yeah, i've been holding off on upgrading from my existing 128 boot drive + mechanical setup to a full 1TB drive. I think I'll keep holding out until PCIe is better supported/better priced.
I see these are MLC drives. Hopefully, Samsung will come out with consumer-level 3D TLC drives that have relatively good reliability, and a price at or below Crucial's drives.
Reliability is a side effect of retreating to 40nm pitch technology. Even if they decided to do TLC with the same cells, they would probably end up being more reliable than 2D nand MLC.
Fantastic drives! That consistency is really remarkable. I don't know if most will pay the price premium for these over the EVO however. The average user probably wouldn't notice a difference in general day to day use.
Good read on the software not taking advantage of SSDs yet. Windows is the biggest offender. I have 8 threads and an SSD and I still have to wait for each of my startup programs to load like a snail 1 at a time after bootup...
"There are some drops, although I am not sure what is causing them"
It looks suspiciously like your values overflowed an unsigned int (prior to being converted from B to KB). Just add ~4.3 million to the 4 mysteriously low values and you have a nicely shaped curve.
Oops.. I pasted the wrong quote. Meant to quote this: "It looks like read performance scales quite linearly until hitting the IO size of 256KB where RAPID stops caching"
I maintain that there is an error in the numbers on the graph :)
if the values were being stored internally as bytes and not kilobytes it might overflow as tetsuo77 mentioned. 4.3 million * kilobytes per sec = billions of bytes/sec which could overflow.
Plus, the MX100 reliability is horrible. Just google MX100 BSOD, disappearing drive. I have 2x MX100 512GB SSDs and I recommend you don't buy one, no matter how cheap they are.
For business purposes, I would rather pay twice as much and get a 10 year warranty vs the 3 year supplied by Crucial. Though, for my daily, I would probably go with the Crucial.
I think there's a slight misunderstanding of manufacturing cost. While the die size may be the same, or even smaller than a competing technology, the 32 level chip does cost more to make per area. There are more masks, more layers, more etching and washing cycles, and more chance of defects.
Right now, I do see why the cost is higher. I can on,y assume that as this technology progresses, that cost will drop per area. But it will always remain higher than an SLC, MLC or TLC chip.
You are correct. I did mention yield and equipment cost in the final paragraph but I figured I won't go into detail about masks and etching since those would have required an in-depth explanation of how NAND is manufactured :)
It would be great if Anand or you do a writeup on 3d NAND & deal with the specific pros & cons of it as compared to traditional 2d NAND & if possible include something related to manufacturing processes of these & how they're different OR more/less expensive, certainly as in case of V-NAND?
You wouldn't need too much detail - just saying that the number of process steps increases by probably around an order of magnitude should make this pretty clear.
It is probably more than that, as Samsung is currently manufacturing 32 layers of cells. Each layer requires multiple operations (deposit, etching, washing...). Their biggest advantage comes from regressing to 40nm: at that technology, each operation is *MUCH* cheaper than the equivalent one at 1X pitch (15~19nm).
So, total cost is an unknown, but should be very competitive, after recovering the initial R&D investment.
And not to mention 3D NAND is still basically bleeding edge. It's still in the stages of where a new DDR generation arrives, much higher costs then current gen.
the 3D structure and design are revolutionary, but the manufacturing technology is actually a very mature one (40nm). This makes it *MUCH* cheaper than the 1X used by their competition.
Samsung has really struck gold with that design, as it allows them to scale in both dimensions, depending on the result and cost of each. While 2D NAND is facing really tough challenges to increase density, V-NAND is allowed to either scale up (more layers) or restart scaling pitch, as manufacturing is *very well* understood from 40nm->~16nm. They just need to experiment with it and see what makes economic sense and good trade-offs.
Silicon ingots cost is marginal. The real cost for scaling is all the R&D necessary to make the pitch smaller. Even using bigger wafers (current ones are 300mm, there have been talks of 450mm for a while, but cost is a deterrent, as a whole fab needs to be re-tooled for the upgrade) only improves yields and costs marginally. NAND scaling down is facing huge challenges, due both to process (who to image those ~15nm line on a wafer) and electrical limits (~3 electrons inside your cell at 15nm). 3D NAND allows to restart the growth by bypassing those challenges (step back to 40nm process and scale in the Z axis). General SOCs are facing similar process limits (there is no solution below 10nm so far, despite the whole industry cooperating to find one), even if their design limits are more relaxed (SOCs are not trapping charges, but cross talk and interference are starting to be challenges too).
"This further suggests that the issue lies in our tests instead of the RAPID software itself as end-users will always run the drive with a partition anyway."
Um, no. I don't care what the end user does, the software shouldn't cause a BSOD. If it can't cache without a partition, it should simply not attempt to cache. This is just a case of Samsung thinking that just because they do some nice hardware, that they're experts in software. They're really not. RAM caching of I/O isn't specific to SSDs anyway, why are they tying it to an SSD launch?
Yeah, that's a huge red flag for me, I definitely wouldn't use the RAPID software. Another layer to go wrong (and apparently it does). I hope Anandtech submits details of the BSODs they saw to Samsung.
Wait a minute, 150TB endurance on a 1TB drive? Only 150 cycles? That doesn't make any sense, that's absurdly low.
Then again, Intel's rating for the 335 doesn't make any sense either. They say 20GB a day for 3 years, or about 22TB... But they also rate it for 3000 cycles, and the media wear indicator on the drive is set to treat 3000 as full wear, and that represents 720TB...
The endurance figures are usually based on a 4KB random write workload and are thus worst-case numbers. 150TB of random writes means a ton of more NAND writes than 150TB, that's why. I explained the calculation of TBW here:
However, as I mentioned in the article, in the client space the endurance is more for guidance (i.e. don't put these in servers!) than an actual technical limit.
The above was meant to not be a reply to yours. In general, there's isn't much, though, because with SATA, we're still stuck needing striped RAID to improve bandwidth. Assuming they can keep costs in check, the rapid shrinking,m and thus capacity increases over the next few years, will give us something much more useful. That said, it's nice that they caught up in terms of consistency.
More relevantly, the 150TB warranty figure applies to all capacities. All things being equal, a 128GB drive with 150TB written will have the same amount of flash wear as a 1TB drive with 1200TB written.
So in practice the 1TB model will have far higher endurance than 150TB. But its warranty will expire when its flash is only 12.5% worn. Samsung aren't the only manufacturer to have low TBW warranty figures; Micron/Crucial does it as well, with the same TBW figure for all drive capacities.
thanks, the XP941 demolishes this thing, why isn't it in the charts of this review? clearly PCIe m.2 would be on the radar for anyone looking for a high end SSD these days... ?
it may not be readily available in the channels, but it is a taste of what is to come... personally I just got a 256GB plextor M6e PCIe m.2 for $220. It's easy to find in the market.
no, its 600MB/s in total in either direction. you can max the 600 read if you arent doing any writes, etc. It isnt 300 max per direction, its 600 max for total data bandwidth
No, it's really "600 MB/s whatever". In reality about 550 MB/s seems to be the maximum. With 300r + 300w full duplex neither read nor write could surpass 300 MB/s.
why I ask this I see a lot op people putting SSDs in RAID0, and I wonder why ... to me, it seems totally pointless the system doesn't do only writes or only reads, so what's the point? well, except in the cases where you copy/move some huge files from one matrix to another
Copying large files, and loading large files. Realistically, you can copy from a new SSD to another at anywhere from 250MBps to 450MBps (including OS/filesystem overhead). That's halved, or worse, copying from an SSD to itself. RAID 0 gives you an approximately linear bandwidth increase. If you don't need a logical or physical separation of data locations, RAID 0 beats a separate drive.
When I'm backing up my Bethesda games directory, I wish I had more speed, as it could be done faster in a RAID. However, I only do that one or two times per month (all my saves for all my games are on automatic backup).
For typical light random IO, it very much is pointless.
"I bet many of you would have liked to see the 850 Pro move to the PCIe interface but I understand Samsung's decision to hold on with PCIe for a little while longer."
I will test that as soon as I get back from Korea, as well as send you the Extreme Pro data. Sorry for not replying to your email earlier, I kind of lost focus on everything else when I started working on this review :)
You should actually put up a little pipeline article with all of this data, because it would be interesting to me, and I am sure lots of other users as well.
Yeah, the most important factor of a SATA SSD is the price, not how it will saturate the 550Mb limit in a scenario that will never happen for most consumers. Now even a mention on the first page, eh? Please put a little pressure for the manufacturers to move on to M.2 and NVME.
**In the Midnight hour, She cried more, more, more!!**
Your EVO has worse endurance, lower IOPS, lower consistency and most of all lower Write speeds (495MB/s write is due to trick called turbowrite). Sequential speeds are only a part of the story.
I always wonder who this kind of drive is supposed to be aimed at. Yes, it's fast; but the competition is fast enough that I really can't imagine anyone *noticing* the difference outside of artificial extremely heavy non-stop I/O. And I don't mean light-load non-stop, I mean I/O's maxed out non-stop so the drives can't take time to garbage collect.
So... a power user is unlikely to notice much difference outside of short bursts of high-I/O apps, and as the rather heavy 2011 light load demonstrates, the drives are already maxing out there too. I just can't think of a real-world load where a human being would notice the performance difference and care about it; that would imply a performance difference of around a factor 2.
