Mushkin Launches Reactor Armor 3D and Triactor 3D 2TB SATA SSDs: 3D NAND, SM2258

Mushkin at CES demonstrated its new SSDs in 2.5”/7 mm form-factor aimed at mainstream PCs with a SATA interface. The new Reactor Armor 3D and Triactor 3D use 3D NAND flash memory, the same controller from Silicon Motion and offer nearly similar performance. The main difference is that the former use 3D MLC, whereas the latter uses 3D TLC memory.

The NAND flash industry is transitioning to various 3D NAND architectures that enable higher densities, lower per-bit costs and higher endurance compared to planar flash made using very thin process technologies. So far it has not been easy for independent makers of drives to secure a supply of 3D NAND memory because some manufacturers are cutting down the share of produced flash they sell on the open market, whereas 3D NAND from others does not suit SSDs well. In the recent months ADATA was the only independent supplier of drives to offer 3D NAND-based drives, but as we observed at CES, the situation is about to change. Mushkin is another company to announce a lineup of SSDs featuring 3D NAND and targeting different market segments, from entry-level to the high-end. Unlike ADATA, Mushkin is announcing all of its 3D NAND SSDs at once, which implies that the company can get enough chips for different kinds of drives.

Mushkin’s Reactor Armor 3D and Triactor 3D SSDs are based on Silicon Motion’s SM2258 controller, but while the former uses 3D MLC NAND, the latter uses 3D TLC NAND from an undisclosed manufacturer. The SM2258 controller has four NAND flash channels, LDPC ECC technology, a SATA interface, a DRAM buffer support as well as pseudo-SLC (pSLC) caching in order to maximize SSD performance. At present, the SM2258 is virtually the only market-ready third-party SSD controller with that supports 3D NAND (technically speaking, the SM2256 also supports 3D NAND, but drive makers prefer the more advanced controller so to address the higher end of the SSD spectrum), so Mushkin’ s choice is not surprising if the company needs rapid time-to-market (which is also why it does not wait for Phison’s PS5008-E8). What is even more interesting is that Mushkin is considering to add 3D NAND-based drives to the Reactor lineup that uses the SM2246EN controller (this one is qualified for 3D MLC as well). It does not look like the company has made any final decisions, but it is considering such possibility in a bid to continue addressing the entry-level segment with the Reactor lineup.

Mushkin does not disclose the name of its 3D NAND flash supplier, but we have a reason to believe that this is Micron. SanDisk and Toshiba are shipping their 64-layer BiCS NAND inside their removable media products and promise to use this memory for their SSDs. But as of now, 64-layer BiCS chips have not been qualified for SSDs. 3D NAND from SK Hynix is also available for various products, but it has not been qualified for SSDs just yet.

The Reactor Armor 3D SSDs will be available in 240 GB to 1920 GB configurations, whereas the Triactor 3D drives will feature 256 GB to 2 TB capacities. The former family will take advantage of MLC and offer slightly better endurance albeit at a higher price, whereas the latter lineup will be more aggressively priced thanks to cheaper memory. At the same time, it is noteworthy that both product lines include high-capacity (~ 2 TB) drives, an indicator that they target customers who need a lot of non-volatile memory and can pay for that.

As for performance, Mushkin rates sequential read speed of both Reactor Armor 3D and Triactor 3D drives at 565 MB/s, whereas sequential write speed is rated at up to 525 MB/s and 520 MB/s (when pseudo-SLC caching is used) respectively. Random performance of the drives is specified at up to 90,000 read IOPS and up to 85,000 write IOPS.

A New Challenger Appears: Palit’s Own-Brand UVS and GFS SSDs Announced

Palit has announced two families of SSDs that it plans to sell under its own brand. The new drives are aimed at entry-level and mainstream gaming PCs, and will be based on controllers from Phison using 3D MLC or 3D TLC NAND flash memory from Micron depending on which drive you pick up. The Palit SSDs will be among the first drives on the market that will use a combination of a Phison controller and 3D NAND memory ICs from Micron, but we expect this combination to spread across several SSD vendors in due course.

Palit Microsystems is one of the world’s largest producers of graphics cards, but it is not entirely new to SSDs too. Palit’s GALAX and KFA2 brands have offered Phison-based SSDs for quite a while, but their lineups have never been large and the whole effort looked more like a brand development rather than an attempt to compete against much of the market. This time, Palit has announced two families of SSDs under its own trademark and with seven drives in total, it plans to address entry-level and mainstream gaming PCs. We do not know Palit’s plans in regards of higher-end drives in M.2 or add-in-card form-factors, but such products are available from other brands that Palit owns and it should not be a problem for the company to expand its own lineup if it needs to.

Palit will initially offer two families of SSDs: First is the Palit UVS family, featuring the Phison S3111-S11 controller and 3D TLC memory for entry-level gaming systems. Then second is the Palit GFS family, based on the same Phison S3111-S11 controller but with 3D MLC NAND flash.

