Connect with us

Peripherals

Whether PCI Express 4.0 is an advantage Ryzen 3000? Check on NVMe SSD

Advertisement
Code: GKB014D. Only $162.99 for 128Bit Desktop Gaming Graphics Card With HDMI DVI DP Interface with $37 Coupon Code GKB014D!

One of the key areas in the marketing strategy to promote AMD processors Ryzen 3000 was the focus on appeared to them, supporting high-speed interface PCI Express 4.0. Indeed, the new Ryzen, built on microarchitecture Zen 2, became the first consumer-level processors that have received this high-speed interface, and, moreover, none of the current Intel platforms of the item of PCI Express 4.0 in the list of features to date has not. Do not use this reason for the protrusion of the progressivity of its products to AMD, of course, could not, and therefore, as in the time of the announcement Ryzen 3000, and after him the representatives of the company did not miss a single case to trump the appropriate opportunity.

The arguments in favor of AMD’s new tires, which allows you to select devices interface with double greater than before throughput was reduced to two theses: “it’s good for graphics cards” and “this is good for SSDs”. However, to be honest, their reasoning is flawless no different.

Speaking about the benefits of moving to PCI Express 4.0 for graphics cards, AMD representatives refer to synthetic test of the capacity of 3DMark, which really shows 69-percent performance improvement when using higher speed version of the bus.

However, this should make two large but which AMD is silent. First, support PCI Express 4.0 only selected video cards, related to the Radeon RX series RX 5700 and 5500, while members of the family GeForce quite successfully cost traditional interface PCI Express 3.0. Secondly, if we talk about real the games load, it obviously does not generate such volumes of data that would not be enough bandwidth common graphic bus PCI Express 3.0 x16, reaching 16 GB/s. moreover, even the use of interface PCI Express 2.0 lowers the frame rate in popular games in 4K resolution on a few percent, and the difference in GPU performance when connecting to the CPU when using the third and fourth version of the tyre in the worst case is a few tenths of a percent.

Transfer SSD to work through the PCI Express 4.0 seems to be a much more sensible step, because modern NVMe SSD really getting to the limit of bus bandwidth PCI Express 3.0 x4. Increase the speed of the interface by adding PCI Express would be too easy a way out, since consumer SSDS are firmly accustomed to the form factor M. 2, which allows to bring to drive no more than four lines. But switching to PCI Express 4.0 can be easily held within the framework of existing ecosystems, and this action seems quite sensible, because it allows you to raise the bar of peak velocities from 3.9 to 7.9 GB/s. this is the point AMD marketing materials, which the company says almost polutorakratnoe increase disk subsystem performance in processor systems Ryzen 3000.

But it’s not so simple. First and foremost you need to understand that we are talking exclusively about the growth of linear velocities, while in the responsiveness of solid state drives due to the transfer at a higher speed interface is definitely not going to win here because this is not determined by external bus, and their internal architecture and capabilities of the flash memory. That is, in the average common scenarios such impressive growth, which is depicted on a marketing slide from AMD, of course, will not.

It is important to take into account another important point: the leading SSD manufacturers do not support PCI Express 4.0 in their products. So if you want to install in your Ryzen 3000 system NVMe drive with the new high-speed interface will have to choose among the proposals of the second-the third echelon. And moreover, all of the SSDS with support for PCI Express 4.0 x4 are similar and rely on the same controller PS5016-E16 independent developer Taiwan’s Phison, which engineers have been unable to mark any notable achievements in the field of building high-performance platforms for consumer SSDS. In other words, much optimism these SSD do not cause.

However, we are not accustomed to rely on some value judgment and everyone is trying to check in practice. Moreover, the question of what kind of SSD is to recommend to a modern system based on Ryzen 3000, seeking to squeeze the maximum out of it possible, not idle. Drives based on controller Phison PS5016-E16 PCI Express x4 4.0 finally made it to local stores. So, a lot of buyers decided to opt for a relevant and promising the AMD platform, you rise before a choice: whether to establish a system of time-tested performance SSD with a PCI Express 3.0 interface; whether to give in to the propaganda AMD and Phison and try to get better performance, contact the new products with the support of PCI Express 4.0. In this article we will try to find out what is better after testing both options in real conditions.

#Why are we even talking about PCI Express 4.0

The PCI Express 3.0 specification was approved in November 2010, and it is not surprising that to date, this tire is firmly entrenched as the base interface for interfacing computing devices. Especially helped in this by two factors: the protracted development of the next version of the standard, on the one hand, and the absence of a request for speed beyond the capabilities of PCI Express 3.0 on the other. However, gradually the industry came to realize the need for further progress, what especially helped with SSDs. The rapid growth in this industry has led to the fact that the SSD was able to rest against the speed limitations of the existing interface, and the output to mass-market PCI Express 4.0 really made sense.

The PCI Express 4.0 was adopted in June 2017. It has improved flexibility and scalability, for example, considered the possibility of a physical realization not only in the form of slots and tracks on the Board, but in the form of external connectors and cables; added additional control commands for the purposes of energy conservation; and most importantly — doubled the speed of data transmission. Equally important, all of these improvements do not entail any fundamental changes in the Protocol. A two-fold increase in throughput achieved by the simple increase in the frequency of data transmission, keeping the old coding scheme of the signal. As a result, between the new and old standards maintained both direct and backward compatibility, and hence, hosts and terminals that support different versions of PCI Express can without any problem to work with each other.

Data exchange via the bus PCI Express 3.0 occurs with a frequency of 8 billion transfers per second, and PCI Express 4.0, this frequency has already reached 16 billion transfers per second. Thus, while the bandwidth of one PCI Express 3.0 lines amounted to 985 MB/s (including 128b/130b encoding) in the PCI Express 4.0 it has doubled – up to 1969 MB/s. Further naturally expanded bandwidth all standard for mass PC tires. Speed of PCI Express x4 4.0 increased to 7.9 GB/s, and PCI Express x16 4.0 was able to transmit data with the maximum capacity of 31.5 GB/s (in each direction).

  Bandwidth, GB/s
×1 ×2 ×4 ×8 ×16
PCI Express 1.0 0,25 0,5 1,0 2,0 4,0
PCI Express 2.0 0,5 1,0 2,0 4,0 8,0
PCI Express 3.0 0,98 Of 1.97 3,94 Of 7.88 15,75
PCI Express 4.0 Of 1.97 3,94 Of 7.88 15,75 31,51

The new version of the platform Socket AM4 using the chipset X570, was the first and only point of the presence of PCI Express 4.0 in the mass systems. Thanks to it, users can build a system with a new high-speed bus today, but to test this high-speed tire, you need three components. First, the processor family Ryzen 3000, PCI Express controller, which supports the fourth version of the Protocol. Second, the motherboard chipset AMD X570, which has optimized wiring signal lines capable of guaranteeing the integrity of information during transmission at the high frequency bus. And third, a terminal device with support for PCI Express 4.0, which can be either a graphics card or, as in our case, the SSD. In the absence of one of these components, the system will still work due to the two-way version compatibility PCI Express, but in a slower-than-4.0, mode.

Thus, the spare tire is still in practice will only in a very small number of desktop systems. Gradually, however, Park platforms with support for PCI Express 4.0 will be expanded. So, in the coming months AMD plans to add this interface in one of your set logic – B550. This will reduce the entry price threshold for systems that support wireless high-speed tires, but the processor family Ryzen 3000 is still needed.

As for Intel, it will support PCI Express 4.0 it’s very easy. Microprocessor giant somehow missed the moment when the new specification was ready, and now far behind AMD in terms of its implementation. Clearly speaks for itself the fact that even in the server segment support PCI Express 4.0 at Intel could not appear before the middle of next year. As usual personal computers on the Intel platform, they speed the tire “Shine” very soon, as planned for 2020 Comet Lake processors and chipsets of the 400 series provides just the usual PCI Express 3.0, and further the company’s plans look very uncertain.

In other words, Intel will soon be playing the role of a brake of progress, and we can say almost certainly that this implementation of PCI Express 4.0 will not happen as rapidly as one would expect. For example, if we talk about SSD at the moment supports this high-speed bus with only a single controller Phison development, and the emergence of any alternatives in the foreseeable future is not expected. Especially skeptical to the idea of issue under NVMe SSD interface PCI Express 4.0 in the near future are the leading manufacturers of scale Western Digital and Samsung. Apparently, they intend to wait for the time when PCI Express 4.0 will appear in the Intel platform, and the mass of the solution with its support will be offered only in anticipation of this event.