So we're left with a super fast drive (good for bragging rights, but what else?), but some apparently intentionally missing features like power-loss protection. Why would even a heavy user choose this over, say, samsungs own 840 EVO, or crucial's MX100/M500?
I just don't see the value here. To me this looks like microoptimization and losing sight of the bigger picture.
Price matters. Features matter. Performance - only until you're fast enough.
Kinga agree with this. If you REALLY need every bit of performance, you won't be looking for SATA drives at all. This probobly has a nieche. But mainstream drives are really fast enough these days for most people.
I second this. The Anandtech SSD tests were designed so that we could tell the difference between drives that are all so fast - that there is no way to tell them apart in ordinary usage scenarios. I see the value of testing the theoretical performance of drives as manufacturers push the technological limits.
That said, at the end of the day user-experience is what matters. I agree with emn13 that the "light workload" test is already more strenuous than anything the average user is likely to do, and looking at the chart, we see that almost every drive is within a range of ~280 to ~380 MB/s. I'm guessing that the range in performance gets even narrower for "real world" workloads.
So keep up the innovative SSD testing, but be sure to put these theoretical performance gains into a real-world context when you get to the Conclusions section of these articles. Not everyone will benefit from these theoretical increases in performance.
Is Samsung planning on doing TLC based V-NAND anytime soon ? It would be great for a mainstream drive, since endurance would be higher (due to older node), speeds would probobly also went up (so no need for gimicks like turbowrite). Or is it not mature enough to scale down to TLC ?
You had me at 10 years warranty. I don't mind the slight premium if I'm not buying another one midway through the cycle. Sure, it will be obsolete well before it dies, but that term signals Samsung is really confident about their reliability.
I must have missed this in the article - are these V.NAND cells as used in 850 Pro drives 2 or 3 bits per cell ? I got the "larger lithography improves endurance" part, I'm just wondering whether they opted for more conservative option (MLC) there as well.
Hi, one question: In the review the idle power consumption for e.g. the 850 128gig is 35 mw. I wanted to compare that to my Samsung 470 so I went to Bench and selected the drives for comparison. There it says that the 850 uses 0.29 Watt. So how comes there is a difference?
Anandtech Bench has four SSD power numbers: SSD Slumber Power (HIPM+DIPM) Drive Power Consumption - Idle Drive Power Consumption - Sequential Write Drive Power Consumption - Random Write
The confusing things are that (1) the review only listed slumber power, not idle power, and (2) Bench lists both numbers but doesn't place the slumber power next to the other power values.
I also find the lack of powerloss protection being a big negative over this hard drive. Until REFS has all the features it needs in Windows that you would get in Linux this is going to be an important feature for anyone who values data integrity. Even after that happens it still might be very important.
The SSD just quitting could brick the drive, and could hose up the FS. Like ECC RAM, if you need power loss protection, you need it regardless of file system. IMO, they should all at least be equipped with enough to gracefully finish in-progress writes and shut down (not necessarily empty buffers, but set up a state that they can be guaranteed to be able to roll back from).
Is it just me, or does the "Performance Consistency" page have interactive charts, all of which are entitled "(SSD name and capacity) - 4KB Random Write (QD32) Performance", but with different results in each of the graphs for the same SSDs and capacities? Also, descriptive text is missing below the last two.
I'm presuming they are supposed to be a read and another read or write one, presumably at a lower queue depth.
I purchased a 512GB MX100 when they first came out for $199. No way the extra speed from the 850 pro is worth over twice the price as the MX100. Nobody is going to keep that drive for 10 years anyway. In 10 years the drive will be obsolete. Most users won't even notice the difference in speed in their every day usage.
As I mentioned elsewhere, it is probably a STEM bright field image after a FIB liftout of the gate stack. Certainly not an X-ray; he should be correct the article...
As soon as it is cheap enough. But, don't get your hopes up about performance. SD cards are mostly limited by the controllers being slow, and in the tiny package they fit in, with the narrow margins they have, there's not a lot of room, physically and economically, to give them fast controllers, even if you get a big one that must have several NAND dies, and are talking about full-size SD, where multiple channels might be viable. It sucks, and I dislike shopping for SD cards as much as anybody, but today, that's how it is.
I think he was talking about V-NAND (3D cells) which is independent of the controller. I would guess it will, as density will continue to scale up which will make it the cheaper technology. It is cutting edge now, but will let Samsung scale higher densities very aggressively in the coming years, replacing all their 2D NAND production (they announced it when presenting the 3D cells).
Personally I have no interest in this kind of performance, and I really hope they focus on reducing prices and increasing capacities. The MX100 is just great for home usage (system and gaming), and I would like to see a 512 GB equivalent for around 100 $ by the end of 2015.
"Oftentimes when cell size is discussed, it is only the actual size of the cell that is taken into account, which leaves the distance between cells out of the conclusion."
Incorrect. Oftentimes what is being discussed is the half pitch. The 16nm, 19nm, 20nm, etc of the die. That is not the cell. The cell is Always defined as the repeatable structure in a memory device, and this includes the space between cells as described. The cell size is incorrectly referenced as being the half pitch.
Then there is marketing gimmick by companies who call their products 19nm when it is really 19nm by 2xnm. A rectangle and not a true 19nm square half pitch.
I never paid much attention, but you're right. If they changed the point shapes, and maybe dashed a couple of the lines, they could take care of that easily.
it's great to see a new drive from samsung and even greater seeing them advancing ssd tech and performance in such substantial ways. keeping that in mind i'm not really surprised about the msrp sammy is asking for its drives. and as always when new devices hit the scene, we're comparing msrp with real market prices here, so the difference should be a bit lower in a couple weeks when enough stock is available.
that said, even if sata3 remains the most important storage interface today, it's kind of a shame seeing such a beautiful drive limited by this "old" interface. i know the new standards like m2, sata3.2 and pci-e-drives are still kind of a mess, but we already saw what higher throughputs in combination with more efficient interface protocols can do and seeing an expensive enthusiast drive like the 850 pro connected to sata3 just makes it seem more limited than it needed to be.
all that said, it doesn't change much for the average user, or advanced users even, since for most people a good sized evo or crucial is all they ever need in the years to come. upgrading to expensive drives like the 850 will only make sense for the most demanding users, for the rest it will only get interesting again when pci based storage gets more affordable.
Minor nit: There's no such thing as "pentalobe torx" it's either one or the other but I'm guessing that it might have been torx security since pentalobe screws have only been used by Apple a couple of years back.
Its great to see its doing well in power consumption area. Which is important in Notebook. I hope we could bring this down to 2W or even 1.5W during operation.
I really do think our SSD storage tier deserve a PCI-E lane direct from CPU. It would be great if the market just settle on 2x PCI-E 3.0 from CPU. We get 2GB/s out of it. That is plenty of headroom to grow until we move to PCI-E 4.0
Well, kinda, I mean, some implementations come from the IOH instead of the CPU. I have heard rumors that future versions of Intel Desktop CPU's will have 20-24 PCIE lanes on them instead of 16. That would be perfect for storage!
"That is very aggressive because it essentially implies that the die capacity will double every year (256Gbit next year, 512Gbit in 2016 and finally 1Tbit in 2017)" - Does this mean a 8TB drive at $600 in 2017 ($75 per TB). Good bye hard drives by then :). What about a 850/850 Evo version?
It certainly does imply 8TB SSDs by 2017. By that point such a SSD will likely have a higher capacity than HDDs of that time.
Of course, price will be very significantly in favour of the HDD still; but if money is no object you could do bulk storage in a portable device if you wanted to.
thanks for the article on VNAND SSD. I think the SSD analysis is good and shows the impact. The details of Planar NAND and VNAND are incorrect in many cases. The overall NAND takeaway should be Samsung VNAND is a 86Gbit device Die level with a very large effective cell size. I still want to buy one... where can I get it?
@extide: so you measured the die to be 67mm^2? how was this measured ? CSAM? XRAY? that seems odd to use such a small dies for SSD. and that would make it even smaller cell size than the one at ISSCC since it was 134 for a 128Gbit with 24 layers (periphery doesnt shrink as fast with lower density)
I'm not claiming that it is an accurate figure, hence the "~" sign in front of it. However, Samsung wouldn't disclose the die size when I asked them during the Q&A, so at this point I don't know for sure. However, I have a picture of the 32-layer wafer and once I get back home I'll do the math of the wafer to figure out the exact die size.
The micrograph you say is an "x-ray", is almost certainly a transmission electron microscopy image. Given the oxides are light it is a bright field image. It could technically be a SEM image but the resolution is a bit to high, so it most likely is a STEM image with a bright field detector.
Perhaps time should be added as a cost efficiency factor. Presumably the die has to stay in the process much longer due to all the layers being added one at a time.
If only this drive would of been a pci-e 3.0 x4 interface with nvme. We would of finally had a worthy upgrade. Yes it's performance is better than the 840 evo but you can get the 840 evo 1tb for 400 dollars even less if you catch it on sale. So you can get 2x 1tb 840 evo for almost the same price as 1 850 pro. If you compare an 840 evo with 25% OP to a regular 7% on the 850 the 840 evo looks just as good so 2x 825GB drives with 25% OP on each drive costs you 750-800 depending on the deal you get vs 700 for 1000GB. I would rather pay 800 for 1650GB than 700 for 1000GB with performance being nearly identical. I get an extra 650GB (which at 50 cents per GB is another 325 dollars worth of ssd) and tons of over provisioning to give the drive equal or better performance for only 100 dollars more (possibly less as the 840 evo has often sales)
Sorry Samsung but 2x 1tb 840 evos with 25% provisioning gives me better or equal performance and a whole 650GB of extra storage for only 100 dollars more. At 50 cents per GB you get 325 dollars worth of more storage capacity for only 100 dollars more and thats with the 25% over provisioning which basically negates the performance increase of the new drives.