Before we start discussing the drives, let’s talk a little bit about the controller itself. Formally, the PS3111-S11 is positioned below the S10 because it has only two NAND channels with 16 CE targets and physically cannot deliver breakthrough performance. As it is a SATA controller, the PS3111-S11 does not have to deliver anything sequentially higher than 550 MB/s and this is something it can do with both MLC and TLC chips (sustained performance is a different comparison). The most important advancement of the controller versus its predecessors is that the PS3111-S11 supports LDPC ECC, and thus can be enabled on SSDs with sufficient endurance. Additionally, the PS3111-S11 supports 3D and 1z MLC/TLC NAND flash and memory with large (8 KB and 16 KB) blocks.

As for the drives, the Palit UVS family will include 120 GB, 256 GB, 480 GB and 512 GB models using 3D TLC NAND (except the 120GB, which is planar TLC). Depending on the model, the drives are rated to deliver up to 560 MB/s sequential read speed and up to 470 MB/s (370 MB/s for the 120 GB version) sequential write speed. As for random performance, the numbers on the box give 72,500 read IOPS and up to 85,000 write IOPS.

The Palit GFS lineup consists of three drives with 120 GB, 128 GB and 240 GB capacities all based on 3D MLC and offering all the endurance-related benefits of such memory. From a performance point of view, the GFS SSDs are slightly faster than the UVS drives: they are rated for up to 560 MB/s sequential read speed and up to 480 MB/s sequential write speed. Palit also states they can also perform up to 75,000 read IOPS and up to 87,500 write IOPS (240 GB version only). Palit may decide to expand the GFS lineup with higher-capacity offerings over time, but right now, its premium drives only offer entry-level capacities.

There are two intrigues about Palit’s SSDs: the memory supplier and actual manufacturer. Typically, Phison ships its controllers with memory and firmware and in many cases even provides assembly and test services (essentially, shipping already made drives). Despite this, Palit has enough SMT lines and can produce virtually everything itself. At present, we do not know whether Palit-branded SSDs are made by Palit, or are manufactured by a third party, but the latter is clearly a possibility here.

The supplier of the NAND is also not obvious and could come from different sources. Palit does not disclose who is their supplier, but it is worth noting that Phison usually ships its controllers primarily with memory from Toshiba. We do know that there are Phison PS3111-S11-based reference designs featuring Toshiba’s BICS2 memory (which is not exactly positioned for SSDs by Toshiba) as well as S11 drives with Micron’s 3D NAND memory.

The Palit SSDs are expected to hit the market in the coming months. We do not have any information about their MSRP of the new drives, but it is logical to assume that Palit will try to make them competitive in terms of pricing.

The Intel Optane SSD DC P4800X (375GB) Review: Testing 3D XPoint Performance

Intel’s new 3D XPoint non-volatile memory technology, which has been on the cards publically for the last couple of years, is finally hitting the market as the storage medium for Intel’s new flagship enterprise storage platform. The Intel Optane SSD DC P4800X is a PCIe SSD using the standard NVMe protocol, but the use of 3D XPoint memory instead of NAND flash memory allows it to deliver great throughput and much lower access latency than any other NVMe SSD.

3D XPoint
The potential significance of 3D XPoint memory is immense. When it was first publicly announced by Intel and Micron in 2015, 3D XPoint memory was a fundamentally different storage technology from the flash memory that dominates the market. It is the first new truly mass market, high-density solid state storage medium to hit the market since NAND flash itself. It comes at a time where the NAND market is booming like never before, but also at a time when we know that there is a definite end of the line for NAND. The ongoing transition to 3D NAND flash is just a temporary postponement of the fundamental limitations of flash memory. Once NAND can no longer scale in density and cost-per-bit, it will fall to paradigm changes and next-generation memory technologies (one of which will be 3D XPoint) to continue to carry the industry forward. There are many other new memory technologies that may compete alongside flash memory and 3D XPoint in the coming years, but 3D XPoint is the one that’s ready to go mainstream now.

In the near term, 3D XPoint is important because it offers a new set of performance tradeoffs entirely unlike NAND; tradeoffs that, for the right applications, can deliver performance far in excess of today’s NAND products. By being able to read and write at the bit or word level – and not the 4K+ page level of NAND – 3D XPoint has the potential to deliver excellent performance across a wide range of workloads, but especially in minimally parallel workloads, which are common in the consumer and enterprise spaces.

The drawback here is that, due to various factors regarding time, production, and scope, 3D XPoint is more expensive than NAND. It also comes in as less dense, to aid in ease of production in this first stage, but this also adds to the cost. For now, due to scale and other factors, it won’t be able to replicate the sheer capacity and cost effectiveness that has made NAND storage so popular in all market segments. Due to the scale, especially as a first-generation version of the technology, the first 3D XPoint products are being aimed at speciality and high-margin markets: enterprise performance, consumer caching, etc. Future products promised from Intel should add non-volatile DIMMs to the mix, and then later on, if everything goes to plan, a potential wholesale replacement of NAND flash (or at least a strong competitor).