That is why today we have to talk about the usefulness of PCI Express 4.0 for consumer NVMe SSD using as an illustration just one example – built on the controller Phison PS5016-E16 drive Gigabyte Aorus Gen4 NVMe SSD.

#The main hero – Gigabyte Aorus Gen4 NVMe SSD

Although we rendered in the subtitle of this section, the name of the drive Gigabyte, who visited us in the lab, everything said below is equally true for any other NVMe SSD controller Phison PS5016-E16, and even now any other available mass storage device under the PCI Express 4.0, as long as the market does not appear the nominal support of this promising interface. To make this generalization allows us production strategy Phison, not much over the past years, no changes. The fact that the developer of controllers offers its partners not chip and SSD with them. Therefore, the diversity of models on the basis of Phison PS5016-E16 – apparent. In fact, all these products are made one of the patterns on the same plant and differ only in the names, bodywork and accessories, such as cooling systems.

This means that any currently available drive with support for PCI Express 4.0 has a completely predictable hardware configuration: it will be based on the controller PS5016-E16, which manages the TLC 3D NAND-memory, belonging to the class BiCS4. This memory has a 96-layer design and comes Toshiba. She, by the way, can be seen not only in the drives with interface PCI Express 4.0, but in some other models NVMe SSD with a traditional interface, for example, in Kingston KC2000. But for a drive that claims to be a significant improvement in speed performance, the use of a three-bit BiCS4-memory – point of principle. Such flash memory is faster than previous versions and is able to better reveal the potential of the external bus with twice the bandwidth.

Of course, to increase the linear velocities provided by the array of flash memory, there is always an extensive way – the increase in the number of channels. But the platform Phison PS5016-E16 continues to rely on the traditional eight-channel scheme. And this is quite logical: adding channels would considerably complicate the circuitry of the drive, and the design of the controller. The same Phison engineers in the development of its first PCI Express 4.0 controller treated it as a transitional solution, and therefore not particularly strained. Architectural chip PS5016-E16 is very similar to the PCI Express 3.0 controller PS5012-E12 — with some improvements in the firmware and only upgraded the hardware unit responsible for the implementation of the external interface.

The rest of Phison PS5016-E16 does not differ almost from the predecessor, with which we are acquainted in detail on the example of Silicon Power P34A80 and the Gigabyte Aorus RGB M. 2 NVMe SSD. It represents the average processing power of a chip based on dual-core 32-bit processor ARM Cortex R5, which offers eight channels to connect a three-bit or four-bit flash memory with 800-MHz interface. Also present in the controller DRAM controller to operate the cache memory in which you can be a DDR4 SDRAM. In addition, it has the usual set of proprietary technologies Phison, including a special co CoXprocessor serving technology and Express entry algorithm for LDPC encoding and error correction of the fourth generation, hardware-based encryption scheme AES256 and so on.

It is significant that in reality power Couple PS5016-E16 to completely fill the bandwidth of the four PCI Express 4.0 is not enough. The official specs of this chip is the maximum speed of the linear reading of 5.0 GB/s and write is 4.4 GB/s, which is a third below the capabilities of the tire. Peak the performance of random operations expected close to the PS5012-E12 and it is 750 and 600 thousand IOPS when reading and writing respectively. In other words, even the Phison is no reason to think that on the basis of PS5016-E16 can collect some really high-performance NVMe SSD, which could be attributed to a number of flagship of new generation solutions. However, there was no choice — the other options are to build a drive that supports x4 PCI Express 4.0 for today simply does not.

The result is to advance the PCI Express 4.0 at the mass market end manufacturers of drives with AMD had to go to a specific fraud and try to build around the platform Phison E16 halo high-end product, which she from a technical point of view can not be. Appears pseudolites, for example, that we have chosen to test the Gigabyte Aorus Gen4 NVMe SSD costs a lot, comes in a large neatly decorated simply packaging and comes with a massive and a quality cooling system made of bars of copper. When I take this device, really starting to think that this is a premium, high performance solid state storage systems to the highest level.

For tests, we took the version of the Gigabyte Aorus Gen4 NVMe SSD capacity of 2 TB. This is the maximum capacity version of this drive. Since the 96-layer crystals Toshiba BiSC4 have a capacity of 512 GB, an array of flash memory 2-TB drive is formed of 32 NAND devices. This, in turn, means that the channels of the controller Phison PS5016-E16 used a fourfold rotation, which ensures the best performance. In other words, for example, the Gigabyte Aorus Gen4 NVMe SSD we will see the maximum of what a capable platform Phison PS5016-E16.

Gigabyte Aorus exposed NVMe SSD Gen4 with the removed radiator looks completely ordinary. This is the familiar gaze of the enthusiast M. 2 card form factor 2280 with a key of type M in stab connector. To understand in her appearance that the drive is able to use four lines of PCI Express 4.0, it is absolutely impossible. However, we should not forget that due to backward compatibility, this SSD can be installed in the system without the support of the new version of the interface: in this case it will use the external bus mode 3.0.

On close acquaintance with the Gigabyte Aorus NVMe SSD Gen4 immediately evident that its circuit Board is virtually identical layout from Gigabyte Aorus RGB M. 2 NVMe SSD on the basis of the last controller PS5012-E12. And this means that the new PCI Express 4.0 controller Phison PS5016-E16 not only inherits from its predecessor, most of the domestic units, but also working with him on the conclusions.

But the controller itself PS5016-E16 look at this very familiar. Its upper surface is closed with Nickel-plated copper plate, then chips, Phison has not previously been observed. The explanation is simple: despite the fact that the controller is made according 28-nm process, it reaches the dissipation under load 6.7 W, and a metal plate may support the heat sink.

 

On-Board SSD in question has four chip flash memory marked TABHG65AWV. They sliced and packaged in the factories of the Taiwanese PTI from silicon semi-Toshiba: this memory – it is a typical option for drive controllers Phison. Usually, the use of such “non-native” chipsets makes us question the quality of the flash memory, because of its selection and sorting in this case is responsible not Toshiba itself, as a mediator. But drives with the Phison controller PS5016-E16, which is now on the market have a high of the declared resource, which for a 2 TB version reaches 3.6 Pb. Apparently, for a flagship product Phison purposefully selects the flash memory of the higher grades.

In addition to the flash memory and the controller on both sides of the drive located on the chip DDR4-1600 SDRAM. Used Gigabyte chips production SK Hynix, that is the amount of memory required for the controller to work effectively with the table of address translation on the Gigabyte Aorus Gen4 NVMe SSD 2 TB is a typical 2 GB.

#Antagonist – 970 Samsung EVO Plus

As products based on the controller Phison PS5016-E16 aiming to occupy the top rung in the hierarchy of consumer NVMe SSD, compare Gigabyte Aorus NVMe SSD logical Gen4 is the fastest and best models use to communicate with the system familiar to the bus PCI Express 3.0. So we thought it fair to oppose the storage Gigabyte is the most popular NVMe SSD Samsung EVO 970 Plus, the benefit of his 2-TB not so long ago appeared on the market.

Give details about the inner workings 970 Samsung EVO Plus there is not much point – we’ve done it several times already. It is therefore only briefly recall the fundamental details. Based on 970 Samsung EVO Plus lies Samsung’s proprietary controller Phoenix, which, as shown, has a good performance potential and phenomenal flexibility. On the first point, it is sufficient to mention that it locked up pachydermia ARM-processor, and the second is to remind you that this controller is included in all contemporary NVMe SSD South Korean manufacturer, including models for MLC- and even on SLC-memory.

After the second controller component, which makes from 970 Samsung EVO Plus fast model, this is a special flash memory. It found the use of corporate TLC 3D V-NAND, Dating back to the fifth generation. Crystals such memory was a monolithic 90-layer design and significantly improved performance characteristics and energy consumption. The result is 970 Samsung EVO Plus long and firmly holds the title of fastest mass NVMe SSD with TLC memory.

 

Interestingly, apparently 970 Samsung EVO Plus capacity 2 TB completely different from the drives of the same model, but smaller capacity. This means that no sophisticated cooling device is not provided here and the manufacturer believes that the overheating protection will be enough a label with an inner layer of copper foil glued to the back side of the drive. Which, incidentally, does not carry any chips.

The uniqueness of the 970 EVO Plus 2 TB is that Samsung managed to gather such a capacious SSD on a unilateral M. 2 2280-Board and whole array of the flash memory is placed just two chips. “Stuff” in a single chip a terabyte of flash memory Samsung has managed through the use of crystals with a size of 512 GB, which are stacked in 16 pieces – able to do so only a few manufacturers. Another chip on the 970 EVO Plus 2 TB chip LPDDR4 SDRAM 2GB, so that all components of the drive are four circuits (including the controller).