The only way samsung could of made this drive worth that money is if they had the drive on a pci-e 3.0 x4 interface with nvme instructions. I'm sure there will be tons of idiots who just buy it cause it's the latest drive. But if you use your brain you can see the 840 evo is still the best SATA drive when the cost/performance ratio is taken into account. 2x 1tb evo's in 25% OP mode gives you 1650GB and costs u 100 dollars extra or less and gives you the same performance or even better compared to 7% 850 pro 1000GB especially when raid 0 is taken into account. I'll take 1650GB over 1000GB if it's only 100 more and performance is equal or better easy choice.
Since samsung is stacking vertically now what they should of done is made a super huge 2tb drive to differentiate themselves from all these other drives. A 2tb drive for 1400 is a little more acceptable than 1tb for 700 simply because it's the only single drive with 2TB capacity.
I can see great things happening with vertical stacked nand tho. When this process matures we should be seeing nand drives surpass spinning hard drives in capacity. When samsung has those 1tbit dies its planning for 2017 we should be able to have 8-12TB SSD's
It's a new product and it's priced highly. Eager early adopters who want to move on to the latest and greatest will buy now, value-concious people will buy the 840s. Inventory of the 840s will get eaten up and the 850s will drop in price.
This is what happens whenever any product is released, basically ever.
Sorry, but neither this nor the EVO will get my money. Performance is all very well, even if only a tiny handful of users with extremely niche workloads will ever notice any difference, but all this extra reliability at a price premium means NOTHING without simple power loss protection. Restricting something so basic to "Enterprise" class products is the real gouging here by Samsung, not the price.
If Crucial can provide that protection on the CHEAPEST drives in their class (M500/M550 and IIRC MX100 too) with performance that is not gimped as a balance, then there is no excuse for Samsung not to. This should be a no-compromise baseline for all SSDs going forward, and Anandtech should push hard for that - users should as well, by voting with their wallets and refusing to buy drives, however fast and powerful, that do not provide power loss protection as a basic feature.
Though I agree it is a highly useful feature, it is far less significant to those of us using a good UPS. So I can't agree that it's a no compromise feature.
What you aren't taking into account is the fact that the 850 Pro has MUCH higher endurance, and it's also more consistent. Those two items bring it more into the Intel DC 3500/3700 series type of class. It's not just a sheer performance comparison.
...and if you thing the 850 Pro is expensive, price out the larger Intel DC S3700 drives.
"Real" Enterprise SSDs don't even have an MSRP. You negotiate with the vendor, and hope for the best. Now that Texas Memory is in IBM, and Fusion-io in SanDisk, with Violin likely to go somewhere. The conundrum is V.NAND's impact on flattening the curve between Enterprise and Commodity/Retail. At one time, a mainframe had bespoke 14" behemoth hard drive subsystems (in the case of IBM, run by the equivalent of a PDP-x). In due time, binned commodity 3.5" drives are now used.
Samsung could well be the driving force to regularize solid state storage. The remaining issue is whether the file system hand waving will be dumped in favor of direct NVM persistence? Samsung, or whoever, likely couldn't care less.
It's been enjoyable to see a lot of "new" flash memory storage vendors pop-up. More competition is always good in that enterprise space.
I've been looking forward to having more SSDs options available to servers at much better pricing. Solid storage advances have a trickle-down effect. If I can put an array of these Samsung 850 Pros into a server and achieve near "enterprise" performance, then that forces Dell/HP/etc. to drop their own SSD pricing.
Well, I was certainly able to upgrade to Magician 4.4, and it did take a couple reboots while it was re-activating RAPID. I am going to guess and say yes it seems to be the new version of RAPID, but I don't see anywhere that it actually says RAPID 2.0 or any indication of how much memory it has available.
Ah ok. When I select the 470 in Bench there is no slumber so I got confused.
Since power cosumption is a big thing in Mobile could you do the following: Bench the impact on the Battery of a Laptop due to a fast drive. E.g. drive A is fast but has the downside that it draws more under load as drive B which is slower. But since its done faster it returns to idle faster. I mean impact on real world scenarios. Or how much power was used for bench x etc. and have a average per hour or something like that.
Prices are not going down, good thing we have Crucial who have best bang for the buck, ofcourse performance wise is not compared to sandisk or samsung, but its still a very fast SSD, for normal users and gamers, Mx100 is the best drive you can get for its price.
Thanks Kristian for such an amazing, in depth review. I especially loved the detailed explanation of current 2D NAND vs 3D NAND, how it all works, and why it's all so important. Possibly one of my favourite Anandtech articles to date!
I don't see this mentioned anywhere, but were the tests performed with RAPID enabled or disabled? I understand that some of the tests could not run with RAPID enabled, but for those other tests which do run on a formatted partition (i.e. not run on the raw disk), its not clear if RAPID is enabled or disabled. Therefore its not clear how RAPID will affect the results in each test.
I always wondered if there is any practical and notable difference between dynamic and static over-provisioning. I mean... since TRIM should blank out the empty LBAs anyway, I don't see the point in leaving unpartitioned space for static over-provisioning for home users. From a general user standpoint, having as much usable space available as possible (even if we try to restrict ourself from ever utilizing it all) seems to be a lot more practical (until it's actually usable with an acceptable speed, so even if notably slower but still fast enough...) than keeping a (significantly more, but still not perfectly) constant random write performance.
So, I always create a system partition as big as possibly (I do the partitioning manually: a minimal size EFI boot partition + everything else at one piece) without leaving unpartitioned space for over-provisioning and I try to leave as much space empty as possible.
However, one time, after I filled my 840 Pro up to ~95% and I kept it like that for 1-2 days, it never "recovered" . Even after I manually ran "defrag c: /O" to make sure the freed up space is TRIMed, sequential write speeds were really slow and random write speeds were awful. I ha to create a backup image with DD, fill the drive with zeros a few times and finally run an ATA Secure Erase before restoring the backup image.
Even though I was never gentle with the drive (I don't do stupid things like disabling swapping and caching just to reduce it's wear, I bought it to use it...) and I did something which is not recommended (filled almost all the user-accessible space with data and kept using it like that for a few days as a system disk), this wasn't something I expected from this SSD. (Even though this is what I usually get from Samsung. It always looks really nice but later on something turns out which reduces it's value/price from good or best to average or worse.) This was supposed to be a "Pro" version.
I don't normally go out of my way to comment on a product but I felt this product deserved the effort. I've been using personal computer since personal computers first came out. I fully expected my upgrade from an old 50gig SSD to be a nightmare.
I installed the new 500gig Evo 850 as a secondary, cloned, switch it to primary and had it booting in about 15 minutes. No problems, no issues, super fast, WOW. Glad Samsung got it figured out. I'll be a lot less concerned my next upgrade and won't be waiting until I'm at my last few megabytes before upgrading again.
I must disagree with the conclusion, "there is not a single thing missing in the 850 Pro". Power-loss protection is a *huge* omission, especially for a "Pro" product.
I bought the 850 Pro 500 GB version. The transfer speeds are around (and just under) 400 MB/s and the IOPS are always around 55-60k. Not impressed - considering returning it just due to the huge gap between the rated specs and actual performance. (Of course Samsung Magician over-reports the transfer speed by quite a large margin).
Someone just tipped me that V-NAND in Samsung 850 Pro is actually TLC not MLC. The first reason is the 86Gbit/die number is odd, if 850 Pro uses a TLC 128Gbit/die and emulates it to MLC then it would be 86Gbit/die, the numbers fit well. The second reason is Chipworks confirmed it in the die shots: http://bit.ly/YTVm9Z http://bit.ly/1uByKcm
I'm just the messenger here, I'm not expert. So...I was wodering is it possible to hack Samsung 850 Pro to get extra storage space in TLC?
Why isn't the lack of power loss protection a showstopper for any of these drives? I love the speed and reliability benefits under normal operation, but how can anyone get excited about a drive that can get corrupted in an instant due to power loss or computer freeze, where a hard shutdown is required? Seems that these drives are more prone to massive data issues much more than HDDs under the same power fault conditions. I keep looking for a good answer, but it seems smart people are willing to look past this seemingly fatal Achilles heal. Not sure what I might be missing. Any thoughts?
Dear Kristian Vättö, Fantastic review of the drives and the technology behind it. This is what readers like me expect to see on Anandtech. Thanks for breaking it down. Keep up the good work!
Please update the endurance to 300TBW for 1TB model. This was my basis for buying this SSD. This is 2nd to Crucial MX200 (320TBW on 1TB model) in terms of endurance - for client/consumer SSDs.