The Intel Optane SSD DC P4800X
The new storage drive, and the focus of today’s review, is the Intel Optane SSD DC P4800X. It uses a new NVMe controller Intel developed specifically for use with 3D XPoint memory. Where Intel’s enterprise NVMe SSDs like the P3700 use a controller with 18 channels for interfacing to their flash memory, the Optane SSD’s controller has only 7 channels. In order to achieve at least parity on peak performance, each of those channels has to provide much higher throughput than on a flash SSD, and it shows that each 3D XPoint memory die is delivering much higher performance than a die of flash memory.

The first capacity of the Optane SSD DC P4800X to ship and the model we’ve tested here offers a usable capacity of 375GB from a total of 28 3D XPoint memory dies (four per channel) for a raw capacity of 448GB. 3D XPoint memory has better endurance than NAND flash, but not enough to get away without wear levelling. The fine-grained accessibility of 3D XPoint memory gets rid of a lot of the wear leveling and write amplification headaches caused by flash pages and erase blocks being larger than the sector sizes exposed by the drives, but the drive still needs some spare area plus storage for error correction overhead, metadata for tracking the mapping between logical blocks and physical addresses, and potential replacement of bad sectors, similar to normal SSDs.

As with most NVMe SSDs, the Optane SSD DC P4800X supports a configurable sector size. Out of the box it emulates 512B sectors for the sake of compatibility, but using the NVMe FORMAT command it can be switched to emulate 4kB sectors. The larger sector size reduces the amount of metadata the SSD controller has to juggle, so it usually allows for slightly higher performance. The NVMe FORMAT command is also the mechanism for triggering a secure erase of the entire drive, and for flash SSDs the format usually consists of little more than issuing block erase commands to the whole drive. 3D XPoint memory does not have large multi-megabyte erase blocks, so a low-level format of the Optane SSD needs to directly write to the entire drive, which takes about as long as filling it sequentially. Thus, while a 2.4TB flash SSD can perform a low-level format in just over 13 seconds, the 375GB Optane SSD DC P4800X takes six minutes and 47 seconds. This is long enough that unsuspecting software tools or SSD reviewers will give up and assume that the drive has locked up.

So far, Intel has only started shipping the 375GB Optane SSD DC P4800X to select customers, and they have not released detailed specifications for the larger capacities that will ship later this year.
It is worth noting that the performance specifications for the P4800X, as provided in the product specification sheets, cover a different set of metrics than Intel usually reports for their enterprise SSDs. Sequential performance is not mentioned at all, but the product brief has quite a bit to say about latency: average latency for QD1 reads and writes, and 99.999th percentile latency for both reads and writes at QD1 and QD16. The fact that Intel is publishing a five-nines QoS metric at all suggests that they plan to set a new standard for performance consistency.
The throughput claims are also remarkable: half a million IOPS or more for reads, writes and a 70/30 read/write mix. There are already drives on the market that can deliver more than 550k random read IOPS, but those SSDs are far larger than 375GB and they require very high queue depths to hit 550k IOPS. There are even a few multi-TB drives that can beat 500k random write IOPS, but they can’t sustain that performance indefinitely. The Optane SSD DC P4800X is promising an unprecedented level of storage performance both in absolute terms and relative to its capacity, so it is interesting to see where Intel is going to lay down its line in the sand.
The P4800X will not really occupy the same niche as the multi-TB monsters that offer comparable throughput. With limited capacity but the highest level of performance, this Optane SSD most closely fits the role of SLC NAND based SSDs. SLC has disappeared from the SSD market as virtually all customers preferred to sacrifice a little bit of performance to double their capacity by using MLC NAND. One of the last high-performance SLC SSDs was the Micron P320h, a PCIe SSD from 2012 that slightly pre-dated NVMe and used 34nm SLC NAND flash. Anyone still using a P320h for its consistent low latency performance will be very interested in the P4800X. Outside of that niche, the Optane SSD will obviously be desirable for its raw throughput, but the low capacity may be problematic for some use cases.
One of the unique and most notable performance advantages of the Optane SSD DC P4800X is that it does not require extremely high queue depths to reach full throughput. Enterprise customers have long had to design their systems around the fact that getting full performance out of the fastest PCIe SSDs requires loading them down with queue depths of 128 or higher, sometimes requiring applications to use dozens of threads for I/O. In the client space achieving such queue depths is outright impossible, and in the enterprise space it doesn’t happen for free. The P4800X’s high performance at low queue depths makes it a much easier drive to get great real-world performance out of.
Intel originally introduced 3D XPoint memory as having far higher write endurance than NAND flash—on the order of 1000x higher. The Optane SSD DC P4800X is rated for 30 drive writes per day (DWPD) for five years, and the current models shipping during this early limited availability period are only rated for three years, rather than the five years it expects the support for the full retail models. Intel says they’re being extremely conservative with a new and unproven technology, and doing the math means that 30 DWPD doesn’t provide any endurance advantage over the most highly over-provisioned flash-based enterprise SSDs. In terms of total petabytes written, the P4800X only has four-fifths the endurance of the SLC-based Micron P320h. Even allowing for Intel’s original comparisons possibly having been relative to lower-endurance contemporary MLC or TLC flash, it seems like this first generation of 3D XPoint memory is not as durable as originally planned – the headline number of 30 DWPD is aimed at alleviating that issue, however for Intel to match its original intentions then the second and third generation parts will have to be a step up, and we look forward to testing them.

