In fact, products Lite-On perfectly familiar to domestic consumers. Moreover, in some years the model of the manufacturer took the first place in various rankings. It is, of course, a solid-state drives Plextor.
For a long time brand Plextor is a subsidiary of the Japanese company Shinano Kenshi, leading its history since 1918 (she Plextor was organized in 1985), but a few years ago, Plextor brand became the property of Lite-On, which until then was for Plextor contractor for the manufacture of it drives. In other words, Lite-On company is present on the Russian market in the past, just not under its brand.
But the market situation is changing, and, apparently, the Plextor brand has started to lose its popularity. A simple observation: in the Forum’s recommendations in its favor are becoming rarer. Apparently, because of that drives Lite-On has started to deliver on our market. Some party fell into retail and occasionally flashed on store shelves in the past, but this line of Lite-On MU-3 came in droves.
At first glance, the specifications of the full order: volume up to 960 GB, linear read speed is 555 MB / s, and the linear recording – 470 MB / s for random reads and writes all also very good. A potential leader? Some understanding of what it is actually not so great, comes with a more thorough understanding of data, resulting in Lite-On.
First, the already alarming that does not indicate any information about the type of flash memory and controller used, although in general it is not a definite indicator. But beyond that it begins the more interesting.
Secondly, the line «DRAM Cache» says only about 32 MB, and for the entire range of volumes, but still with a note «SDR». In fact, in such a way designated by the so-called “bezbufernye» NAND controllers that to reduce the cost by simplifying PCB layout and cancel the installation, a separate memory chips are equipped with a built-in buffer a small volume.These controllers are not only lower costs, but not the best indicators of performance. Phison PS3109-S9, Phison PS3111-S11, Samsung MFX, Marvell 88NV1120 – if under simple loads they serve is still relatively good, in difficult situations are often at the bottom of the indicator tables.
Third, it makes you think the value of the declared write speed: 470 MB / s, even in the youngest modification of 120 GB. Sorry, but there are no miracles: all current flash memory, used in modern models of SSD-drives, such indicators can not boast. 330-350 MB / s – this is a ceiling, and is 15-16 nm MLC NAND, which budget decisions is not widely applied, replaced by TLC NAND. The controller SandForce SF-2281 with its compression believe there is little, it is too old, and not referred to as «32MB SDR». Thus, uniquely applied here “fast write” mode (SLC-mode).
Fortunately, in our hands was a copy of Lite-On MU-3 of 240 GB, and therefore we can on real sample to see what lies under the hood, and assess how well in practice, the creation of the company’s engineers. Of great interest is also a version for 960 GB, but at this point in the domestic retail neither she nor the version of 480 GB are not represented (and get them to test it is impossible), so we were limited to one sample.
Page on the manufacturer’s website: Lite-On MU-3 240 GB (PH4-CE240) ( the datasheet ).
Packing Lite-On MU-3 repeats the judges at Plextor drives to the smallest detail: there is even a label on the side, which contains the date of manufacture, batch, serial version of firmware microcode and serial number.
A delivery in this case is also absent.
The user is prompted only the device itself in the form of transparent plastic and an information booklet.
But further showed surprise.
Interesting building, is not it? Regular readers must have understood what I mean: body Lite-On MU-3 is identical housings Phison drives. And not just very similar, and repeats in every detail, down to the pulling out of individual small items.
Traditionally constructive budget: no screw, the clutch halves is provided exclusively latch system. Note the complete absence of any seals.
The short circuit board – another way to reduce some production costs. Of course, I already knew that inside Phison S11, but the fact that it is a complete “Reference» Phison until the memory chips (and typical warranty label on the memory chip) is very surprised.
Torments nagging question: what is this from Lite-On? It is possible that the firmware, but not the fact that everything in it is not opposed to the designation of the version that can be and change. Phison, in itself, though a major Taiwanese developer drives his own does not produce, the charging session is one of the ODM-manufacturers, which, according to unofficial data, is Powertech Technology Inc. (PTI) and which in turn has a similar Toshiba Services (PTI OCZ acquired plants after bankruptcy last and now produces a series of SSD-drives Toshiba OCZ) and Intel.
A funny situation: one of the largest ODM- and OEM-manufacturers, the company Lite-On, has its own strong engineering and manufacturing capabilities, acquires finished goods from other similar products in order to sell it under its own brand.
Flash memory chip with a strange marking TT69G51ARA, incidentally, also PTI assembly.Self-cutting and packing of NAND memory chips allows for the selection of the resulting chips in their requirements to the final characteristics of NAND crystals (the percentage of broken cells, levels of operating voltage, etc.). This makes it possible to change the percentage between the shares of the finished products and scrap, reaching at the cost of finished products. Plus, among other things, the absence of the original brand on chips also give some savings.
So what lies behind the name TT69G51ARA? As a rule, based on the accumulated experience of SSD test, a chip marked, beginning with «T», is Toshiba memory. Already logically clear that this is a budget solution, and hence the conclusion follows: TLC NAND, besides, most likely made on the most important technical process at the moment. These are for Toshiba is 15 nm.
ID 98; 3c; 99; a3; 7a; 51 in fact corresponds to 15 nm TLC ToggleNAND Toshiba production.Thus, based Lite-On MU-3 of 240 GB is controlled by the controller Phison S11 array of flash memory, sixteen typed crystals capacity of 128 Gbps each, and the actual volume of the drive is 256 GB.
But part of the standard array is allocated in the hidden reserves, and the amount indicated in the decimal system (used to indicate the volume of 1 GB is equal to one billion rather than 1,073,741,824 bytes). Therefore, in reality, only 223.57 GB available to the user, and the remaining amount of the firmware of the controller operates for official purposes: for wear leveling, as reserve pool for the replacement of defective memory cells and others.