Samsung is using 40nm over Micron's 16nm. Doing the math implies that for every 2.5 16nm Micron nodes in planar section, Samsung only does 1 40nm node. If we scale vertically to 32 layers, then Micron (or others for that matter) still do only 2.5 nodes for every 32 nodes of Samsung. Which means for every 16nm node, Samsung provides 12.8 nodes. Meaning the density scaling factor for every 32 layer increments is a multiple of 12.8. Assuming the current die size for 32 layers to be 128Gbit, then the density advantage for 256Gbit is 12.8x2 times, 512Gbit is 12.8x4 times and for 1Tbit in 2017 should be 12.8x8 times for 256 layers of nodes stacked on top of each other. So the density advantage is approximately 102.4% (theoretically). Samsung can theoretically produce a 1Tbit die at a cost advantage/space advantage of nearly 100 times compared to planar and manufacturers. It is almost like you are able to earn 100 times the profit for the same die provided the cost per bit is scaled along without passing on the price advantage to the end users until other players enter 3D market.
The same is the case with Intel's 3D Cross Point Technology. They are having a new tech in their hands that is faster than NAND and closer to DRAM. So they are also planning to price it exactly between NAND and DRAM. We are already paying high costs to shift from platters to NAND and are going to pay even more to make a shift from NAND to 3D xPoint. I just wish I can jump a few years to the future, grab a high capacity NAND/xPoint drive for cheap and come back to the present and use it. Sigh !
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
160 Comments
Back to Article
Homeles - Monday, June 30, 2014 - link
Man, the 850 Pro is killer. Samsung really knocked it out of the park. Those prices are just completely out of touch, though.Awful - Monday, June 30, 2014 - link
Yeah the prices are high for now. Game changing stuff though; and prices can only come down. PCIe V-NAND? Yes please!Hung_Low - Tuesday, July 1, 2014 - link
Even better, Intel style NVMe controller + this v-nand!! orgasmicavyshue - Tuesday, July 1, 2014 - link
yeah, i've been holding off on upgrading from my existing 128 boot drive + mechanical setup to a full 1TB drive. I think I'll keep holding out until PCIe is better supported/better priced.Angrychair - Tuesday, July 1, 2014 - link
The prices are in line for class leading performance and reliability.The reliability is the critical part, these are drives that are unlikely to wear out in any system almost no matter how heavily taxed.
Ken_g6 - Tuesday, July 1, 2014 - link
I see these are MLC drives. Hopefully, Samsung will come out with consumer-level 3D TLC drives that have relatively good reliability, and a price at or below Crucial's drives.frenchy_2001 - Tuesday, July 1, 2014 - link
Reliability is a side effect of retreating to 40nm pitch technology.Even if they decided to do TLC with the same cells, they would probably end up being more reliable than 2D nand MLC.
joelypolly - Thursday, July 3, 2014 - link
They already have drives in testing that has written over 8 petabytes so I think the reliability is theremkozakewich - Saturday, July 5, 2014 - link
On that note, I'm wondering how 4-bit MLC would perform compared to 2D NAND.RaistlinZ - Monday, June 30, 2014 - link
Fantastic drives! That consistency is really remarkable. I don't know if most will pay the price premium for these over the EVO however. The average user probably wouldn't notice a difference in general day to day use.alacard - Monday, June 30, 2014 - link
Fascinating stuff, thanks for the in depth analysis.Iketh - Tuesday, July 1, 2014 - link
Good read on the software not taking advantage of SSDs yet. Windows is the biggest offender. I have 8 threads and an SSD and I still have to wait for each of my startup programs to load like a snail 1 at a time after bootup...tetsuo77 - Monday, June 30, 2014 - link
"There are some drops, although I am not sure what is causing them"It looks suspiciously like your values overflowed an unsigned int (prior to being converted from B to KB). Just add ~4.3 million to the 4 mysteriously low values and you have a nicely shaped curve.
tetsuo77 - Monday, June 30, 2014 - link
Oops.. I pasted the wrong quote. Meant to quote this: "It looks like read performance scales quite linearly until hitting the IO size of 256KB where RAPID stops caching"I maintain that there is an error in the numbers on the graph :)
Gigaplex - Tuesday, July 1, 2014 - link
32 bit unsigned integers support around 4.3 billion, not million.lyeoh - Friday, July 4, 2014 - link
if the values were being stored internally as bytes and not kilobytes it might overflow as tetsuo77 mentioned. 4.3 million * kilobytes per sec = billions of bytes/sec which could overflow.nirwander - Monday, June 30, 2014 - link
I cant see how they aim at mainstream with these prices.Crucial MX100 512 is already fast enough for SATA 6 Gbps and.. twice as cheap!
Technology geeks will probaly go for Intel PCIe NVMe drives.
Gigaplex - Monday, June 30, 2014 - link
And if you really need the performance, just get two of the MX100s and RAID 0 them.willis936 - Tuesday, July 1, 2014 - link
Unless you care about storage latency at all.Gigaplex - Tuesday, July 1, 2014 - link
Fair point, but SSDs are so far ahead of hard drives in terms of latency that it hardly matters.Squuiid - Saturday, March 14, 2015 - link
Plus, the MX100 reliability is horrible. Just google MX100 BSOD, disappearing drive.I have 2x MX100 512GB SSDs and I recommend you don't buy one, no matter how cheap they are.
nightauthor - Tuesday, July 1, 2014 - link
For business purposes, I would rather pay twice as much and get a 10 year warranty vs the 3 year supplied by Crucial. Though, for my daily, I would probably go with the Crucial.TheWrongChristian - Wednesday, July 2, 2014 - link
No current SATA drives push low queue depth random IOs to the point of saturating SATA II, let alone SATA III.At high queue depths, perhaps. But then, that is not a typical workload for most users, desktop or server.
Plus, it's a new drive, prices will come down.
jwcalla - Monday, June 30, 2014 - link
Unless they're doing 5% OP the capacities are kinda... off.melgross - Monday, June 30, 2014 - link
I think there's a slight misunderstanding of manufacturing cost. While the die size may be the same, or even smaller than a competing technology, the 32 level chip does cost more to make per area. There are more masks, more layers, more etching and washing cycles, and more chance of defects.Right now, I do see why the cost is higher. I can on,y assume that as this technology progresses, that cost will drop per area. But it will always remain higher than an SLC, MLC or TLC chip.
So there is a balance here.
Kristian Vättö - Tuesday, July 1, 2014 - link
You are correct. I did mention yield and equipment cost in the final paragraph but I figured I won't go into detail about masks and etching since those would have required an in-depth explanation of how NAND is manufactured :)R0H1T - Tuesday, July 1, 2014 - link
It would be great if Anand or you do a writeup on 3d NAND & deal with the specific pros & cons of it as compared to traditional 2d NAND & if possible include something related to manufacturing processes of these & how they're different OR more/less expensive, certainly as in case of V-NAND?MrSpadge - Tuesday, July 1, 2014 - link
You wouldn't need too much detail - just saying that the number of process steps increases by probably around an order of magnitude should make this pretty clear.frenchy_2001 - Tuesday, July 1, 2014 - link
It is probably more than that, as Samsung is currently manufacturing 32 layers of cells. Each layer requires multiple operations (deposit, etching, washing...). Their biggest advantage comes from regressing to 40nm: at that technology, each operation is *MUCH* cheaper than the equivalent one at 1X pitch (15~19nm).So, total cost is an unknown, but should be very competitive, after recovering the initial R&D investment.
Spatty - Tuesday, July 1, 2014 - link
And not to mention 3D NAND is still basically bleeding edge. It's still in the stages of where a new DDR generation arrives, much higher costs then current gen.frenchy_2001 - Tuesday, July 1, 2014 - link
the 3D structure and design are revolutionary, but the manufacturing technology is actually a very mature one (40nm). This makes it *MUCH* cheaper than the 1X used by their competition.Samsung has really struck gold with that design, as it allows them to scale in both dimensions, depending on the result and cost of each. While 2D NAND is facing really tough challenges to increase density, V-NAND is allowed to either scale up (more layers) or restart scaling pitch, as manufacturing is *very well* understood from 40nm->~16nm. They just need to experiment with it and see what makes economic sense and good trade-offs.
toyotabedzrock - Tuesday, July 1, 2014 - link
Almost seems like it would be cheaper to ramp up the production of silicon ingots and drive that cost down further than the r&d for this.frenchy_2001 - Tuesday, July 1, 2014 - link
Silicon ingots cost is marginal. The real cost for scaling is all the R&D necessary to make the pitch smaller. Even using bigger wafers (current ones are 300mm, there have been talks of 450mm for a while, but cost is a deterrent, as a whole fab needs to be re-tooled for the upgrade) only improves yields and costs marginally.NAND scaling down is facing huge challenges, due both to process (who to image those ~15nm line on a wafer) and electrical limits (~3 electrons inside your cell at 15nm). 3D NAND allows to restart the growth by bypassing those challenges (step back to 40nm process and scale in the Z axis).
General SOCs are facing similar process limits (there is no solution below 10nm so far, despite the whole industry cooperating to find one), even if their design limits are more relaxed (SOCs are not trapping charges, but cross talk and interference are starting to be challenges too).
UltraWide - Monday, June 30, 2014 - link
Will there be a version with PCIe or M.2?Gigaplex - Monday, June 30, 2014 - link
If you'd read the article, you'd know the controller doesn't support PCIe.Gigaplex - Monday, June 30, 2014 - link
"This further suggests that the issue lies in our tests instead of the RAPID software itself as end-users will always run the drive with a partition anyway."Um, no. I don't care what the end user does, the software shouldn't cause a BSOD. If it can't cache without a partition, it should simply not attempt to cache. This is just a case of Samsung thinking that just because they do some nice hardware, that they're experts in software. They're really not. RAM caching of I/O isn't specific to SSDs anyway, why are they tying it to an SSD launch?