Pricing
The MSRP for the 375GB P4800X is $1520, though it will be quite some time before it can readily be ordered from major online retailers. At slightly more than $4/GB, the P4800X will be almost twice as expensive per GB as Intel’s next most pricey SSD, the P3608 (which is really two drives in one plus a PCIe switch). Compared to Intel’s fastest single SSD (the P3700), the P4800X will be more than three times as expensive per GB. In the broader SSD market, $4/GB is not completely unprecedented, but most companies selling drives in this price range don’t even pretend to have a retail price.
This Review
For this review of the Intel Optane SSD DC P4800X, first, we are going to take a deeper dive into what 3D XPoint actually is. Then we go through our testing suite for enterprise drives, testing Intel’s claims on performance.
It is worth noting that there is no such thing as a general-purpose enterprise SSD. Enterprise storage workloads are far more varied than client workloads and it is impossible to make general statements about whether random or sequential performance is more important, what kind of mix of reads and writes to expect, or what queue depth is apporpriate to test with. Real-world application benchmarks are difficult to construct and typically end up being far more narrowly applicable than we would hope. Our strategy for this review is to provide a very broad range of synthetic tests with the knowledge that not all results will be relevant to all use cases. Enterprise customers must know and understand their own workload. Since this is our first time testing anything with 3D XPoint memory, this review includes some new benchmarks that would probably not be applicable to a flash SSDs review, making for some interesting numbers.

TechInsights Publishes Preliminary Analysis of 3D XPoint Memory

Now that the Intel Optane Memory M.2 SSDs are readily available on the open market, anyone with an electron microscope and the skills to use it can begin to probe the secrets of 3D XPoint memory that Intel and Micron have been keeping tightly under wraps since announcing the new technology in August 2015. The reverse engineering experts at TechInsights have been doing just that, and they recently published their initial findings.

With some of the first high-resolution die photographs of 3D XPoint, TechInsights has provided precise measurements of the die size and memory density. The 128Gb 3D XPoint die is 206.5 mm2, much larger than is typical for modern NAND flash or DRAM but comparable to Intel’s 128Gb 20nm planar MLC NAND. A large total die size is typical for Intel and Micron, as they have historically not catered to the mobile market with their NAND flash while competitors like Samsung and Toshiba have strived to ensure their flash will physically fit in devices like smartphones. (That trend is changing with the introduction this year of 64-layer 3D NAND where Intel and Micron are producing both a larger 512Gb TLC part and a smaller 256Gb TLC part.)

The Intel Optane SSD DC P4800X is using memory of similar density to the Intel SSD DC P3700 that it is displacing as the flagship of Intel’s SSD product line. When comparing similar chips, die size is a strong predictor of manufacturing cost, but 3D XPoint memory is quite different from NAND flash memory, both older planar NAND or newer 3D NAND. Still, there’s some value in noting that the P4800X is arriving with a price tag about 25% higher than the P3700 initially carried. This suggests that the manufacturing process for 3D XPoint is either more expensive than planar NAND or that 3D XPoint yields are not mature enough, but a lot of the markup can also be explained by the lack of high-performance competition for Optane SSDs.

TechInsights calculates that 91.4% of the 3D XPoint die area is occupied by the memory array itself. This is a much higher figure than for NAND flash, where the record is 84.9% for Intel/Micron 3D NAND with its “CMOS under the array” design that puts a large portion of the peripheral circuitry underneath the memory array instead of alongside. Samsung’s current 48-layer 3D V-NAND manages an array efficiency of just 70%, and 3D NAND from Toshiba and SK Hynix has been comparable. This means that once Intel gets around to increasing they layer count in future generations of 3D XPoint memory, they should be able to get much closer to the ideal capacity scaling than 3D NAND memory can currently achieve.

The analysis from TechInsights confirms that 3D XPoint memory is manufactured using a 20nm process, with the same pitch in both the bitline and wordline directions of the memory array. The DRAM market is only just moving beyond this milestone, so comparing the density of 3D XPoint to current DRAM highlights the fundamental capacity advantage 3D XPoint enjoys: around 4.5 times higher density compared to typical 20nm DRAM, and about 3.3 times higher than the most advanced 1Xnm DDR4 on the market. This gap is likely to widen with future generations of 3D XPoint.

The materials and construction of an individual 3D XPoint memory cell have not been fully analyzed, but it appears to be a phase change memory element with a doped chalcogenide selector switch. The 3D XPoint memory array is constructed between the fourth and fifth metal interconnect layers above the silicon die.

Toshiba’s 768Gb 3D QLC NAND Flash Memory: Matching TLC at 1000 P/E Cycles?

Toshiba last week announced its first 3D NAND flash memory chips featuring QLC (quadruple level cell) BiCS architecture. The new components feature 64 layers and developers of SSDs and SSD controller have already received samples of the devices, which Toshiba plans to use for various types of storage solutions.