As for performance, version 970 EVO Plus 2 TB formally, even a bit faster modifications younger volume. But this is manifested only in the performance melkoplodnyj operations, which is stated at 620 and 560 thousand IOPS when reading and writing respectively, which is slightly lower than the promise drives on the controller Phison PS5016-E16. Linear speed, of course, limited to the bandwidth of the PCI Express 3.0 interface and is 3.5 and 3.3 GB/s for reading and writing. In other words, on paper it looks like Samsung EVO Plus 970 vs Gigabyte Aorus Gen4 NVMe SSD has no chance. However, we are not just caught on Phison inflating passport numbers, so do not be surprised if in practice the picture is not the same.

What 970 Samsung EVO Plus definitely plays a daring competitors with support for PCI Express 4.0 is in terms of warranty service: the South Korean manufacturer is not ready to allow users such as hard to load your drives. The declared resource of 970 EVO Plus 2 TB is only 1.2 Pb – three times less than that of the SSD based on Phison controller PS5016-E16.

#Specifications table

All the promises Phison and its partners regarding the performance of the new generation PCI Express 4.0 drives based on controller PS5016-E16, we have compiled the following table, where the passport specifications Gigabyte Aorus Gen4 NVMe SSD 2 TB are compared with the characteristics of 970 Samsung EVO Plus a similar volume. And let such data are usually poorly correlated with actual performance SSD, but it is clear bluster, armed with which manufacturers drives on the platform Phison E16 trying to convince the owners of the new Socket AM4-systems in the rationality of choice to use in conjunction with the processor family Ryzen 3000 such media.

Manufacturer Gigabyte Samsung
Series Aorus Gen4 NVMe SSD 970 EVO Plus
Model number GP-ASM2NE6200TTTD MZ-V7S2T0
Form factor M. 2 2280
Interface 4.0 PCI Express x4 NVMe 1.3 PCI Express 3.0 x4 NVMe 1.3
Capacity, GB 2000 2000
Configuration
Memory chips: type, interface, process technology, manufacturer Toshiba 96-layer 512 GB TLC 3D NAND (BiCS4) Samsung 90-layer 512 GB TLC 3D V-NAND
Controller Phison PS5016-E16 Samsung Phoenix
Buffer: type, volume DDR4-1600, 2 GB LPDDR4, 2 GB
Performance
Max. sustained sequential read, MB/s 5000 3500
Max. sustained sequential write speed, MB/s 4400 3300
Max. speed of random read (blocks 4 KB) IOPS 750 000 620 000
Max. speed random write (blocks 4 KB) IOPS 700 000 560 000
Physical characteristics
Power consumption: inaction/read-write, W 0,02/6,6 0,03/6,0
MTBF (mean time between failures), million h 1,77 1,5
Resource record, TB 3600 1200
Dimensions: D × H × D, mm 23,5 × 80,5 × 11,4 22,15 × 80,15 × 2,38
Weight, g 95 8
Warranty period, years 5 5

#Features SLC-caching

Every time we talk about the performance claimed by the manufacturers, we have to make a disclaimer that they belong to the SSD with SLC caching, in which all accesses to the flash memory occur in a single bit mode. In some cases this is appropriate, since many SSDs offer SLC cache is quite sufficient for very extended operations, but it is still not true for any consumer NVMe SSD. For example, disk drives based on controllers Phison, always suffered from the fact that SLC cache they worked on a static algorithm and had a very small size, resulting in users often had to face a steep drop in the speed when copying or writing large files.

Fortunately, the firmware for the controller PS5016-E16 developers Phison finally was able to fix the old problem and for the first time implemented a progressive dynamic algorithm SLC-caching. This means that the information is written to the drive in SLC mode, as long as it allows an open space, and transfer of cells TLC 3D NAND to established a three-bit mode in idle moments. In other words, high speed recording, the Gigabyte Aorus Gen4 NVMe SSD you can not worry – it really can be achieved for very large amounts of data.

In this respect, the drive platform Phison E16 are the envy of even 970 Samsung EVO Plus is not as effective combined scheme of Express entry Intellegent TurboWrite. For example, in dvuhterabaytny version 970 EVO Plus SLC cache consists of a 6 GB static and variable dynamic, but its maximum volume even on a completely clean drive is limited by the size of 78 GB. At the same time in the Gigabyte Aorus Gen4 NVMe SSD of the same capacity in the accelerated mode, it is possible to fill the third space, that is, on an empty SSD capacity SLC cache can reach 667 GB.

To illustrate all this very easy using the graph of maximum continuous speed of sequential write on a clean drives Gigabyte Aorus Gen4 NVMe SSD 2 TB Samsung EVO 970 Plus 2 TB.

Here you can really see how much more information may take the Gigabyte Aorus NVMe SSD Gen4 2 TB, compared to 970 Samsung EVO Plus 2 TB at its specification speed. Although in fairness it should be noted that of the 970 EVO Plus the size of the SLC cache is that to go beyond it in real conditions will not work in 99.9% of cases. But even if we talk about the performance of a linear recording in the SLC mode, the drive platform Phison E16 looks more attractive thanks to the support of PCI Express 4.0. Here its relevance is undeniable: the high-speed interface allows you to write information to the drive faster as eight-channel array of the flash memory in the SLC mode has a greater bandwidth than four lines PCI Express 3.0.

However, in a native TLC mode, the array of flash memory in 970 Samsung EVO Plus is much more productive. The storage capacity of 2 TB, it provides the write speed at 1.3 Gbytes/s, while in the Gigabyte Aorus Gen4 NVMe SSD 2TB write speed after filling in SLC-Kesha drops to 530 MB/s. However, this result is unlikely to have any practical value: we’re talking about the drives for desktops, and in such systems there are no tasks that require continuously keep on to the media hundreds of gigabytes as fast as possible.

Another thing – read operation. Phison is among the manufacturers that have mastered the optimization of its products under benchmarks. It lies in the fact that the files written to the drive, remain in the SLC cache for some time in order to appeal to them to run after recording, were served faster. This simple technique brings no benefit during actual operation, but strongly influences the test results because the measurement of speed of reading from the file immediately after creating such a file is a typical scenario in which in almost all benchmarks.

To illustrate how it works, this magic trick is very simple. The following graph shows the speed of random melkoplodnogo read data from a file on the Gigabyte Aorus Gen4 NVMe SSD 2 TB Samsung EVO 970 Plus 2 TB as soon as it is created, and once on the SSD after the test file contains some more information.

It is clearly seen that in 970 Samsung EVO Plus fresh file is immediately sent from the SLC cache to the main array of TLC-memory storage provides for him the speed melkoplodnogo read queue of requests at the level of 46.5 MB/sec. the same Drive platform Phison E16 saves the new file in the SLC cache for some time, and thus the speed of reading random blocks is phenomenally high, at around 56 MB/s. But not for long. If after the test file to the drive to save some amount of information, he will move into the bulk of TLC memory and speed melkoplodnogo access drops to 40 MB/s – that such a level of performance and will deal with the real users.

All this means the following: don’t believe the popular benchmarks like CrystalDiskMark. They NVMe SSD controller Phison PS5016-E16 will show impressive numbers, but these numbers do not actually describe the real performance, but just to illustrate that complex optimizations, which is in drives like the Gigabyte Aorus Gen4 NVMe SSD. In fact, view themselves (left – score, the Gigabyte Aorus Gen4 NVMe SSD 2TB; right – 970 Samsung EVO Plus 2 TB).

Gigabyte Aorus NVMe Gen4 SSD 2 Тбайт

Gigabyte Aorus NVMe SSD Gen4 2TB

 
Samsung 970 EVO Plus 2 Тбайт

970 Samsung EVO Plus 2 TB

It seems that the Gigabyte Aorus Gen4 NVMe SSD won over 970 Samsung EVO Plus 2TB landslide victory, defeating his on all graded in CrystalDiskMark performance. But don’t jump to conclusions: we will show that in real conditions of storage on the basis of Phison PS5016-E16 is not as good as it first seems, and certainly can not boast of an indisputable superiority over 970 Samsung EVO Plus.

SOURCE

Advertisement
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.
Code: FASTMWP50. You must use promocode. The promotion applies to Managed Word Press. The promotion is available for all customers of the store.
Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Peripherals

Overview NVMe drive Silicon Power P34A80: reference Phison E12 and the new firmware

Advertisement
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.