Donuts123 - Wednesday, July 2, 2014 - link
Yeah, that's a huge red flag for me, I definitely wouldn't use the RAPID software. Another layer to go wrong (and apparently it does). I hope Anandtech submits details of the BSODs they saw to Samsung.RAPID probably just uses the Samsung SSD as a dongle. Presumably RAPID is derived from Samsung's acquisition of NVELO, see http://www.anandtech.com/show/6518/samsung-acquire...
Guspaz - Monday, June 30, 2014 - link
Wait a minute, 150TB endurance on a 1TB drive? Only 150 cycles? That doesn't make any sense, that's absurdly low.Then again, Intel's rating for the 335 doesn't make any sense either. They say 20GB a day for 3 years, or about 22TB... But they also rate it for 3000 cycles, and the media wear indicator on the drive is set to treat 3000 as full wear, and that represents 720TB...
Kristian Vättö - Tuesday, July 1, 2014 - link
The endurance figures are usually based on a 4KB random write workload and are thus worst-case numbers. 150TB of random writes means a ton of more NAND writes than 150TB, that's why. I explained the calculation of TBW here:http://www.anandtech.com/show/7947/micron-m500-dc-...
However, as I mentioned in the article, in the client space the endurance is more for guidance (i.e. don't put these in servers!) than an actual technical limit.
emn13 - Tuesday, July 1, 2014 - link
...but outside of server-like workloads, what's going to benefit from this performance?Cerb - Tuesday, July 1, 2014 - link
"cheese in in" -> "cheese in it"?Cerb - Tuesday, July 1, 2014 - link
The above was meant to not be a reply to yours. In general, there's isn't much, though, because with SATA, we're still stuck needing striped RAID to improve bandwidth. Assuming they can keep costs in check, the rapid shrinking,m and thus capacity increases over the next few years, will give us something much more useful. That said, it's nice that they caught up in terms of consistency.Donuts123 - Sunday, July 6, 2014 - link
More relevantly, the 150TB warranty figure applies to all capacities. All things being equal, a 128GB drive with 150TB written will have the same amount of flash wear as a 1TB drive with 1200TB written.So in practice the 1TB model will have far higher endurance than 150TB. But its warranty will expire when its flash is only 12.5% worn. Samsung aren't the only manufacturer to have low TBW warranty figures; Micron/Crucial does it as well, with the same TBW figure for all drive capacities.
8steve8 - Tuesday, July 1, 2014 - link
Yawn. Another SATA 2.5" SSD.How does this compare to m.2 PCIe drives like the XP941 and M6e?
Will we see m.2 PCIe NVMe drives before xmas?
Kristian Vättö - Tuesday, July 1, 2014 - link
The 850 Pro is now in the Bench so you can compare the two easily:http://www.anandtech.com/bench/product/1201?vs=125...
8steve8 - Tuesday, July 1, 2014 - link
thanks, the XP941 demolishes this thing, why isn't it in the charts of this review? clearly PCIe m.2 would be on the radar for anyone looking for a high end SSD these days... ?it may not be readily available in the channels, but it is a taste of what is to come... personally I just got a 256GB plextor M6e PCIe m.2 for $220. It's easy to find in the market.
Kristian Vättö - Tuesday, July 1, 2014 - link
I didn't want to put the XP941 in the graphs because the availability and support are relatively limited and it would have broken some of the graphs.zmeul - Tuesday, July 1, 2014 - link
just a quick dumb question:the bandwidth for SATA 3.0 - 6 Gbit/s is 600 MB/s right? so, it's 300MB/s read and 300MB/s write
correct ?
thesavvymage - Tuesday, July 1, 2014 - link
no, its 600MB/s in total in either direction. you can max the 600 read if you arent doing any writes, etc. It isnt 300 max per direction, its 600 max for total data bandwidthzmeul - Tuesday, July 1, 2014 - link
yes, I forgot to mention, duplex (at the same time) 300r+300wMrSpadge - Tuesday, July 1, 2014 - link
No, it's really "600 MB/s whatever". In reality about 550 MB/s seems to be the maximum. With 300r + 300w full duplex neither read nor write could surpass 300 MB/s.zmeul - Tuesday, July 1, 2014 - link
why I ask thisI see a lot op people putting SSDs in RAID0, and I wonder why ... to me, it seems totally pointless
the system doesn't do only writes or only reads, so what's the point? well, except in the cases where you copy/move some huge files from one matrix to another
Cerb - Tuesday, July 1, 2014 - link
Copying large files, and loading large files. Realistically, you can copy from a new SSD to another at anywhere from 250MBps to 450MBps (including OS/filesystem overhead). That's halved, or worse, copying from an SSD to itself. RAID 0 gives you an approximately linear bandwidth increase. If you don't need a logical or physical separation of data locations, RAID 0 beats a separate drive.When I'm backing up my Bethesda games directory, I wish I had more speed, as it could be done faster in a RAID. However, I only do that one or two times per month (all my saves for all my games are on automatic backup).
For typical light random IO, it very much is pointless.
457R4LDR34DKN07 - Wednesday, July 2, 2014 - link
Easy, say you have 2 sata 2.0 SSD but want sata 3.0 speeds.Kevin G - Tuesday, July 1, 2014 - link
Typo?"I bet many of you would have liked to see the 850 Pro move to the PCIe interface but I understand Samsung's decision to hold on with PCIe for a little while longer."
I think the second PCIe should be SATA.
stickmansam - Tuesday, July 1, 2014 - link
of the "on" could be a "off'Solid State Brain - Tuesday, July 1, 2014 - link
How many P/E cycles does the NAND on this drive support, by SMART data?Kristian Vättö - Tuesday, July 1, 2014 - link
I will test that as soon as I get back from Korea, as well as send you the Extreme Pro data. Sorry for not replying to your email earlier, I kind of lost focus on everything else when I started working on this review :)Solid State Brain - Tuesday, July 1, 2014 - link
No problem, thanks!extide - Tuesday, July 1, 2014 - link
You should actually put up a little pipeline article with all of this data, because it would be interesting to me, and I am sure lots of other users as well.beginner99 - Tuesday, July 1, 2014 - link
Sounds great until you see the price. Even for an enthusiast a crucial MX100 is probably the more reasonable choice. It's half the price...juhatus - Tuesday, July 1, 2014 - link
Yeah, the most important factor of a SATA SSD is the price, not how it will saturate the 550Mb limit in a scenario that will never happen for most consumers. Now even a mention on the first page, eh? Please put a little pressure for the manufacturers to move on to M.2 and NVME.**In the Midnight hour, She cried more, more, more!!**
boogerlad - Tuesday, July 1, 2014 - link
Are you guys going to review the SM1715? Really curious to see how it stacks up against the Intel, especially in client workloads.pesho00 - Tuesday, July 1, 2014 - link
I realy hoped to see 2,3,4 TB drives with this technology :(But we will weith ;)
Nice drive, not so nice price ;)
MrSpadge - Tuesday, July 1, 2014 - link
Not much point in offering them now if you consider the price of the 1 TB version.Samus - Tuesday, July 1, 2014 - link
Those Koreans...they're just killer engineers.trumanhw - Tuesday, July 1, 2014 - link
I don't get it. I get about 505 read and 495 write on Evo. This goes up by TEN percent in performance and then all but DOUBLES the price?I'm confident someone here has understanding of this I'm missing--please reply and just point me to the parts that change my view.
Thanks
hojnikb - Tuesday, July 1, 2014 - link
Your EVO has worse endurance, lower IOPS, lower consistency and most of all lower Write speeds (495MB/s write is due to trick called turbowrite).Sequential speeds are only a part of the story.
emn13 - Tuesday, July 1, 2014 - link
I always wonder who this kind of drive is supposed to be aimed at. Yes, it's fast; but the competition is fast enough that I really can't imagine anyone *noticing* the difference outside of artificial extremely heavy non-stop I/O. And I don't mean light-load non-stop, I mean I/O's maxed out non-stop so the drives can't take time to garbage collect.So... a power user is unlikely to notice much difference outside of short bursts of high-I/O apps, and as the rather heavy 2011 light load demonstrates, the drives are already maxing out there too. I just can't think of a real-world load where a human being would notice the performance difference and care about it; that would imply a performance difference of around a factor 2.
So we're left with a super fast drive (good for bragging rights, but what else?), but some apparently intentionally missing features like power-loss protection. Why would even a heavy user choose this over, say, samsungs own 840 EVO, or crucial's MX100/M500?
I just don't see the value here. To me this looks like microoptimization and losing sight of the bigger picture.
Price matters. Features matter. Performance - only until you're fast enough.
hojnikb - Tuesday, July 1, 2014 - link
Kinga agree with this. If you REALLY need every bit of performance, you won't be looking for SATA drives at all.This probobly has a nieche.
But mainstream drives are really fast enough these days for most people.
TrackSmart - Tuesday, July 1, 2014 - link
I second this. The Anandtech SSD tests were designed so that we could tell the difference between drives that are all so fast - that there is no way to tell them apart in ordinary usage scenarios. I see the value of testing the theoretical performance of drives as manufacturers push the technological limits.That said, at the end of the day user-experience is what matters. I agree with emn13 that the "light workload" test is already more strenuous than anything the average user is likely to do, and looking at the chart, we see that almost every drive is within a range of ~280 to ~380 MB/s. I'm guessing that the range in performance gets even narrower for "real world" workloads.
