Over the last three years, Samsung has become one of the most dominant players in the SSD industry. Samsung's strategy has been tight vertical integration ever since the beginning, which gives Samsung the ability to be in the forefront of new technologies. That is a massive advantage because ultimately all the parts need to be designed and optimized to work properly together. The first fruit of Samsung's vertical integration was the SSD 840, which was the first mass produced SSD to utilize TLC NAND and gave Samsung a substantial cost advantage. Even today, the SSD 840 and its successor, the 840 EVO, are still the only TLC NAND based SSDs shipping in high volume. Now, two years later, Samsung is doing it again with the introduction of the SSD 850 Pro, the world's first consumer SSD with 3D NAND.

For years it has been known that the scalability of traditional NAND is coming to an end. Every die shrink has been more difficult than the previous as the endurance and performance have decreased with every node, making it less and less efficient to scale the size down. Scaling below 20nm was seemed as a major obstacle but the industry was able to cross that with some clever innovations in the NAND design. However, the magic hat is now running out of tricks and a more signficant change to the NAND design is required to keep scaling the cost. 

The present solution to the scalability problem is 3D NAND, or V-NAND as Samsung calls it. Traditionally NAND and other semiconductors are scaled horizontally along the X and Y axes but due to the laws of physics, there is a limit of how small the transistors can be made. To solve the problem, 3D NAND introduces a Z-axis i.e. a vertical dimension. Instead of cramming transistors horizontally closer and closer to each other, 3D NAND stacks layers of transistors on top of each other. I will be going through the structure and characteristics of 3D NAND in detail over the next few pages.

By stacking transistors (i.e. cells when speaking about NAND) vertically, Samsung is able to relax the process node back to a much more convenient 40nm. When there are 32 cells on top of each other, it is obvious that there is no need for a 10nm-class node because the stacking increases the density, allowing production costs to scale lower. As we have seen with the history of NAND die shrinks, a higher process node provides more endurance and higher performance, which is what the 850 Pro and V-NAND is all about.

Fundamentally the only change in the 850 Pro is the switch to V-NAND. The interface is still SATA 6Gbps and the controller is the same triple-core MEX from the 840 EVO, although I am still waiting to hear back from Samsung whether the clock speed is the same 400MHz. The firmware, on the other hand, has gone through a massive overhaul to adopt the characteristics of V-NAND. With shorter read, program and erase latencies and higher endurance, the firmware needs to be properly optimized or otherwise the full benefits of V-NAND cannot be utilized. 

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 off with PCIe for a little while longer. The market for aftermarket PCIe SSDs is still relatively small as the PC industry is figuring out how to adopt the new interface, so for the time being Samsung is fine with watching from the side. The XP941 is and will continue to be available to the PC OEMs but for now Samsung will be keeping it that way. From what I have heard, Samsung could bring the XP941 to the retail market rather quickly if needed but Samsung has always been more interested in the high volume mainstream market instead of playing in the niches. 

Samsung SSD 850 Pro Specifications
Capacity 128GB 256GB 512GB 1TB
Controller Samsung MEX
NAND Samsung 2nd Gen 86Gbit 40nm MLC V-NAND
DRAM (LPDDR2) 256MB 512MB 512MB 1GB
Sequential Read 550MB/s 550MB/s 550MB/s 550MB/s
Sequential Write 470MB/s 520MB/s 520MB/s 520MB/s
4KB Random Read 100K IOPS 100K IOPS 100K IOPS 100K IOPS
4KB Random Write 90K IOPS 90K IOPS 90K IOPS 90K IOPS
Power 2mW (DevSLP) / 3.3W (read) / 3.0W (write)
Encryption AES-256, TCG Opal 2.0 & IEEE-1667 (eDrive supported)
Endurance 150TB
Warranty 10 years
Availability July 21st

The performance figures in the table above give us the first glimpse of what V-NAND is capable of. Typically modern 128GB SSDs are only good for about 300MB/s but the 850 Pro is very close to saturating the SATA 6Gbps bus even at the smallest capacity. This is due to the much lower program times of V-NAND because write performance has been bound by NAND performance for quite some time now. 

Endurance Comparison of High-End SSDs
Samsung SSD 850 Pro Intel SSD 730 SanDisk Extreme Pro OCZ Vector 150
150TB 91TB (240GB)
128TB (480GB)
80TB 91TB

The other major improvement from V-NAND is the endurance. All capacities, including the smallest 128GB, are rated at 150TB, which is noticeably higher than what any other consumer-grade SSD offers. Moreover, Samsung told me that the endurance figure is mainly meant to separate the 850 Pro from the enterprise drives to guide enterprise clients to the more appropriate (and expensive) drives as the 850 Pro does not have power loss protection or end-to-end data protection for example. However, I was told that the warranty is not automatically denied if 150TB is reached under a client workload. In fact, Samsung said that they have a 128GB 850 Pro in their internal testing with over eight petabytes (that is 8,000TB) of writes and the drive still keeps going, so I tip my hat to the person who is able to wear out an 850 Pro in a client environment during my lifetime.

Another interesting aspect of V-NAND is its odd capacity per die. Traditionally NAND capacies have come in powers of two, such as 64Gbit and 128Gbit, but with V-NAND Samsung is putting an end to that trend. The second generation 32-layer V-NAND comes in at 86Gbit or 10.75GB if you prefer the gigabyte form. I will be covering the reason behind that in more detail when we look at V-NAND more closely in the next few pages but as far as I know there has never been a strict rule as to why the capacities have scaled in powers of two. I believe it is just a relic from the old days that has stayed in the memory industry because deep down binary is based on powers of two but the abnormal die capacity should have no effect on the operation of the NAND or the SSD as long as everything is optimized for it. 

NAND Configurations
  128GB 256GB 512GB 1TB
# of NAND Packages 4 (?) 4 8 (?) 8
Package Configurations 2 x 4 x 86Gbit
2 x 2 x 86Gbit
2 x 8 x 86Gbit
2 x 4 x 86Gbit
4 x 8 x 86Gbit
4 x 4 x 86Gbit
4 x 16 x 86Gbit
4 x 8 x 86Gbit
Raw NAND Capacity 129GiB 258GiB 516GiB 1032GiB
Over-Provisioning 7.6% 7.6% 7.6% 7.6%

Due to the odd die capacity, the die configurations are also quite unusual. I found two different capacity packages inside my review samples and with Samsung’s NAND part decoder I was able to figure out the die configurations for each capacity. Unfortunately, Samsung did not send us the 512GB model and I could not get the 128GB model open as Samsung uses pentalobe Torx screws and I managed to wear out the screw while trying to open it with an inappropriate screw driver (it worked for the other models, though), so thus there are question marks at those capacities in the table. However, this should not impact the raw NAND capacities as long as all capacities follow the same 7.6% over-provisioning trend but the package configurations may be different. I will provide an update once I receive a confirmation from Samsung regarding the exact configurations for each capacity. 

The 850 Pro also switches to smaller PCB designs. The PCB in the 1TB model populates around two thirds of the area of the chassis, while the 256GB PCB comes in at even smaller size. The reason for the different PCB sizes is the amount of NAND packages as the 256GB only has four, whereas to achieve the capacity of 1TB eight NAND packages are required.

Why We Need 3D NAND
Comments Locked


View All Comments

  • 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:


    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?

Log in

Don't have an account? Sign up now