Until recently, we too often took for related solid-state drives offered by companies from the second or third tier. This was due primarily to the fact that these products are too unpredictable in hardware platform: depending on what the chip manufacturer was able to purchase on the open market, today the particular model of drive could be used some components, and tomorrow – quite different. The producers usually do not consider it necessary to inform users about changes in product configuration nor on your website nor in the name or on the label. A result of doing the reviews of these SSD was kind of pointless and sometimes even harmful. It is written article threatened to lose its truthfulness at any time, and we could not even know.

But it has the ability to change, and in the case of drives with NVMe interface situation with the permanence of the hardware now looks much better. The fact that the main independent developers controllers, Phison and Silicon Motion, offer for progressive NVMe SSD relatively small number of platforms, and they clearly ranked according to the level of performance and price. Therefore, the space for throwing the producers of second and third tier are left, and their NVMe products receive a fixed and deterministic components. In other words, NVMe SSD, even produced the most savvy firms, no mess with the controllers and memory are not observed.

Besides familiarity with the products of manufacturers of the second-third tier has now acquired a distinct sense also for the reason that leading players in the SSD market have switched to using either completely new or highly customized versions of public modules. Therefore, the SSD supplied by smaller companies today have little in common with the production scale of the giants of Samsung, Western Digital, Intel, or Crucial. But their characteristics make them believe that the performance and reliability they do not yield to the suggestions of market leaders. That means they may deserve to be considered separately.

At the moment, independent developers controllers offer two fully ready to implement the platform for productive NVMe SSD interface PCI Express 3.0. First – SMI SM2262EN, and it is actively used in their drives, ADATA, HP, Kingston and Transcend. Second – Phison E12. This controller appeared a little later, but managed to gain wider popularity. It can be found in the products of Corsair, Gigabyte, Lite-On, Patriot, PNY, Seаgate, Silicon Power, Team Group and many other smaller firms. It is about the realization of this platform in the end products we will discuss in more detail today.

But before we get to the point of the story, it is necessary to clarify that the scheme of work Phison over the last few years have not changed. In fact, this company supplies to its partners, not the chip controllers and ready drives in the collection, and end-producers can only add to them radiators, marketing stickers and your own packaging. From this it follows that the fundamental difference between drives of different brands on the basis of Phison E12 does not exist: differences if there is, only in the exterior. Accordingly, testing one drive on this platform, we can summarize the obtained results and all related products.

In the laboratory 3DNews already visited one of the drives with this controller – the Gigabyte Aorus RGB NVMe M. 2 SSD, and he made a special impression. Nevertheless, we decided to give the platform Phison E12 second chance. The fact that the developers have improved the Phison firmware, and the effectiveness of the controller was to increase. To check, whether so it actually, we took another test NVMe SSD based on Phison E12 platform. They found Silicon Power P34A80.

#Specifications

So Silicon Power P34A80 – drive model, built on the controller PS5012-E12, the first chip Phison for NVMe SSD, which can be used where the user is interested in something else besides low prices. Engineers Taiwanese developer could not design some really productive solution, but with the E12 platform, it seems, something began to happen. At least on the website Phison controller PS5012-E12 promised linear speeds exceeding 3 GB/s, while the previous child of this developer, PS5008-E8, with difficulty squeezed only 1.6 and 1.1 GB/s for read and write respectively.

Since the new Phison chip is noticeably faster than previous solutions, it is no surprise that domestic architecture he finally approached the modern controllers for productive NVMe drives other developers. So, PS5012-E12 is based on a dual-core ARM processor with a frequency of 667 MHz, offers eight channels to connect flash memory, supports four lines PCI Express 3.0 to communicate with the system and provided with DDR4/DDR3-interface to implement the internal buffer. The actual chip supports standard NVMe 1.3, and in addition, it implements error correction based on LDPC codes, the cryptographic algorithm AES-256 and a number of proprietary technologies Phison to improve the reliability of data storage. For the production of chips PS5012-E12 meets TSMC, is used to manufacture 28-nm process technology.

In Silicon Power P34A80 this controller works in tandem with flash memory Toshiba, which is quite natural, because between the Phison and Toshiba has a long and close relationship. In this case, the memory array is composed of 64-layer crystals BiCS3 (TLC 3D NAND), what makes considering the drive to any other NVMe SSD based on Phison E12 platform.

Allocates the same Silicon Power P34A80 newer firmware version 12.3: all the same drives that wound up in our hands before, used the firmware versions 12.1 or 12.2. Of course, that doesn’t make this SSD is unique. Actually Phison started to distribute the firmware 12.3 among its partners in the middle of summer, but the transfer of serial products to a newer firmware takes time. Silicon Power has been able to resolve this issue more quickly, but owners of other drives based on the controller Phison PS5012-E12 can preupgrades independently online, it is easy to find information on how to do it.

In passport characteristics of the Silicon Power P34A80 foolish to look for any surprises, everything is exactly the same as in the other SSD on the platform Phison E12 — the only difference is that the speed performance of their product manufacturer to the max caught up with fog.

Manufacturer Silicon Power
Series PCIe Gen3x4 P34A80
Model number SP256GB P34A80M28 SP512GB P34A80M28 SP001TB P34A80M28 SP002TB P34A80M28
Form factor M. 2 2280
Interface PCI Express 3.0 x4 NVMe 1.3
Capacity, GB 256 512 1024 2048
Configuration
Memory chips: type, interface, process technology, manufacturer Toshiba 64-layer 256-Gbit 3D-NAND TLC (BiCS3)
Controller Phison PS5012-E12
Buffer: type, volume DDR3L-1600,
256 MB
DDR3L-1600,
256 MB
DDR4-2400,
1024 MB
DDR4-2400,
2048 MB
Performance
Max. sustained sequential read, MB/s Up to 3400 MB/s
Max. sustained sequential write speed, MB/s Up to 3000 MB/s
Max. speed of random read (blocks 4 KB) IOPS N/a
Max. speed random write (blocks 4 KB) IOPS N/a
Physical characteristics
Power consumption: inaction/read-write, W N/a
MTBF (mean time between failures), million h 2,0
Resource record, TB N/a
Dimensions: D × H × D, mm 22 × 80 × 2,38
Weight, g 8
Warranty period, years 5

To remedy the lack of information about the performance Silicon Power P34A80 possible, based on the characteristics of the controller. Phison promises that drives the platform base E12 versions the maximum amount can reach up to 600 thousand IOPS, and polterabayta modification can provide 360 thousand IOPS when reading and 500 thousand IOPS when writing. All this can be described as a decent enough level of performance: in the spec the same 970 Samsung EVO Plus show up very similar numbers.

However, when we are talking about NVMe SSD built on the basis of an independent Taiwanese developers always need to keep in mind that their characteristics are a lot of places for additions, forgeries and other fraud. Beautiful the promised number, as a rule, can only be obtained under special conditions by special tests, but in reality the performance is lower. Phison such tricks is famous for a long time, and solutions on the platform Phison E12 exception to their rules did not.

First, it applied the technology SLC-cache, accelerating the entry of small amounts of data. Therefore, high speed recording from Silicon Power P34A80 can be observed only in the case when the overall amount of recorded data is not more than 5% of the total capacity of the SSD. SLC-cache in this case is static, its size is about 24 GB on each 256 GB of storage capacity, and indicated in the specifications the rate of linear and random writes are achievable only within this region.

All this is easy to illustrate the graph, which shows the speed of continuous serial write on a blank Silicon Power P34A80 a capacity of 512 GB.

The speed entries in the SLC cache reaches 2.0 GB/s, but this performance is very long, on the main array of flash memory write speed is only about 580 MB/s. And in fact it is a relatively low performance for the drive on the basis of BiCS3 memory. The same WD Black SN750 with the same flash in the same configuration shows a remarkably high speed direct writes to the TLC memory. But in fairness it should be noted that some work to improve performance recording in IOS 12.3 is still carried: the size of the SLC cache at Silicon Power P34A80 now twice.

The second feature of the platform Phison E12 is that the SLC cache is activated not only when writing but also when reading. But the algorithm of its operation in this case is very primitive and at the same time cheat. About any pre-sample data it is not just the information that ends up in the SLC cache, it is transferred from the main array in flash memory immediately, but with a small time delay. This directly affects the performance in the tests, since the vast majority of benchmarks on the SSD first, write a test file and then immediately turns to him to measure the speed, thus inflated results. In actual use any disk read operations are performed from files recorded not only that, the speed in this case will be much lower.