So keep up the innovative SSD testing, but be sure to put these theoretical performance gains into a real-world context when you get to the Conclusions section of these articles. Not everyone will benefit from these theoretical increases in performance.
hojnikb - Tuesday, July 1, 2014 - link
Is Samsung planning on doing TLC based V-NAND anytime soon ?It would be great for a mainstream drive, since endurance would be higher (due to older node), speeds would probobly also went up (so no need for gimicks like turbowrite).
Or is it not mature enough to scale down to TLC ?
artifex - Tuesday, July 1, 2014 - link
You had me at 10 years warranty. I don't mind the slight premium if I'm not buying another one midway through the cycle. Sure, it will be obsolete well before it dies, but that term signals Samsung is really confident about their reliability.Gigaplex - Tuesday, July 1, 2014 - link
Since it's twice the price of competition like the MX100, you're better off replacing mid way through the cycle.Arnulf - Tuesday, July 1, 2014 - link
I must have missed this in the article - are these V.NAND cells as used in 850 Pro drives 2 or 3 bits per cell ? I got the "larger lithography improves endurance" part, I'm just wondering whether they opted for more conservative option (MLC) there as well.extide - Tuesday, July 1, 2014 - link
These are MLC, or 2 bit per cell.It would be interesting if the non pro 850 comes out with TLC V-NAND!
himem.sys - Tuesday, July 1, 2014 - link
Heh, we are waiting for tests 850pro vs 840pro, because there are no bigger differences "on paper".sirvival - Tuesday, July 1, 2014 - link
Hi,one question:
In the review the idle power consumption for e.g. the 850 128gig is 35 mw.
I wanted to compare that to my Samsung 470 so I went to Bench and selected the drives for comparison.
There it says that the 850 uses 0.29 Watt.
So how comes there is a difference?
KAlmquist - Tuesday, July 1, 2014 - link
Anandtech Bench has four SSD power numbers:SSD Slumber Power (HIPM+DIPM)
Drive Power Consumption - Idle
Drive Power Consumption - Sequential Write
Drive Power Consumption - Random Write
The confusing things are that (1) the review only listed slumber power, not idle power, and (2) Bench lists both numbers but doesn't place the slumber power next to the other power values.
mutantmagnet - Tuesday, July 1, 2014 - link
I also find the lack of powerloss protection being a big negative over this hard drive. Until REFS has all the features it needs in Windows that you would get in Linux this is going to be an important feature for anyone who values data integrity. Even after that happens it still might be very important.Cerb - Tuesday, July 1, 2014 - link
The SSD just quitting could brick the drive, and could hose up the FS. Like ECC RAM, if you need power loss protection, you need it regardless of file system. IMO, they should all at least be equipped with enough to gracefully finish in-progress writes and shut down (not necessarily empty buffers, but set up a state that they can be guaranteed to be able to roll back from).sonicology - Tuesday, July 1, 2014 - link
Typo?"Scaling below 20nm was seemed as a major obstacle but the industry was able to cross that..."
seemed a major obstacle or seen as a major obstacle, but not seemed as a major obstacle
Bladen - Tuesday, July 1, 2014 - link
Is it just me, or does the "Performance Consistency" page have interactive charts, all of which are entitled "(SSD name and capacity) - 4KB Random Write (QD32) Performance", but with different results in each of the graphs for the same SSDs and capacities? Also, descriptive text is missing below the last two.I'm presuming they are supposed to be a read and another read or write one, presumably at a lower queue depth.
MrSpadge - Tuesday, July 1, 2014 - link
They're the same data in different scales.paesan - Tuesday, July 1, 2014 - link
I purchased a 512GB MX100 when they first came out for $199. No way the extra speed from the 850 pro is worth over twice the price as the MX100. Nobody is going to keep that drive for 10 years anyway. In 10 years the drive will be obsolete. Most users won't even notice the difference in speed in their every day usage.MrSpadge - Tuesday, July 1, 2014 - link
BTW: that image is surely not an X-Ray but "just" an ordinary SEM image (scanning electron microscope).Spatty - Tuesday, July 1, 2014 - link
This is correct. SEM image after a FIB.drwho9437 - Wednesday, July 2, 2014 - link
As I mentioned elsewhere, it is probably a STEM bright field image after a FIB liftout of the gate stack. Certainly not an X-ray; he should be correct the article...Cerb - Tuesday, July 1, 2014 - link
"The wear out causes the insulators to lose their insulating characters" - shouldn't that be characteristic(s)?jann5s - Tuesday, July 1, 2014 - link
I'm wondering If this technology will also end up in SD (XD) memory cards?Cerb - Tuesday, July 1, 2014 - link
As soon as it is cheap enough. But, don't get your hopes up about performance. SD cards are mostly limited by the controllers being slow, and in the tiny package they fit in, with the narrow margins they have, there's not a lot of room, physically and economically, to give them fast controllers, even if you get a big one that must have several NAND dies, and are talking about full-size SD, where multiple channels might be viable. It sucks, and I dislike shopping for SD cards as much as anybody, but today, that's how it is.frenchy_2001 - Tuesday, July 1, 2014 - link
I think he was talking about V-NAND (3D cells) which is independent of the controller.I would guess it will, as density will continue to scale up which will make it the cheaper technology.
It is cutting edge now, but will let Samsung scale higher densities very aggressively in the coming years, replacing all their 2D NAND production (they announced it when presenting the 3D cells).
Harry Lloyd - Tuesday, July 1, 2014 - link
Personally I have no interest in this kind of performance, and I really hope they focus on reducing prices and increasing capacities. The MX100 is just great for home usage (system and gaming), and I would like to see a 512 GB equivalent for around 100 $ by the end of 2015.Spatty - Tuesday, July 1, 2014 - link
"Oftentimes when cell size is discussed, it is only the actual size of the cell that is taken into account, which leaves the distance between cells out of the conclusion."Incorrect. Oftentimes what is being discussed is the half pitch. The 16nm, 19nm, 20nm, etc of the die. That is not the cell. The cell is Always defined as the repeatable structure in a memory device, and this includes the space between cells as described. The cell size is incorrectly referenced as being the half pitch.
Then there is marketing gimmick by companies who call their products 19nm when it is really 19nm by 2xnm. A rectangle and not a true 19nm square half pitch.
Larry Endomorph - Tuesday, July 1, 2014 - link
Good review. Bad charts. All of these are useless to color blind people:http://images.anandtech.com/doci/8216/NAND%20overv...
http://images.anandtech.com/doci/8216/cell%20inter...
http://images.anandtech.com/doci/8216/V-NAND_1.png
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
http://images.anandtech.com/doci/8216/850%20Pro%20...
Cerb - Tuesday, July 1, 2014 - link
I never paid much attention, but you're right. If they changed the point shapes, and maybe dashed a couple of the lines, they could take care of that easily.fokka - Tuesday, July 1, 2014 - link
it's great to see a new drive from samsung and even greater seeing them advancing ssd tech and performance in such substantial ways. keeping that in mind i'm not really surprised about the msrp sammy is asking for its drives. and as always when new devices hit the scene, we're comparing msrp with real market prices here, so the difference should be a bit lower in a couple weeks when enough stock is available.that said, even if sata3 remains the most important storage interface today, it's kind of a shame seeing such a beautiful drive limited by this "old" interface. i know the new standards like m2, sata3.2 and pci-e-drives are still kind of a mess, but we already saw what higher throughputs in combination with more efficient interface protocols can do and seeing an expensive enthusiast drive like the 850 pro connected to sata3 just makes it seem more limited than it needed to be.
all that said, it doesn't change much for the average user, or advanced users even, since for most people a good sized evo or crucial is all they ever need in the years to come. upgrading to expensive drives like the 850 will only make sense for the most demanding users, for the rest it will only get interesting again when pci based storage gets more affordable.
Daniel Egger - Tuesday, July 1, 2014 - link
Minor nit: There's no such thing as "pentalobe torx" it's either one or the other but I'm guessing that it might have been torx security since pentalobe screws have only been used by Apple a couple of years back.iwod - Tuesday, July 1, 2014 - link
Its great to see its doing well in power consumption area. Which is important in Notebook. I hope we could bring this down to 2W or even 1.5W during operation.I really do think our SSD storage tier deserve a PCI-E lane direct from CPU. It would be great if the market just settle on 2x PCI-E 3.0 from CPU. We get 2GB/s out of it. That is plenty of headroom to grow until we move to PCI-E 4.0
hojnikb - Tuesday, July 1, 2014 - link
Thats what sata-express is doingextide - Tuesday, July 1, 2014 - link
Well, kinda, I mean, some implementations come from the IOH instead of the CPU. I have heard rumors that future versions of Intel Desktop CPU's will have 20-24 PCIE lanes on them instead of 16. That would be perfect for storage!smithrd3512 - Tuesday, July 1, 2014 - link
Gotta love that warranty. 10 years on the drive. Might be worth the extra cost just for that alone.rahuldesai1987 - Tuesday, July 1, 2014 - link
"That is very aggressive because it essentially implies that the die capacity will double every year (256Gbit next year, 512Gbit in 2016 and finally 1Tbit in 2017)" - Does this mean a 8TB drive at $600 in 2017 ($75 per TB). Good bye hard drives by then :). What about a 850/850 Evo version?DarkXale - Tuesday, July 1, 2014 - link
It certainly does imply 8TB SSDs by 2017. By that point such a SSD will likely have a higher capacity than HDDs of that time.Of course, price will be very significantly in favour of the HDD still; but if money is no object you could do bulk storage in a portable device if you wanted to.
