage the flash memory to itself. And SSD controller is not required.
Instead, read and write eMMC data using a FPGA, which exchanges with the external interface.
V.This has two advantages.
One is that the stable speed exits.
In order to manage the flash memory chips are dispersed at the eMMC, the speed of reading or writing is stabilized as a whole.
Another, that you can easily change the capacity and read and write speed.
Number of mounting the eMMC, and by increasing or decreasing the communication channel connecting eMMC and FPGA, it is possible to change the size and speed.
VI.Customer needs (I want you to to 8 channels in order to ensure the speed, I want you to to less expensive because sufficient 3 channel, etc.) can be easily realized.
They rather than sell what their have made, to make those customers covet, I am feeling.
(I am not good native English, please forgive Where difficult to read.)
>I.The SSD of easily customizable new structure the speed and capacity
So is asic based solition. You make a base configuration and fuse off features that are not needed. You dont need fpga for that
>II.First, put the target for video devices, and medical devices,etc.
I wouldnt put fpga based device with 52 independed emmc controllers in a medical device. Or any envirment, that needs high reliability. Remember, the more controllers you have, the bigger the risk of faulire (think of it as raid0).
>III.Was disassembled the Fixstars SSD-3000M. It will be 52 pieces equipped with eMMC chip of 64G bytes
So, it has some redundancy afterall. But question is, how does it work (can it sustain a failed emmc chip ?)
>IV. eMMC is equipped with a controller to manage the flash memory to itself. And SSD controller is not required.
Instead, read and write eMMC data using a FPGA, which exchanges with the external interface.
Which is never a good approach for a high performance device. eMMC is designed for emmbended devices such as mobile phones or other portable devices, not for ssds, where there will be lots of random data. Remember, emmc controllers are very simplictic (no dram, very simple wearlevelling and garabage collection algorithms) and are not designed for random envirment.
And having 52 of them means, that you have 52 controllers managing only one chip instead of one main chip controlling 52 packages. This has buch of downsides, one of them being inefficient wearlevelling (since every eemc controller is only aware of its own nand chip and nothing else)..
Even if you complain to me, I can not even how to.
Simply, the features of this product is so.
Also, this reminds me of another similar solution using the sledgehammer emmc approach (only sataII though)
https://www.sage-micro.com/chip-5.html