Illustrate how this trick works, it is also very easy. The following graph shows the speed of random melkoplodnogo read data from a file on P34A80 Silicon Power 512 GB as immediately after creation, and after this the SSD was recorded some more information.

When the fresh test file is excluded from the SLC cache and then writing the additional 12 GB of data, the read speed of it is reduced by about a quarter. This means that the standard benchmarks that measure the performance by using calls to the file you just created will show for Silicon Power P34A80 significantly higher than it will give real work.

The model range of Silicon Power P34A80 formally includes four versions with a capacity from 256 GB to 2 TB, and they all really commercially available. Besides, they are relatively inexpensive — such a drive may attract those buyers who do not want to spend money on Samsung 970 970 EVO and the EVO Plus or WD Black SN750. While P34A80 looks like quite a budget option: Silicon Power gives him quite a full five-year warranty, not limiting the user in the amount of the rewritable information.

#Exterior and interior devices

For testing, we took a sample of Silicon Power P34A80 a capacity of 512 GB. Drive his performance did not surprise: it is a charge in the standard M. 2 form factor 2280 with a key of type M in stab connector that uses to communicate with the system bus PCI Express 3.0 x4.

Moreover, cost Silicon Power P34A80 has a full reference design and the blue color of the PCB is an indirect indication that she made at the company PTI under the direct control Phison. And this in turn means that to seek out some unique features drive Silicon Power is completely useless, it is identical analogous solutions Corsair, Patriot, Seagate, and other drives, built on the platform of Phison E12.

Many vendors NVMe drives on the basis of the chip Phison PS5012-E12 add to their products some or other cooling system. Silicon Power has decided to dispense with such frills, but the sticker with the marking on P34A80 has a thin layer of aluminum foil that is likely to contribute to the heat sink.

On the sticker label is not too much information: is given only to the article, SSD capacity and serial number. No information about date of production or filled at the factory firmware on it.

 

Board Silicon Power P34A80 512 GB has a one-way design, but this applies only to drives with a capacity of 256 and 512 GB is more capacious modifications are chips on both sides of the PCB.

The circuitry of the drive, in addition to the controller IC includes two chips of the flash memory chip and a DRAM buffer. Immediately it is worth noting that usually the SSD platform Phison E12 use DDR4 memory, but Silicon Power P34A80 512GB chosen the cheaper option – chip DDR3L-1600 SDRAM Nanya production, and half lower-than-usual volume – 256 MB.

But the flash memory in this drive is completely standard: it is a TCBBG55AIV chips, each of which is assembled of eight 256-Gigabit semiconductor crystals 64-layer TLC 3D NAND authorship Toshiba Packed by company PTI. This memory is a little worse chipsets with native Toshiba markings, as in this case, for the cutting and selection of crystals of the outside manufacturer, which has its own notions of acceptable for an SSD as flash memory. However, purebred memory Toshiba found only in more expensive products.

Capacity Silicon Power P34A80 512 GB after formatting the operating system 476 is “honest” GB.

#Software

Now, almost all SSD manufacturers offer a utility service that allows you to monitor the status of SSD and manage their work. At Silicon Power, this is the role utility of the SP ToolBox, however, from the point of view of functionality must be attributed to the number of unsuccessful examples of such programs: it is not able to practically nothing.

The only thing you can do with this program is to see General information about the SSD and access the S. M. A. R. T. telemetry. In other words, we can assume that no specialized tools and no.

SOURCE

Advertisement
Weak wifi signal causing you a headache? Get the signal strength you need to be productive and efficient on the internet with this Mini WiFi USB Signal Booster. Compatible with any device that has a USB port like your laptop or tablet, this easy-to-use signal booster helps your device pick up a wifi
Save $150 and receive free two-day shipping on Focal Sib Evo Dolby Atmos-Enabled 5.1.2 Home Cinema System!
Continue Reading

Peripherals

Overview NVMe drive Lexar NM610: return of the legend

Advertisement

So, the reason for the appearance of this article on our website was the fact that in the Russian retail now you can meet solid state drives Lexar. Lexar is a legendary name in the market of memory cards, is familiar to many enthusiasts with experience: this company originated in the mid-nineties, and her best time is in the middle of the two thousandth. In 2004, Lexar entered into a partnership agreement with Kodak, and in the next few years formed the tandem offered truly outstanding at the time of the performance and capacity of memory card for use in digital cameras. Orientation it is on Lexar digital photography was for the company absolutely clear priority, and in this field she was able to achieve great success. Even the chief marketer of the company at the time was a professional photographer Jeff cable (Jeff Cable). Not surprisingly, Lexar had and still has enormous weight among the users, one way or another related to photography.

However, in 2017, the company Acer, which was owned by Lexar in 2006, decided to completely abandon retail sales of removable drives based on flash memory. The result is Lexar brand as an independent ceased to exist, and the products under this name disappeared from the market. But not for long. Literally within a month became useless for Dell Lexar assets, primarily at a popular and reputable brand, found a buyer. It was the company Longsys – visible Chinese ODM-manufacturer of solid state drives, and other built and flash memory media.

So Lexar brand has found a new life. Last year Longsys decided to try to enter retail market — and is now actively pursuing it to life, using the legendary brand. As a result, first products Lexar returned to the North American region, and then came to Russia.

Longsys Electronics Shenzhen Company is a large contract manufacturer, the main products of which are solid – state drives for data centers. The firm operates from the beginning of 2000-x years and has now grown to the point that became one of the leading SSD manufacturers in China. The company has a development center, which employs more than two hundred highly qualified engineers. And that this team is able to create interesting and original products, we are already convinced. For example, it Longsys released the first fully Chinese SSD, which uses components exclusively of local origin: the 64-layer flash memory YMTC and controller Goke GK2302.

Office Longsys

However, products under the Lexar brand, which Longsys offers on the international market is not as exotic products. The company decided to begin to return the distinguished brand to life through traditional products Lexar memory cards and USB flash drives. But also added to them and a wide range of affordable SSD, with a senior representative which we decided to meet today. Arrived in our laboratory model Lexar NM610 is the best SSD brand PCI Express 3.0 x4, which can be purchased at a very attractive price. In this article we will try to assess if this product be of interest to buyers and why he was compared to the majority of similar alternatives.

#Specifications

Won’t create intrigue where there is none. NVMe drive Lexar NM610, is not rare in our part of the beast, and it is typical and widespread solution, and not high-level. Producer reasonably considered that to start a life journey with some exotics can be dangerous, and therefore decided to enter the market by the well-trodden path, used by many Chinese companies. So Lexar NM610 is unbuffered NVMe storage device on the basis of well-proven Silicon Motion controller SM2263XT similar to numerous NVMe SSD, which is rife with Aliexpress. However, you need to understand that the difference NM610 from numerous conventionally-unnamed Chinese alternatives is that in this case the manufacturer provides a constant hardware configuration, a certain level of quality of the used flash memory and the normal three-year warranty which is valid in Russia.

Otherwise, everything is quite ordinary. An array of flash memory to Lexar NM610 collected from the TLC 3D NAND chips produced by Intel. Used 64-layer TLC 3D NAND devices of the second generation, which have a capacity of 256 GB and, as the results of previous trials, provide relatively good performance. These devices are combined into a four-band array, operated by the controller SM2263XT. He does not know how to work with the DRAM interface, and for that reason the Lexar NM610 no own dynamic memory to store “fast” copy of the table of address translation. But he supports HMB (Host Memory Buffer, the buffer on the host side), which allows to buffer the table in the computer memory and refer to it on the bus PCI Express 3.0 x4 SSD under consideration through which communicates with the main system.

As we already know, in the case of storage platform, Silicon Motion SM2263XT this whole scheme quite well manifest himself in the case, but only in Windows 10 operating system – in other OC technology HMB is not supported.

If we turn to our recent testing of terabyte NVMe SSD, it would appear that Lexar NM610 in its hardware configuration similar to AMD Radeon R5 NVMe SSD or Transcend MTE110S. However, in passport characteristics Lexar somehow appear a little more high performance.