CalaverasGrande - Tuesday, July 1, 2014 - link
This may become an exhibit in some future dispute between Samsung and Apple. Those prices are easily Apple territory.extide - Tuesday, July 1, 2014 - link
Ehh, those prices were par for the course 18-24+ months ago!toyotabedzrock - Tuesday, July 1, 2014 - link
Does the height of these 32 layers make the cells more delicate when subjected to horizontal movement?And is this mlc or TLC?
MrSpadge - Wednesday, July 2, 2014 - link
Do, it doesn't. The height scale is still in the µm range, which is pretty much stable on macroscopic sclaes.emvonline - Tuesday, July 1, 2014 - link
thanks for the article on VNAND SSD. I think the SSD analysis is good and shows the impact. The details of Planar NAND and VNAND are incorrect in many cases. The overall NAND takeaway should be Samsung VNAND is a 86Gbit device Die level with a very large effective cell size. I still want to buy one... where can I get it?Kristian Vättö - Tuesday, July 1, 2014 - link
"The details of Planar NAND and VNAND are incorrect in many cases."Can you elaborate on that? I'm not saying that there can't be mistakes but it doesn't help me unless you explain what you think is incorrect.
emvonline - Tuesday, July 1, 2014 - link
is the die size for the NAND chip 67mm^2? I assumed you measured it (pretty easy to do). I would think it would be much larger than thatGigaplex - Tuesday, July 1, 2014 - link
It's difficult to measure the die when multiple of them are on the same package.emvonline - Tuesday, July 1, 2014 - link
send me the package... I promise to have exact die size in a week ..... TEM cross sections in two weeks :-)Kristian Vättö - Tuesday, July 1, 2014 - link
Send me an email at [email protected] and let's work this out :)extide - Tuesday, July 1, 2014 - link
Yes, 67mm^2, but remember that is 32 "deep"emvonline - Tuesday, July 1, 2014 - link
@extide: so you measured the die to be 67mm^2? how was this measured ? CSAM? XRAY? that seems odd to use such a small dies for SSD. and that would make it even smaller cell size than the one at ISSCC since it was 134 for a 128Gbit with 24 layers (periphery doesnt shrink as fast with lower density)Kristian Vättö - Tuesday, July 1, 2014 - link
No I didn't because I don't have the equipment to do that. The method I used to calculate the die size is explained on the fifth page of the review:http://www.anandtech.com/show/8216/samsung-ssd-850...
I'm not claiming that it is an accurate figure, hence the "~" sign in front of it. However, Samsung wouldn't disclose the die size when I asked them during the Q&A, so at this point I don't know for sure. However, I have a picture of the 32-layer wafer and once I get back home I'll do the math of the wafer to figure out the exact die size.
emvonline - Tuesday, July 1, 2014 - link
Got it thanks. I am mainly wondering about redundancy, extra blocks and ECC overhead.drwho9437 - Tuesday, July 1, 2014 - link
The micrograph you say is an "x-ray", is almost certainly a transmission electron microscopy image. Given the oxides are light it is a bright field image. It could technically be a SEM image but the resolution is a bit to high, so it most likely is a STEM image with a bright field detector.GTVic - Tuesday, July 1, 2014 - link
Perhaps time should be added as a cost efficiency factor. Presumably the die has to stay in the process much longer due to all the layers being added one at a time.GTVic - Tuesday, July 1, 2014 - link
Wondering if the 3D V-NAND has an effect on heat produced by each chip?Laststop311 - Wednesday, July 2, 2014 - link
If only this drive would of been a pci-e 3.0 x4 interface with nvme. We would of finally had a worthy upgrade. Yes it's performance is better than the 840 evo but you can get the 840 evo 1tb for 400 dollars even less if you catch it on sale. So you can get 2x 1tb 840 evo for almost the same price as 1 850 pro. If you compare an 840 evo with 25% OP to a regular 7% on the 850 the 840 evo looks just as good so 2x 825GB drives with 25% OP on each drive costs you 750-800 depending on the deal you get vs 700 for 1000GB. I would rather pay 800 for 1650GB than 700 for 1000GB with performance being nearly identical. I get an extra 650GB (which at 50 cents per GB is another 325 dollars worth of ssd) and tons of over provisioning to give the drive equal or better performance for only 100 dollars more (possibly less as the 840 evo has often sales)Sorry Samsung but 2x 1tb 840 evos with 25% provisioning gives me better or equal performance and a whole 650GB of extra storage for only 100 dollars more. At 50 cents per GB you get 325 dollars worth of more storage capacity for only 100 dollars more and thats with the 25% over provisioning which basically negates the performance increase of the new drives.
The only way samsung could of made this drive worth that money is if they had the drive on a pci-e 3.0 x4 interface with nvme instructions. I'm sure there will be tons of idiots who just buy it cause it's the latest drive. But if you use your brain you can see the 840 evo is still the best SATA drive when the cost/performance ratio is taken into account. 2x 1tb evo's in 25% OP mode gives you 1650GB and costs u 100 dollars extra or less and gives you the same performance or even better compared to 7% 850 pro 1000GB especially when raid 0 is taken into account. I'll take 1650GB over 1000GB if it's only 100 more and performance is equal or better easy choice.
Laststop311 - Wednesday, July 2, 2014 - link
Since samsung is stacking vertically now what they should of done is made a super huge 2tb drive to differentiate themselves from all these other drives. A 2tb drive for 1400 is a little more acceptable than 1tb for 700 simply because it's the only single drive with 2TB capacity.I can see great things happening with vertical stacked nand tho. When this process matures we should be seeing nand drives surpass spinning hard drives in capacity. When samsung has those 1tbit dies its planning for 2017 we should be able to have 8-12TB SSD's
althaz - Wednesday, July 2, 2014 - link
It's a new product and it's priced highly. Eager early adopters who want to move on to the latest and greatest will buy now, value-concious people will buy the 840s. Inventory of the 840s will get eaten up and the 850s will drop in price.This is what happens whenever any product is released, basically ever.
asmian - Wednesday, July 2, 2014 - link
Sorry, but neither this nor the EVO will get my money. Performance is all very well, even if only a tiny handful of users with extremely niche workloads will ever notice any difference, but all this extra reliability at a price premium means NOTHING without simple power loss protection. Restricting something so basic to "Enterprise" class products is the real gouging here by Samsung, not the price.If Crucial can provide that protection on the CHEAPEST drives in their class (M500/M550 and IIRC MX100 too) with performance that is not gimped as a balance, then there is no excuse for Samsung not to. This should be a no-compromise baseline for all SSDs going forward, and Anandtech should push hard for that - users should as well, by voting with their wallets and refusing to buy drives, however fast and powerful, that do not provide power loss protection as a basic feature.
bsd228 - Wednesday, July 2, 2014 - link
Though I agree it is a highly useful feature, it is far less significant to those of us using a good UPS. So I can't agree that it's a no compromise feature.romrunning - Wednesday, July 2, 2014 - link
What you aren't taking into account is the fact that the 850 Pro has MUCH higher endurance, and it's also more consistent. Those two items bring it more into the Intel DC 3500/3700 series type of class. It's not just a sheer performance comparison....and if you thing the 850 Pro is expensive, price out the larger Intel DC S3700 drives.
FunBunny2 - Wednesday, July 2, 2014 - link
"Real" Enterprise SSDs don't even have an MSRP. You negotiate with the vendor, and hope for the best. Now that Texas Memory is in IBM, and Fusion-io in SanDisk, with Violin likely to go somewhere. The conundrum is V.NAND's impact on flattening the curve between Enterprise and Commodity/Retail. At one time, a mainframe had bespoke 14" behemoth hard drive subsystems (in the case of IBM, run by the equivalent of a PDP-x). In due time, binned commodity 3.5" drives are now used.Samsung could well be the driving force to regularize solid state storage. The remaining issue is whether the file system hand waving will be dumped in favor of direct NVM persistence? Samsung, or whoever, likely couldn't care less.
romrunning - Wednesday, July 2, 2014 - link
It's been enjoyable to see a lot of "new" flash memory storage vendors pop-up. More competition is always good in that enterprise space.I've been looking forward to having more SSDs options available to servers at much better pricing. Solid storage advances have a trickle-down effect. If I can put an array of these Samsung 850 Pros into a server and achieve near "enterprise" performance, then that forces Dell/HP/etc. to drop their own SSD pricing.
watersb - Wednesday, July 2, 2014 - link
Fantastic detail of 3D NAND design and why it matters. Thanks very much!Pastuch - Wednesday, July 2, 2014 - link
When will we see Vnand in smart phones? This 16gb Nexus 5 is brutal! Always out of space.ajlueke - Wednesday, July 2, 2014 - link
Page 1, "Scaling below 20nm was seemed", I believe you intended "Scaling below 20nm was seen".Automaticman - Wednesday, July 2, 2014 - link
Can current EVO 840 users upgrade to Magician 4.4 and get the benefits of RAPID 2.0 (assuming they have >16GB DRAM)?bsd228 - Wednesday, July 2, 2014 - link
just as RAPID support was extended to the 840 PRO, I would expect this to support the 840 pro/evo soon, if not right off the bat.Automaticman - Sunday, July 6, 2014 - link
Well, I was certainly able to upgrade to Magician 4.4, and it did take a couple reboots while it was re-activating RAPID. I am going to guess and say yes it seems to be the new version of RAPID, but I don't see anywhere that it actually says RAPID 2.0 or any indication of how much memory it has available.sirvival - Wednesday, July 2, 2014 - link
Ah ok.When I select the 470 in Bench there is no slumber so I got confused.