Toshiba’s first 3D QLC NAND chips feature 768 Gb (96 GB) capacity and uses 64 layers, just like the company’s BICS3 chips with 256 Gb and 512 Gb capacities launched in 2016 and 2017. Toshiba does not share further details about its 3D QLC NAND IC (integrated circuit), such as page size, the number of planes as well as interface data transfer rate, but expect the latter to be high enough to build competitive SSDs in late 2018 to early 2019 (that’s our assumption). Speaking of applications that Toshiba expects to use its 3D QLC NAND ICs, the maker of flash memory mentions enterprise and consumer SSDs, tablets and memory cards.

Endurance++
Besides intention to produce 768 Gb 3D QLC NAND flash for the aforementioned devices, the most interesting part of Toshiba’s announcement is endurance specification for the upcoming components. According to the company, its 3D QLC NAND is targeted for ~1000 program/erase cycles, which is close to TLC NAND flash. This is considerably higher than the amount of P/E cycles (100 – 150) expected for QLC by the industry over the years. At first thought, it comes across a typo – didn’t they mean 100?. But the email we received was quite clear:

– What’s the number of P/E cycles supported by Toshiba’s QLC NAND?
– QLC P/E is targeted for 1K cycles.

It is unclear how Toshiba managed to increase the endurance of its 3D QLC NAND by an order of magnitude versus initially predicted. What we do know is that signal processing is more challenging with QLC than it is with TLC, as each cell needs to accurately determine sixteen different voltage profiles (up from 2 in SLC, 4 in MLC, and 8 in TLC).

The easiest way to handle this would be to increase the cell size: by having more electrons per logic level, it is easier to maintain the data and also read from it / write to it. However, the industry is also in a density race, where bits per mm^2 is an issue. Also, to deal with read errors from QLC memory, controllers with very advanced ECC capabilities have to be used for QLC-based SSDs. Toshiba has its own QSBC (Quadruple Swing-By Codes) error correction technique, which it claims to be superior to LDPC (low-density parity-check) that is widely used today for TLC-powered drives. However, there are many LDPC implementations and it is unknown which of them Toshiba used for comparison against its QSBC. Moreover, there are more ECC methods that are often discussed at various industrial events (such as FMS), so Toshiba could be using any or none of them. The only thing that the company tells about its ECC now is that it is stronger than 120 bits/1 KB used today for TLC. In any case, if Toshiba’s statement about 1000 P/E cycles for QLC is correct, it means that that the company knows how to solve both endurance and signal processing challenges.

The main advantage of QLC NAND is increased storage density when compared to TLC and MLC, assuming the same die size. As was perhaps expected, die size numbers were not provided. However, last year Toshiba and Facebook talked about a case study QLC-powered SSD with 100 TB of capacity for WORM (write once read many) applications and it looks like large-capacity custom drives and memory cards will be the first to use QLC for cold storage. P/E cycles and re-write endurance isn’t a concern for WORM at this stage.

Toshiba has begun to sample its 3D QLC NAND memory devices earlier this month to various parties to enable development of SSDs and SSD controllers. Taking into account development and qualification time, Toshiba plans to mass produce its BiCS3 768 Gb 3D QLC NAND chips around the same time it starts to make its the next generation BiCS4 ICs. The latter is set to hit mass production in 2018, but the exact timeframe is yet to be determined.

Viking Ships UHC-Silo SSDs: 25 – 50 TB Capacity, Custom eMLC, SAS, $0.4 per GB

Viking Technology has started shipping their new lineup of ultra high capacity (UHC) SSDs designed to replace 3.5” HDDs in capacity-demanding applications that can take advantage of flash memory. The Viking UHC-Silo drives use planar eMLC NAND memory in custom packaging with raw NAND capacities of 25 TB and 50 TB, and consequently are currently the highest capacity SSDs available on the market.

An increasing number of datacenters these days use both SSDs and HDDs, balancing the high performance of SSDs with HDDs’ ability to store huge amounts of data relatively cheaply. Meanwhile, there is an emerging category of all-flash or hybrid storage systems that either do not use hard drives at all, or use HDDs mostly for things like “cold” archives. Such systems are rather energy efficient and offer high performance thanks to the heavy use of solid state storage.

Nevertheless, when it comes to bulk storage, their requirements are similar to the requirements of datacenters using HDDs: maximum capacity per cubic meter, maximum capacity per watt, high availability, and predictable cost per GB. Viking’s UHC-Silo SSDs were designed for the aforementioned kinds of applications — in some cases, they are going to replace hard drives for huge databases or even “cold” storage, in other cases they are going to sit between “warm” and “cold” storage. Given that many applications may benefit from SSDs, demand for high-capacity flash storage devices is growing in general.