Manufacturer Lexar
Series NM610
Model number LNM610-250RB LNM610-500RB LNM610-1TRB
Form factor M. 2 2280
Interface PCI Express 3.0 x4 NVMe 1.3
Capacity, GB 250 500 1000
Configuration
Memory chips: type, interface, process technology, manufacturer Intel 64-layer 256-Gbit 3D TLC NAND
Controller SMI SM2263XT
Buffer: type, volume No
Performance
Max. sustained sequential read, MB/s 2000 2100 2100
Max. sustained sequential write speed, MB/s 1200 1600 1600
Max. speed of random read (blocks 4 KB) IOPS 110 000 188 000 188 000
Max. speed random write (blocks 4 KB) IOPS 151 000 156 000 156 000
Physical characteristics
Power consumption: inaction/read-write, W N/a
MTBF (mean time between failures), million h 1,5
Resource record, TB 125 250 500
Dimensions: D × H × D, mm 80 × 22 × 2,25
Weight, g 9
Warranty period, years 3

Overall the specifications look quite typical for products of this class. It’s not NMe’s flagship SSD, and the budget solution with modern and advanced interface. Therefore, linear speed is limited by chetyrehkantnomu array of flash memory, and the lack of DRAM buffer has the latency and performance when random access. However, if you compare Lexar NM610 with suggestions from the same price category, its performance is quite outstanding. It stands out especially on indicators for the design of the resource: the manufacturer allows you to overwrite almost half of the total capacity of the SSD (more specifically, 46 %) daily for a three-year warranty period. Typically, such a vitality promise to offers more high-end, but Lexar did not stint on quality memory Intel and now can afford not to restrict the user in how much information he is allowed to write to the drive.

However, you need to understand that this drive is still not designed for high loads. But it’s not in the endurance of flash memory, and that with heavy use, the performance of the SSD is markedly reduced. To maintain good responsiveness of the controller Lexar NM610 requires a sustained pause, during which he will be able to release SLC cache and to reorganize a table of address translation, the size of which in this case is smaller than in models with a private DRAM.

All this is easy to illustrate the results of practical experiments. To see the algorithms SLC-caching in action, we looked at how changing the speed of continuously writing data to the Lexar NM610 500 GB (the measurements were carried out on a clean SSD).

The write speed of Lexar NM610 in SLC-mode up to about 1.6 GB/s, and to free drive with such performance, you can record about 65 GB of data. In a slow TLC-mode consider SSD only goes after that. And while writing new data in the TLC mode, the controller need to free up space – compress the data that was recorded in a single bit SLC mode. As a result, the write speed is not only reduced to the level of about 135 MB/s but loses any stability. Fortunately, under normal home use such situations will have to face often – they can occur only when continuous recording very large amounts of data.

It is easy to trace and how affects the performance of the fact that in-memory with fast access Lexar NM610 is only part of the table address translation. It is enough to observe the rate at random melloblocco reading, which requires multiple search matches in the table of address translation. For example, we follow the reduced speed random reads with increasing of the volume data which are read requests.

To interpret these results is very simple. As long as the total size of the files accessed by the storage controller, remains so that any required portion of the table of address translation is placed in the HMB-buffer in RAM, we see the constant speed random read (queue requests) at 63 MB/s. But as soon as the active part table in HMB buffer fit ceases, the performance immediately drops. In the worst case, the speed of random access can be reduced by about half, to a relatively low level of 33 MB/s.

Based on these data, we can assume that for his HMB-buffer Lexar NM610 reserves a memory region size of 64 MB. Therefore, sets of files smaller than 64 GB it works with good speed, and with the further increase of the volume leads to a drop in performance.

In addition, the firmware Lexar NM610 we found algorithms to increase its performance in benchmarks. Their essence lies in the fact that the information got to the SLC cache is delayed it for some time to provide the best performance when accessing files that were recorded just. You can see this with a simple experiment, during which we check the speed of random melkoplodnogo read data from a small file right after burning it on a flash drive, and after this the SSD was added some more information.

Speed melkoplodnogo reading at 63 MB/s Lexar NM610 shows only when working with the newly written, “fresh” test file. If after write this file in the SSD was recorded 28 GB of data, he’s from SLC cache is displaced and the performance of the access it falls in half. Another drop in performance occurs after a write to the SSD 64 GB of data — it is due to the displacement of the information about the source file of HMB-buffer, which stores a copy of the table of address translation.

Thus, the architecture of the drive is such that it cannot provide the same quick access to all the information that it stores. High performance is achieved only in the case when working with relatively small amounts of data. In other words, should not be regarded as NM610 Lexar high-speed solution for high loads — the drive for them did not fit.

#Appearance and internal structure

For testing we received from Lexar sample NM610 capacity of 500 GB. This volume should be significantly faster than the 250 GB version and is close in speed to the terabyte drive. Cost Lexar NM610 500 GB in local shops is less than 5 thousand rubles.

About external performance Lexar NM610 to tell especially there is nothing. This is a typical accumulator-based low-cost controller SMI SM2262XT. Very similar in design to the SSD offer and many other companies – in this regard, Lexar did not invent anything original. Fee NM610 is a M. 2 module form factor 2280, designed for installation in the slot supplied with four lines of PCI Express 3.0. Regular cooling the drive has.

 

On the front side is pasted the label with the model name and serial number. No additional useful information, such as production date or version of firmware on it. The reverse side of the PCB NM610 free from components that allows you to use this SSD in particular in thin laptops.

An array of flash memory is represented by four chips, inside each of which contains four crystal 64-layer TLC 3D NAND produced by Intel. Such crystals have a capacity of 256 GB, therefore, four-channel array formed with a fourfold alternating devices NAND channels, which causes a maximum capacity of 500 GB version among the various capacity options Lexar NM610.

 

You should pay attention to the fact that the chips flash memory does not have the Intel logo and labeled with some obscure sequence of characters. The fact that the testing and packing of crystals in the chip Lexar (or rather Longsys) working independently, receiving from Intel flash memory in the form of uncut wafers. This reduces the cost of the final product.

In addition, against the background of similar products lineup Lexar NM610 stands out somewhat atypical set of dimensions that are multiples of 250 GB. Thus, to the backup area in the storage allocated about 9 % of the total flash-memory. There are no fundamental advantages it does not, but having a slightly larger pool of “spare” cells in certain circumstances can extend the life time of the SSD.

Nothing more interesting in the design of the Lexar NM610 no. Perhaps someone surprised by the small size of the chip SM2263XT or the lack of the usual circuits of dynamic memory, but for those who have seen the drives on low-cost Silicon platform Motion, all of this is quite ordinary things. In other words, the flagship offer Lexar was indeed reduced NVMe product, which is built on the stripped-down characteristics of the platform. But it is quite appropriate money.

#Software

Now, almost all SSD manufacturers offer a utility service that allows you to monitor the status of SSD and manage their work. From Lexar, this role is assigned to the utility SSD Dash, however, from the point of view of functionality, this development should be attributed to the number of not the most successful examples of such programs: its possibilities are very limited.

The only thing you can do with this utility is to see General information about the SSD to access its S. M. A. R. T.-telemetry, and run Secure Erase. One can only hope that the program will evolve in the future will acquire additional features.

SOURCE

Advertisement
Samsung J7 V just $5 mo. New device payment purchase req'd. Plus, free shipping.
Save $120 on iPhone SE 16 GB with Sprint Flex Lease when you visit Sprint.com. $7.09/mo. for 18 mo. after $120 instant savings for well-qualified customer with new-line activation.
Continue Reading

Peripherals

Testing hard disks with capacity up to 14-16 TB: not only more, but better

Advertisement
Electronic Gift Ideas, Free Shipping!

The capacity of hard drives continues to increase, but the pace of growth in recent years has been steadily declining. So, in order to release the first drive with a capacity of 4 TB after the sale there was a 2-terabyte HDD, the industry has spent only two years to conquer the level of 8 TB took three, and to double the volume of 3.5-inch hard disk again was only for five years.

The last breakthrough was accomplished through a list of innovative solutions. Today, even conservatives such as Toshiba, which until recently refused from helium, forced to release the hard drives in sealed enclosures, and the number of plates on the mandrel was increased to nine pieces — though once, a long time, five plates was considered a reasonable limit. In specific niches, the technology of so-called shingled recording (SMR, Shingled Magnetic Recording), in which track sectors on the plate partly overlap. Finally, in order to shift the limit of hard drive capacity from 14 to 16 TB without the use of SMR, manufacturers had to introduce one of emerging technologies, gradually shrinking the list of which we reproduce in the annual final articles, the reading track multiple heads at the same time (TDMR, Two-Dimensional Magnetic Recording). Further movement sooner or later will require more extensive changes to the basics of HDD such as a heated plate with a laser or microwaves (HAMR/MAMR, Heat/Microwave-Assisted Magnetic Recording) at the time of passage of the recording head.