Since power cosumption is a big thing in Mobile could you do the following:
Bench the impact on the Battery of a Laptop due to a fast drive.
E.g. drive A is fast but has the downside that it draws more under load as drive B which is slower. But since its done faster it returns to idle faster.
I mean impact on real world scenarios.
Or how much power was used for bench x etc. and have a average per hour or something like that.
sirvival - Wednesday, July 2, 2014 - link
damn this was to be a reply tohttp://www.anandtech.com/comments/8216/samsung-ssd...
sorry
Nickat - Thursday, July 3, 2014 - link
Thank you so much. You explained everything so well.Stokkolm - Thursday, July 3, 2014 - link
Newegg still has them for preorder at the more expensive price, hopefully they drop those before the release date.skarthikeyan - Monday, July 7, 2014 - link
Hi, How come random read is 106.8MB/sec and random write is 292.4MB/sec for the SSD 850 Pro 256 GB? Aren't writes supposed to be slower than reads?YazX_ - Monday, July 7, 2014 - link
Prices are not going down, good thing we have Crucial who have best bang for the buck, ofcourse performance wise is not compared to sandisk or samsung, but its still a very fast SSD, for normal users and gamers, Mx100 is the best drive you can get for its price.soldier4343 - Thursday, July 17, 2014 - link
My next upgrade the Pro 850 512gb version over my OCZ 4 256gb.bj_murphy - Friday, July 18, 2014 - link
Thanks Kristian for such an amazing, in depth review. I especially loved the detailed explanation of current 2D NAND vs 3D NAND, how it all works, and why it's all so important. Possibly one of my favourite Anandtech articles to date!DPOverLord - Wednesday, July 23, 2014 - link
Looking at this it does not seem to be a HUGE difference than raid 0 of 2 Samsung Pro 840 512GB (1tb in raid 0).To upgrade at this point does not make the most sense.
Nickolai - Wednesday, July 23, 2014 - link
How are you implementing over-provisioning?joochung - Tuesday, July 29, 2014 - link
I don't see this mentioned anywhere, but were the tests performed with RAPID enabled or disabled? I understand that some of the tests could not run with RAPID enabled, but for those other tests which do run on a formatted partition (i.e. not run on the raw disk), its not clear if RAPID is enabled or disabled. Therefore its not clear how RAPID will affect the results in each test.Rekonn - Wednesday, July 30, 2014 - link
Anyone know if you can use the 850 Pro ssds on a Dell PERC H700 raid controller? Per documentation, controller only supports 3 Gb/s SATA.janos666 - Thursday, August 14, 2014 - link
I always wondered if there is any practical and notable difference between dynamic and static over-provisioning.I mean... since TRIM should blank out the empty LBAs anyway, I don't see the point in leaving unpartitioned space for static over-provisioning for home users. From a general user standpoint, having as much usable space available as possible (even if we try to restrict ourself from ever utilizing it all) seems to be a lot more practical (until it's actually usable with an acceptable speed, so even if notably slower but still fast enough...) than keeping a (significantly more, but still not perfectly) constant random write performance.
So, I always create a system partition as big as possibly (I do the partitioning manually: a minimal size EFI boot partition + everything else at one piece) without leaving unpartitioned space for over-provisioning and I try to leave as much space empty as possible.
However, one time, after I filled my 840 Pro up to ~95% and I kept it like that for 1-2 days, it never "recovered" . Even after I manually ran "defrag c: /O" to make sure the freed up space is TRIMed, sequential write speeds were really slow and random write speeds were awful. I ha to create a backup image with DD, fill the drive with zeros a few times and finally run an ATA Secure Erase before restoring the backup image.
Even though I was never gentle with the drive (I don't do stupid things like disabling swapping and caching just to reduce it's wear, I bought it to use it...) and I did something which is not recommended (filled almost all the user-accessible space with data and kept using it like that for a few days as a system disk), this wasn't something I expected from this SSD. (Even though this is what I usually get from Samsung. It always looks really nice but later on something turns out which reduces it's value/price from good or best to average or worse.) This was supposed to be a "Pro" version.
stevesy - Friday, September 12, 2014 - link
I don't normally go out of my way to comment on a product but I felt this product deserved the effort. I've been using personal computer since personal computers first came out. I fully expected my upgrade from an old 50gig SSD to be a nightmare.I installed the new 500gig Evo 850 as a secondary, cloned, switch it to primary and had it booting in about 15 minutes. No problems, no issues, super fast, WOW. Glad Samsung got it figured out. I'll be a lot less concerned my next upgrade and won't be waiting until I'm at my last few megabytes before upgrading again.
basil.bourque - Friday, September 26, 2014 - link
I must disagree with the conclusion, "there is not a single thing missing in the 850 Pro". Power-loss protection is a *huge* omission, especially for a "Pro" product.Krakadoom - Saturday, September 27, 2014 - link
I bought the 850 Pro 500 GB version. The transfer speeds are around (and just under) 400 MB/s and the IOPS are always around 55-60k. Not impressed - considering returning it just due to the huge gap between the rated specs and actual performance. (Of course Samsung Magician over-reports the transfer speed by quite a large margin).stealth_lee - Wednesday, October 1, 2014 - link
Someone just tipped me that V-NAND in Samsung 850 Pro is actually TLC not MLC.The first reason is the 86Gbit/die number is odd, if 850 Pro uses a TLC 128Gbit/die and emulates it to MLC then it would be 86Gbit/die, the numbers fit well.
The second reason is Chipworks confirmed it in the die shots:
http://bit.ly/YTVm9Z
http://bit.ly/1uByKcm
I'm just the messenger here, I'm not expert.
So...I was wodering is it possible to hack Samsung 850 Pro to get extra storage space in TLC?
wcatlan - Saturday, October 25, 2014 - link
Why isn't the lack of power loss protection a showstopper for any of these drives? I love the speed and reliability benefits under normal operation, but how can anyone get excited about a drive that can get corrupted in an instant due to power loss or computer freeze, where a hard shutdown is required? Seems that these drives are more prone to massive data issues much more than HDDs under the same power fault conditions. I keep looking for a good answer, but it seems smart people are willing to look past this seemingly fatal Achilles heal. Not sure what I might be missing. Any thoughts?futurefilm - Monday, December 1, 2014 - link
Today, Cyber Monday deals on Amazon, the 850 Pro 256 is going for $150. The 128 for $100. Get it now while it's hot.saagar - Thursday, January 22, 2015 - link
Dear Kristian Vättö,Fantastic review of the drives and the technology behind it. This is what readers like me expect to see on Anandtech. Thanks for breaking it down. Keep up the good work!
gsuburban - Wednesday, April 8, 2015 - link
As of April 8, 2015, the 850 Pro 256GB SSD can be had for about $144 if you look hard enough.rockfella79 - Saturday, June 27, 2015 - link
I love my 850 Pro 128 GB SSD :)KDT - Thursday, March 24, 2016 - link
Please update the endurance to 300TBW for 1TB model. This was my basis for buying this SSD. This is 2nd to Crucial MX200 (320TBW on 1TB model) in terms of endurance - for client/consumer SSDs.BimmerInd - Sunday, June 26, 2016 - link
Samsung is using 40nm over Micron's 16nm. Doing the math implies that for every 2.5 16nm Micron nodes in planar section, Samsung only does 1 40nm node. If we scale vertically to 32 layers, then Micron (or others for that matter) still do only 2.5 nodes for every 32 nodes of Samsung. Which means for every 16nm node, Samsung provides 12.8 nodes. Meaning the density scaling factor for every 32 layer increments is a multiple of 12.8. Assuming the current die size for 32 layers to be 128Gbit, then the density advantage for 256Gbit is 12.8x2 times, 512Gbit is 12.8x4 times and for 1Tbit in 2017 should be 12.8x8 times for 256 layers of nodes stacked on top of each other. So the density advantage is approximately 102.4% (theoretically). Samsung can theoretically produce a 1Tbit die at a cost advantage/space advantage of nearly 100 times compared to planar and manufacturers. It is almost like you are able to earn 100 times the profit for the same die provided the cost per bit is scaled along without passing on the price advantage to the end users until other players enter 3D market.The same is the case with Intel's 3D Cross Point Technology. They are having a new tech in their hands that is faster than NAND and closer to DRAM. So they are also planning to price it exactly between NAND and DRAM. We are already paying high costs to shift from platters to NAND and are going to pay even more to make a shift from NAND to 3D xPoint. I just wish I can jump a few years to the future, grab a high capacity NAND/xPoint drive for cheap and come back to the present and use it. Sigh !
BimmerInd - Sunday, June 26, 2016 - link
By the way this is just a rough calculation and is not to be taken literally.