The Viking UHC-Silo SSDs come in a 3.5” form-factor (a rarity for any kind of SSD) and utilize a SAS 6 Gbps interface, two features that make the drives particularly well-suited for replacing high capacity HDDs. The drives are designed for mixed workloads that do not generate more than 1 DWPD and do not require very high performance. The UHC-Silo SSDs offer sustained sequential read/write speed of 500/350 MB/s as well as up to 60,000/15,000 random read/write IOPS, which is in line with other extreme capacity SATA SSDs and is a result of their internal architecture and limitations of contemporary controllers. In fact, given the UHC-Silo’s performance limitations (350 MB/s sustained write speed), it’s impossible to write more than 30 TB of data in a single day. So while the 25 TB version can physically support 1 DWPD over five years (this is what Viking guarantees), the 50 TB model cannot physically support more than 0.6 DWPD. The latter fact essentially means that, assuming Viking’s sustained performance figures are reasonably accurate, the TBW rating of the 50 TB SKU (91.25 PB) cannot be physically exceeded during the warranty period.

The UHC-Silo drives are based on a custom-built proprietary controller that’s paired with eMLC NAND from SK Hynix, which is acquired in wafer form and then cut, tested and packaged in house. In fact, Viking’s proprietary packaging is what enables these drives of rather extreme capacities. SK Hynix officially sells 2048 Gb (256 GB) MLC packages containing 16 128 Gb NAND devices. If Viking used these off-the shelf packages, it would require 100 of them for the 25 TB drive and 200 of them for the 50 TB SSD. Since it is impossible to pack 100 or 200 chips into a 3.5” SSD, Viking uses proprietary NAND packages to build its UHC-Silo drives. The company does not disclose information about its custom NAND chips and does not freely show the internals of the drives because the packaging is one of its key trade secrets.

The proprietary controller that Viking uses supports BCH-based 55 bit/512 byte ECC with end-to-end CRC protection to enable a <1 in 1017 bits read bit error rate. The drives also support data recovery from sector, page and block failure, but for some reason the standard versions listed on the company’s web site do not have any power failure data protection. Meanwhile, since such drives are usually bought for a particular project and are built-to-order, clients may ask Viking to add this feature (and not only this) for an additional fee. Viking stresses that the controller supports NAND from different vendors and thus it can switch between suppliers if it needs to. Viking says that its UHC-Silo SSDs are the highest-capacity SSDs available on the market today. From raw capacity standpoint, the UHC-Silo are exactly what Viking claims them to be: nobody else offers 25 and 50 TB SSDs in a 3.5” form-factor. Meanwhile, Viking had to make significant tradeoffs between performance, capacity, power and compatibility, which is why its drives are considerably slower than some of their direct rivals, such as NGD’s Catalina 24 TB SSD (up to 3.9 GB/s throughput) and Samsung’s PM1633a 15.36 TB drive (up to 1.9/0.9 GB/s read/write). Since there are loads of potential customers requiring massive SSDs for their existing 3.5” SAS backplanes, Viking had to use this interface and could design the drives' capabilities around its limitations. But perhaps most surprisingly, despite the industry-leading capacity of their drives, Viking is keeping the overall prices of the drives relatively reasonable. The actual prices of the Viking UHC-Silo SSDs are not published (remember that they can be customized), but the manufacturer says that the two drives are priced at around $0.40 per GB. This would put the 25 TB drive at approximately $10,000, whereas the 50 TB version would run for $20,000. By contrast, the aforementioned Samsung PM1633a 15.36 TB SSD costs ~$11,300, or $0.73 per GB.

ADATA Launches ISSS314 and IM2P3388 Industrial SSDs: 3D NAND, Extreme Temps

ADATA has introduced two new families of 3D NAND-based SSDs aimed at industrial applications. Dubbed the ISSS314 and the IM2P3388, these drives are designed to handle extreme temperatures as well as humidity levels, allowing them to work reliably in very tough environmental conditions. The more powerful IM2P3388 drives use a PCIe interface and offer high performance levels along with a powerful ECC engine and encryption, whereas the less speedy ISSS314 uses a SATA interface and offers very low power consumption that barely tops 2.5 W.

The IM2P3388: M.2, High Performance, Extreme Temps, Encryption, TCG Opal
The ADATA IM2P3388 is an M.2 drive that uses a NVMe PCIe 3.0 x4 interface and is based on 3D MLC NAND. This specific drive is designed to withstand ESD and EMI, up to 20 G vibration and 1500G/0.5ms shock, extreme temperatures from –40°C to +90°C, as well as high humidity (5%-95% RH, non-condensing). To put it into perspective: the IM2P3388 drives can operate in Antarctica or in the Lut Desert in Iran. In the real world, ADATA’s new SSDs will serve inside space-constrained industrial or commercial PCs, servers, military-grade systems, and embedded computers.

The IM2P3388 drives are based on a Silicon Motion controller that ADATA does not name, we suspect is the SM2260 with some additional customization. As for the NAND, the IM2P3388 SSDs use carefully selected 3D MLC that can handle high temperatures for prolonged amounts of time. The IM2P3388 takes advantage of all the capabilities of the controller and therefore supports AES-256 encryption, TCG Opal 2.0 spec, end-to-end data protection, and so on. In addition, the drive has multiple sensors that monitor its condition.