However, you will notice that all of the techniques are aimed primarily at the increase in density and increase in volume on one spindle, although many of them have a beneficial side effect in the form of increased speed of linear read and write data. This parameter modern HDD broke for the border 250 MB/s and is comparable to the early consumer SSDs. But the speed of access to random sectors of the magnetic disks almost never progresses, and in terms of volume, the number of operations per second becomes less. Along with this are higher requirements for fault tolerance, because the more data is stored on one spindle, the more important not to lose them and the longer to recover.

But this challenge from the creators of magnetic storage was the answer. We took three hard drive with volume from 14 up to 16 TB, to see how to adapt technology 64-year-old to the realities of 2019, and noticed several trends. Champion’s examples of modern 3.5-inch drives, produced for rackmount server and storage solutions, have something in common with solid-state drives — from the principles of addressing to direct integration of the flash chips in the local memory stack. And consumer models, in turn, become closer in its characteristics to the server-side counterparts, and even the description “desktop HDD” is not so much to say about the performance and reliability of the device. But the task of this review is not limited to the General words. We intend to figure out how the new trends in design of hard drives are expressed in hard numbers measuring performance.

#Specifications test participants

Before we start analyzing the test results, you should carefully study the characteristics of the devices with which we have to deal with. This time they do not so much as usually happens in our group test, but we have fulfilled the main conditions, without which the comparison hard drives can claim to completeness. In the survey participated the products of all three manufacturers — Seagate, Toshiba and Western Digital, and they belong to different categories: consumer and server. The main characteristics that unite them, is the volume at 14 or 16 TB, the sealed housing filled with helium, and a spindle speed of 7200 rpm And for comparison with the heavyweights in testing involved three of the now familiar device of a smaller volume (10 and 12 TB), was designed for use in servers, home or office NAS.

Manufacturer Seagate Toshiba Western Digital
Series BarraCuda Pro Exos X10 IronWolf MG08 S300 Ultrastar DC HC530
Model number ST14000DM001 ST10000NM0016 ST12000VN0008 MG08ACA16TE HDWT31AUZSVA WUH721414ALE6L4
Form factor 3.5 inch 3.5 inch 3.5 inch 3.5 inch 3.5 inch 3.5 inch
Interface SATA 6 GB/s SATA 6 GB/s SATA 6 GB/s SATA 6 GB/s SATA 6 GB/s SATA 6 GB/s
Capacity, GB 14 000 10 000 12 000 16 000 10 000 14 000
Configuration
The spindle rotation speed, rpm./min 7 200 7 200 7 200 7 200 7 200 7 200
The useful storage density, GB/plate 1 750 1 429 1 500 1 778 1 429 1 750
The number of plates/heads 8/16 7/14 8/16 9/18 7/14 8/16
Sector size, bytes 4096 (emulated 512 byte) 4096 (emulated 512 byte) 4096 (emulated 512 byte) 4096 (emulated 512 byte) 4096 (emulated 512 byte) 4096 (emulated 512 byte)
Buffer size, MB 256 256 256 512 256 512
Performance
Max. sustained sequential read, MB/s 250 249 210 ND 248 267
Max. sustained sequential write speed, MB/s 250 249 210 ND 248 267
Average seek time: read/write, MS ND ND ND ND ND 7,5/ND
Fault tolerance
The estimated burden of TB/g 300 ND 180 550 180 550
Unrecoverable read error, the number of cases on the amount of data (bits) 1/10^15 1/10^15 1/10^15 10/10^16 10/10^14 1/10^15
MTBF (mean time between failures), h ND 2 500 000 1 000 000 2 500 000 1 000 000 2 500 000
AFR (probability of failure per year), % ND 0,35 ND ND ND 0,35
The number of cycles Parking the heads 300 000 600 000 600 000 600 000 600 000 600 000
Physical characteristics
Power consumption: inaction/read-write, W 4,9/6,9 4,5/8,4 5,0/7,8 ND 7,15/of 9.48 5,5/6,0
Noise level: idle/search, B ND ND 1,8/2,8 2,0/ND 3,4/ND 2,0/3,6
Maximum temperature, °C: drive on/drive is disabled 60/70 60/ND 70/70 55/70 70/70 60/70
Shock: drive on /drive is disabled ND 40 g (2ms) / 250 g (2 MS) 70 g (2ms) / 250 g (2 MS) 70 g (2ms) / 250 g (2 MS) 70 g (2ms) / 250 g (2 MS) 70 g (2ms) / 300 g (2 MS)
Dimensions: D × H × D, mm 147 × 101,9 × 26,1 147 × 101,9 × 26,1 147 × 101,9 × 26,1 147 × 101,9 × 26,1 147 × 101,9 × 26,1 147 x 101,6 x 26,1
Weight, g 690 650 690 720 770 690
Warranty period, years 5 5 3 5 3 5
Retail price (US without tax) $ From 549 (newegg.com) From 289 (newegg.com) From 351 (newegg.com) ND From 301 (newegg.com) From 439 (amazon.com)
Retail price (Russia), RUB. From 34 348 (market.yandex.ru) 17 498 (market.yandex.ru) 26 320 (market.yandex.ru) ND 19 784 (market.yandex.ru) 27 495 (market.yandex.ru)

The first model in our modest collection of hard drives indiscreet volume — BarraCuda Pro 14 TB — is a storage for desktop and DAS, but not simple, and “professional”. On the one hand, this means that the BarraCuda Pro is subject to the typical limitations of desktop hard drives. For example, it is not intended to be combined in RAID arrays, because it is desirable to have TLER (Time-Limited Error Recovery) — configure the firmware that prevents W: a from HDD array because of the long attempts of the microcontroller is considered a problematic sector. In addition, the chassis BarraCuda Pro is ill-suited to work in the shelf or NAS with multiple baskets, because there is no compensation for rotational vibration.

But on the other hand, unlike most other desktop hard drives, HDD of this brand have increased year resource loading — up to 300 TB of overwrites, — ready to work 24/7 and are accompanied by a five year warranty. Performance probably also don’t have to complain (at least, in problems with a predominantly linear data access): with eight plates by 1.75 TB device achieves a sustained throughput of 250 MB/s. in addition, the manufacturer promises that the speed and random access to BarraCuda Pro should be higher in comparison with the ordinary disk drives for desktop computers and energy consumption, in contrast, is lower than that of most 3.5-inch models. However, all statements of Seagate, we’ll see.

In order to conquer such a high turn-density data in the standard perpendicular recording without the use of niche technology SMR (Shingled Magnetic Recording), Seagate had to implement one of the promising methods, which we from year to year in writing of our final articles — the so-called two-dimensional recording (Two-Dimensional Magnetic Recording). But contrary to its name, TDMR has nothing to do with the procedure of data recording as such, and is designed to increase the ratio signal/noise in conditions of high track density on a magnetic plate due to the simultaneous reading of track two reading heads: the latter spaced so that the field captures the adjacent track, and to compensate for the interference becomes easier. In the future, hard drives with TDMR will gain even more heads, along with the reliability of read data may increase its speed, but it is the future.

Drives BarraCuda Pro differ in many ways from the related devices of the Junior series without the prefix Pro — from the fact that all manufacturers of desktop HDD standard models stuck in the amount of 6-8 TB. Disk BarraCuda Pro can rather be described as the offspring of server branches Seagate, devoid of functions, associated with arrays. But the price of the device has soared to the level of corporate models, and even higher in Russia 14-terabyte model will not find a cheaper 34 348 rubles, and retail venues in the United States — $549. Even nearline model Seagate of the same size cost less — from $375 or 28 936 RUB.

 

Seagate BarraCuda Pro 14 TB

The following experiment, a 14-terabyte disk Ultrastar DC HC530 — this drive is nearline class, which represented the best of what the engineers of Western Digital until there is a new model with capacity of 16 TB. But in practice 3DNews he became the first Winchester brand Ultrastar without the usual HGST letters in name: the company moved all server models under its own brand after the assets of HGST are completely dissolved in the combined Corporation. In their key characteristics is a device like a BarraCuda Pro of the same size: inside of a hermetic enclosure Ultrastar DC HC530 are eight useful magnetic platters with a capacity of 1750 GB, and read data from closely spaced paths provides technology TDMR. But for the rest of the parameters and the plurality of additional functions inherent in enterprise HDD, Ultrastar DC HC530 not be put on a par with the desktop models, let BarraCuda Pro and not representative of its category.