As for performance, ADATA specifies the drive to offer up to 2.5 GB/s sequential read speeds and up to 1.1 GB/s sequential write speeds (when pSLC caching is used), but does not specify random performance. ADATA’s IM2P3388 will be available in 128 GB, 256 GB, 512 GB, and 1 TB configurations. Keeping in mind the high density of modern flash chips, expect the entry-level models to be slower than their higher-capacity counterparts. In general, expect performance of the IM2P3388 to be comparable to the XPG SX8000 drives featuring the SM2260 and 3D MLC.

The ISSS314: 2.5”, Extreme Temps, Low Power, Starting at 32 GB
The ADATA ISSS314 SSDs come in a traditional 2.5”/7 mm drive form-factor and use a SATA 6 Gbps interface. In order to satisfy the diverse needs of customers, ADATA will offer the ISSS314 in 32 GB, 64 GB, 128 GB, 256 GB, and 512 GB configurations. The higher-end models will provide up to 560 MB/s sequential read and up to 520 MB/s sequential write speeds, whereas the entry-level drives will be considerably slower. As for power consumption, the new SSDs are rated to only use up to 2.5 W, which puts them into the energy efficient category.

The ISSS314 SSDs are based on an unknown controller as well as 3D MLC and 3D TLC NAND memory sorted using ADATA’s proprietary A+ testing methodology to find the higher quality chips. The industrial ISSS314 drives based on 3D MLC memory are rated to withstand shock, EMI, and extreme temperatures from –40°C to +85°C, and thus are aimed at industrial applications. By contrast, commercial 3D MLC ISSS314 SSDs are rated for –10°C to +80°C operation. Meanwhile, the 3D TLC-powered ISSS314 is guaranteed to work in a temperature range from 0°C to +70°C, but can also withstand shocks, ESD, EMI, and so on. As for features, all the ISS314 SSDs have S.M.A.R.T, a temperature sensor, hardware power detection, and flash protection.

ADATA does not publish recommended prices for its industrial and commercial SSDs. Since such products rarely show up in mainstream retail, their actual prices for customers typically fluctuate depending on the order size and other factors.

Micron Introduces 9200 Series Enterprise NVMe SSDs

Today at Flash Memory Summit, Micron is announcing their next generation of high-end enterprise NVMe SSDs. The new Micron 9200 series is the successor to last year’s 9100 series and uses Micron’s 32-layer 3D TLC NAND flash and a new generation of Microsemi SSD controllers. As with the 9100 series, Micron’s 9200 series covers a wide range of capacities, but adds a third tier of write endurance: ECO joins the PRO and MAX tiers, respectively aimed at read-heavy workloads, mixed workloads, and write-intensive workloads.

The Micron 9200 series will be available in either 2.5″ U.2 form factor or PCIe add-in card. Thanks to the new generation of SSD controllers, the add-in card version can now use a PCIe x8 interface and offer significantly higher sequential access performance than the U.2 version, with read speeds reaching up to 5.5GB/s. The range of capacities is also far different from the 9100 series, which topped out at 3.2TB for the 9100 PRO and 2.4TB for the 9100 MAX. The 9200 MAX now offers up to 6.4TB, the PRO up to 7.68 TB, and the new 9200 ECO is available in 8TB and 11TB capacities.

Micron’s enterprise SATA SSD lineup moved to 3D TLC NAND early this year with the introduction of the 5100 series. Micron’s 7100 series of lower-power enterprise NVMe SSDs has not been replaced with a 3D NAND-based successor and it appears Micron is phasing out the current generation.

ADATA Launches XPG SX9000: 2.8 GB/s Seq. Read, Marvell Controller, Up to 1 TB of MLC

ADATA has announced its new SSD aimed at the very high end of the market. The new flagship XPG SX9000 drives are based on the Marvell 88SS1093 BTB2 controller and are paired with Toshiba’s 2D MLC NAND flash memory. Later on, the company plans to switch to Toshiba’s 3D MLC NAND for a product that will succeed the SX9000 SSD series.

The ADATA XPG SX9000 SSDs use the Marvell 88SS1093 BTB2 controller, which sports three processor cores and 8 NAND channels, with 4 banks per channel for 32 targets in total. The IC is an improved version of the 88SS1093 with higher frequencies and performance to boost speeds of higher-end SSDs. The 88SS1093 BTB2 supports a Marvell’s third-generation ECC technology based on the LDPC algorithm and uses PCIe 3.0 x4 interface.

The new XPG SX9000 drives are to be available in 256 GB, 512 GB, and 1 TB configurations in the M.2-2280 form-factor. The SSDs use DRAM buffers for additional performance, and come with a very basic heat spreader to further prop up performance in systems that provide adequate cooling. Speaking of performance, ADATA promises up to 2.8 GB/s sequential read speed as well as up to 1.45 GB/s sequential write speed for the top-of-the-range 1 TB model. As for random read/write performance, ADATA lists 310K/240K IOPS for the most advanced model.

Reliability is another thing that ADATA is taking serious when it comes to the XPG SX9000. The drives are rated for up to 1 PBW (terabytes to be written) and two million hours MTBF, which in turn is coupled with a five-year warranty.

ADATA has not set recommended prices of the XPG SX9000 series just yet. What we do know is that the drives are hitting the shelves in the coming weeks and expect their prices to be competitive against the obvious rivals — the Samsung 960 Pro and the Samsung 960 Evo families of SSDs.

Otherwise, as previously stated, ADATA is also looking at releasing 3D NAND versions of the drive farther down the line. 3D NAND has a number of advantages over 2D NAND, but it’s not ideal for all possible applications at the moment, particularly due to its high density, which conflicts with the need for multiple NAND packages to maximize parallelism and performance on high-end SSDs. All things considered, this is why ADATA decided to go with a new Marvell controller as well as Toshiba’s 2D MLC NAND for the XPG SX9000 SSD. Eventually, the company promises to use the same controller for a high-end 3D NAND-powered drive, but that is something that is going to happen towards the end of the year at best.

CPU Buyer’s Guide: Q2 2017

In our series of Buyer Guides, here’s the latest update to our recommended CPUs list. All numbers in the text are updated to reflect pricing at the time of writing (7/6). Numbers in graphs reflect MSRP.

CPU Buyer’s Guide: Q2 2017
So far the first half of this year has been mildly insane. Back in 2016, we had the best part of 1.5-2 platform launches and it was a quiet year on the CPU side. So far this year we have had three or four launches, with another few in the pipeline to come. With Kaby Lake, Ryzen 7 and Ryzen 5 out of the way, in comes Kaby Lake-X and Skylake-X to the party, with Skylake-SP, ThreadRipper, EPYC and potentially more still on the invite list later this year. As much as I love writing about CPUs and testing the newest hardware, sometimes a doorman needs a rest (ed: 2018… maybe).

For consumers, it can be a fun time. With new platforms comes an opportunity to upgrade, either through the increase in performance or just because you want the latest and greatest. The idea is that the newest processors are more performant, or lower power, or fit into a particular niche better (and hopefully are the same or lower cost overall). For users who wanted to invest in AMD this year, Ryzen has been a good offering and ThreadRipper is around the corner. For users looking to upgrade that i7-2600K or i5-2500K, Intel is trying hard to tempt them with Kaby Lake processors and even Kaby Lake-X, with the best part of 25-35% IPC and some extra MHz as well. For anyone that wanted a 10-core CPU and thought $1721 was too much for the i7-6950X, Intel has you covered with the Core i9-7900X at $999-$1049 now as well.

CPU Reviews
Our big CPU reviews for 2017 have covered all the launches so far, and well worth a read:

The Intel Core i7-7700K (91W) Review
The Intel Core i5-7600K (91W) Review: The More Amenable Mainstream Performer
The Intel Core i3-7350K (60W) Review: Almost a Core i7-2600K

The Intel Skylake-X Review: Core i9 7900X, i7 7820X and i7 7800X Tested
Intel Announces Kaby Lake-X Processors

The AMD Zen and Ryzen 7 Review: A Deep Dive on 1800X, 1700X and 1700
The AMD Ryzen 5 1600X vs Core i5 Review: Twelve Threads vs Four at $250
In the pipe coming in the next few weeks include our Kaby Lake-X reviews (unfortunately one of our CPUs died, hence the delay), as well as retesting Ryzen with the latest updates, Skylake-X on gaming, and a few more projects underway.

One of the overriding issues so far this year worth mentioning is platform maturity. With new platforms come new challenges, and as far as we understand, extreme deadlines. Motherboard manufacturers, for both AMD and Intel, have had to rush through some of the production of their initial motherboards at launch. When we reviewed Ryzen 7 and Kaby Lake-X, both of those reviews did not have gaming results due to erroneous results on young hardware. At this point we expect both platforms to be running smoothly, but as an indication that this year is about time to market, it’s a big one to note for early adopters (and reviewers that end up wanting to throw products out a window).

The majority of our recommendations aim to hit the performance/price curve just right, with a side nod to power consumption as well. Here’s a breakdown of those recommendations:

Peak Gaming / VR
In the midst of the launches this year, the talk of CPUs that are suitable for Virtual Reality has died down to some extent. Now that AMD has parts on the shelf that are unquestionably suitable, it just comes down to what price a user can enter into VR, or at what level a user can be future proof as VR gaming becomes more demanding. Even with this is mind, a non-VR gaming machine that wants to be ahead of the curve has similar demands, especially as DirectX12 titles are in the pipeline. Single thread performance still helps here, especially for the simpler casual games and driving high frame rates.

The new king of the crop is the Intel Core i7-7740X. It boasts the highest per-core performance of any x86 processor, and then heaps on a lot of frequency as a result. A good processor will run up to 5 GHz with a nod for overclocking, giving a user the best premium VR experience today. At $350 list price, plus some more for a good cooler and a decent motherboard, it should provide a premium gaming system for several years to come when a user wants the Peak experience.