So, a useful bit density on the plates Pro and BarraCuda Ultrastar DC HC530 one and the same as the spindle speed, but WD product guarantees a higher sustainable speed of linear read and write data up to 267 MB/s (it is unclear where did the difference, but the tests will show whether it exists in reality). Delays in random access helps to reduce the two-stage actuator new, third generation and a large buffer of 512 MB, and most importantly, Media Cache — reserve areas to blocks scattered over the surface of the plates. The latter feature unites modern nearline disks with solid-state drives, which, too, there is variable correlation between physical sectors and logical blocks. And since 10-terabyte models Ultrastar DC HC330, WD uses for caching write operations have a small amount of flash memory. Note that at the same time with (potentially) extremely high by the standards of magnetic storage performanceproduct WD is characterized by moderate power — in fact, it is the device with the lowest power consumption among all test takers, judging by his passport options.

Drives of this class is constructed with the expectation of continuous operation in conditions of server rack: double-sided mounting of the spindle, the compensation of rotational vibration and other structural features of Ultrastar DC HC530 allowed to bring the design load of the disc up to 550 TB/year, and the time between failures is typical for nearline models of 2.5 million hours. In the unlikely event of a failure while updating the firmware on the controller Board is soldered a spare chip. The drive comes in versions with native access to the marking 4 KB or emulation of 512-byte sectors with the SATA or SAS. In the latter case, an option is also available through data encryption.

The retail price for the WD Ultrastar DC HC530 configuration with SATA port and emulation hereditary 512-byte marking comply with the advanced features and technology of this device: 27 495 rubles in the Russian Internet-stores and $439 on Amazon.

 

WD Ultrastar DC HC530

To assemble a collection of winchesters with a volume of 14 TB for comparative testing was difficult, and the appropriate device of the third manufacturer — Toshiba — we were never able to get. But instead we got a model of the next generation, to 16 TB. Now all three companies manufacturing hard drives, offer drives of this size, but it is the product Toshiba series MG08 was among them very first. The record of the Japanese company relies on a plate with roughly, if not exactly the same physical recording density in hard disks BarraCuda Pro and Ultrastar on 14 TB, but for the first time, Toshiba was able to pack nine “pancakes” in a standard 3.5-inch case. Not without TDMR technology, which has become essential for conquering new frontiers of capacity. Bandwidth Toshiba MG08 in the operations of the linear read/write should be at the level of the WD Ultrastar DC HC530, but, oddly enough, the manufacturer did not disclose any details about the device performance.

But we know that Toshiba also took measures in order to increase reliability and to reduce the latency of write operations: chip flash memory on Board MG08 in the case of accidental power outage to protect data sent from the host controller to record, but judging by the test results, and even performs the function of the second level cache memory after the buffer DRAM. But this technology (Persistent Write Cache) appears only in the specs of the drives emulate 512-byte markup, which is an additional source of danger during a power failure (and to some extent stealing performance) because of the need to perform the operation read-modify-write when each record logical blocks that do not coincide with the boundaries of physical sectors. But the MG08 series also includes models with native access to 4-kilobyte sectors. Does this mean that the latter does not have flash, or it just removed the backup feature, we do not know. But regardless, PWC, Toshiba MG08, and other storage devices from this company, uses algorithms Dynamic Cache, which, according to the manufacturer, the optimal sharing of buffer space between reads and writes. Any information about them we also do not have.

Other sources of increased fault tolerance in the design of the Toshiba MG08 are mounting the spindle on both sides and the rotational vibration sensors. These drives are designed to record up to 550 TB of data per year, have a standard corporate device time to failure of 2.5 million hours and a five-year warranty period. Available to order several different configurations of the drive with SATA or SAS drives and optional end-to-end encryption. However, for the price we I can not Orient: 16-terabyte disk Toshiba was introduced in January, but it is still a rare beast in retail sales.

Toshiba MG08 16 TB

Now, when we met with the three main parties to the test, look at the hard drives of smaller capacity, with whom we have to compare the new 14-to 16-terabyte model. One of them, the Exos X10 volume of 10 TB, is a nearline storage device that contains seven magnetic platters in a sealed enclosure. Although since the useful capacity of the plate is increased from 1429 to 1750 GB or more, the speed of sequential access to the hard drives must also increase, in this parameter Exos X10 is practically not inferior to the same BarraCuda Pro 14 TB according to the specifications of both drives. Something clearly does not fit the specs of the Seagate hard drives, but we have the opportunity to learn in practice.

In order to increase the speed of operations with any available in a series of Exos have developed a caching mechanism recording AWC (Advanced Write Caching), reducing the response time. In the framework of the AWC write operations are grouped into the buffer DRAM, as happens in any other hard disk, but the buffer keeps a copy of data after they dropped on the plate, and mirror the contents of the buffer can be immediately read by the host controller. In server hard disk Seagate 2.5-inch form factor AWC includes the following in tier — reserved areas on the plate surface, where the data from the DRAM are written in sequential order (Media Cache), and a small amount of nonvolatile memory for saving data from the buffer when power is lost. But Exos X10 flash memory is missing, and maybe Media Cache along with it.

From consumer hard drives for desktop computers and NAS drives Exos series are high performance time between failure (2.5 million hours) and the load (550 TB/year), the ability to operate in a server rack with no limitations on the number of baskets, as well as a five-year warranty period. Hard drive with model number ST10000NM0016 inherited to the test, in addition, refers to Hyperscale-modifications that have reduced power consumption compared to other members of the family of Exos, but are released only with SATA interface and emulate 512-byte sectors. In configurations with SAS connectors among the models Exos there are options and with native access to the 4-Kbyte sectors, as well as function through full-disk encryption.

 

Seagate Exos X10 10 TB

Hard drive Seagate IronWolf not so long ago appeared in our review of the new representatives of this brand along with solid state storage Seagate network attached storage. The 12-terabyte model IronWolf, apparently, comes with plates with the same density of the physical layout as Exos X10, only here they are one more. However, Seagate estimates the performance of their offspring in sequential read and write is much lower — only 210 Mbytes/s And no sophisticated technology designed to compensate for high latency of the response inherent in magnetic storage devices, not here either.

But all hard drives IronWolf, since the volume of 4 TB borrowed from the Exos series a number of hardware features that contribute to increased resiliency. Block of the magnetic plates of each Winchester balanced in two planes, and the rotational vibration sensors provide stable performance in rack storage or separate NAS with a number of disc bays, up to eight. IronWolf is designed for moderate mode of operation with design capacity of 180 TB/year and is characterized by the time between failures of 1 million hours. As a consequence, and the warranty period from IronWolf is not as long as at more serious models in the catalog Seagate three years.

 

IronWolf Seagate 12 TB

Badged S300 Japanese company Toshiba has released a series drives for video surveillance systems — these winchesters too devoted to their own review pages 3DNews. Due to the expansion of the data transfer Protocol for ATA Streaming Command Set older model Toshiba S300 guarantee simultaneous video recording from up to 64 cameras, but at its core are typical drives for NAS and DAS with the ability to operate 24/7 and a decent resource MTBF: like IronWolf, it is 1 million hours and a warranty period — the same three years. Thanks to the constructive advantages of the S300 chassis — mounting of the spindle on both sides and active compensation of rotational vibration allowed the installation of more than eight such devices in one rack shelf or stand-alone NAS.

Model S300 chosen for comparison with novelties volume 14-16 TB, built on the basis of hardware chassis server drives MD06ACA-V and contains seven magnetic plates and specs on the device is typical of modern large HDD speed of random read/write 248 MB/s. But from the techniques that are used in the Toshiba server hard drives to reduce latency, the S300 was only a function of Dynamic Cache.

In contrast to all other test participants, S300 even with a thick stack of seven plates without helium and carried out in a standard ventilated enclosure. It seems that for this reason, 10-terabyte model belongs to the highest value of power consumption in consolidated specification table of test participants, and this option, though in itself important only for administrators of data centers that directly detects the temperature of the HDD. Real consumption S300 we’d like to see yourself, but until we take the item note.

 

Toshiba S300 10 TB

SOURCE

Advertisement
Samsung J7 V just $5 mo. New device payment purchase req'd. Plus, free shipping.
Code: VZWDEAL. Enter this coupon code at checkout to get $100 discount on Samsung Galaxy Note 8. Includes free shipping. Restrictions may apply. Device payment purchase required.
Continue Reading

Deals

Advertisement
5.5��� 3G+16G MSM8939 8 Core Phone
Code: measy. Measy T8C Intel X-Z8300 Tv Stick
Code: HOTSPOT. A Night Without Wifi Free Tee with Order $50+! Use code: HOTSPOT

Trending