Unfortunately, there is no software support from the Lite-On MU-3 240 GB MU-3 Series drives do not. So we can use a universal application to view the state of the parameters SMART.
That of the Lite-ON MU-3 eleven. Among the most obvious possible to note 01 (read errors), 02 (hours of work), 0C (the number of closing and opening cycles), the C0 (power failure), the E7 (the remainder inferred resources as a percentage) and F1 (the amount recorded on the SATA interface data keeping in gigabytes).
Temperature monitoring is not working, the display value of + 33 ° C is trivial software “plug” that is often found in low-end models of SSDs.
With CrystalDiskMark (64bit) 3.0.1 in random mode, the data is held four times performance measurement:
The initial state of the new drive ( “zero” state);
After the entire test cycle is filled with data from a disk user article previous partition (Word files, photos, audio and video), so that the total amount of data has been recorded at least three times the total volume of storage;
Thirty minute easy, for which is not made any transactions with the SSD – for background functions Drive Cleaning “garbage”;
Performing TRIM command forces the operating system.
Thus, we can see how well the drive firmware is up to the task of maintaining performance levels on low volume simultaneously written and read data – for the operating conditions in the domestic environment, this is enough.
Next, the complete cleaning of the drive, and then run AIDA64 Disk Benchmark test in «Write» mode (block size is set to 1 MB) – This test is linear recording the total volume of the carrier, simultaneously displays information about the recording process in an easy schedule.This test allows us to see how the whole drive is stable, there is no overheating and whether any may have algorithms “fast recording” implemented in the firmware.
Finally, after the TRIM command is the whole volume of storage, testing is done using Iometer.
Simulated work load drive in conditions close to the server (continuous random 4KB blocks record the entire volume from the request queue depth of 32) in the absence of TRIM. That is how, for example, the database work: create one or unspecified number of large files, which are carried out within the read / write operation, the TRIM command generation is not happening. The test is carried out continuously for two hours with every second performance indicators are removed. The results of this test allow us to see the possibility of an experimental in its “pure” and the “use” conditions (achievement of “performance established” state).
Upon completion of this test, which has done one more goal will be finding out how well the algorithms “garbage collection» (Garbage Collection). At the final graph present speed performance drive in four situations: a “clean” cell array, after the continuous load for two hours in the absence of TRIM command, after being idle for 30 minutes, which should be sufficient drive for development of internal algorithms “garbage collection”, after TRIM command execution on the entire volume of the drive. The test is quite specific, and its results are important for those who aim to operate in the conditions of work without TRIM.
The fall recording velocities in filled condition is caused by the fact that in our test, we fill the data storage device so that the free space is only about 20 GB, and hardware platform performance is not enough for an array of memory clearing required for SLC-mode.Ultimately, the performance falls on write operations.
However, it should be noted that the already “not to waste time” during the test firmware and engaged in parallel “garbage collection”: by the last test write speeds are restored. After downtime performance and all are almost identical to the original.
Linear overwrite the entire volume shows the classical picture of drives, built on the basis of the hardware configuration with TLC NAND flash memory and the implementation of the algorithm pseudo-SLC: some of the data is recorded at high speed, then the controller switches to the normal recording mode, and the speed drops. Size SL-cache is a little more than one percent of the entire array of flash memory storage.
Simply copying large files confirms the results in AIDA64: outside SLC-cache write speed is about 75 MB / s.
Unfortunately, as it was before with Phison controllers, we observe the relative volatility of performance: in the process of recording every few seconds, the speed drops to almost zero – the graph is clearly visible marks corresponding line. The average speed of the drive is at a level of 15 000 IOPS.
Interestingly, the performance drop starts at ~ 203 GB of data recorded. But not dramatically, but only to about 12,000 IOPS. Obviously, at this point it starts the firmware to consolidate data recorded in the SLC-mode. The final drop accounted for 208 GB of data recorded – a little earlier, “the situation.” Final stabilization – about 4000 IOPS.
In general, the performance slightly better than one would expect for this kind of heavy load (just in case I emphasize in a home PC this kind of load does not occur, the test is more of academic interest).
This schedule is different scale, so we can clearly see the systematic performance drop to zero. In the normal state, when the drive comes to the TRIM command, the amount of SLC-cache of 2.84 GB. But the firmware of the hardware configuration can not boast of a good implementation of “garbage collection” algorithms, resulting in storage in the absence of TRIM will take only about 800 MB of data.
It is not at all “desktop” motherboard supports DIPM team that takes the drive to “deep sleep” mode, whereby power consumption drops to extremely low levels. In relative terms, the difference is impressive: up to five to seven times, but in actual relation to it is a value of about one watt or less. Last for a desktop PC does not play any role.
But at the same time, SSDs are often put in laptops, and the question of support for this command in specific models of interested users in a completely practical light: DevSleep mode, which switches SSD with the active support of DIPM, you can add to an autonomous work an extra five or ten minutes, it is sometimes critical.
In the process of testing used two motherboards: ASRock Z170 Extreme6, does not support DIPM, and Zotac Z77-ITX WiFi (Z77ITX-AE), where the necessary support is implemented. It was somewhat easier than looking for a motherboard with the necessary features “in one.”
And in order to avoid the back board from the processor socket (as it is known, processor socket type LGA is quite fragile and designed for a limited number of CPU re-installation), it was decided to collect two nearly full test configuration: motherboards directly assembled with CPU, RAM, and other simply rearranged to stand as needed. The general was only PSU – Corsair HX750W power of 750 watts.
Creating a file system standard tools of the Windows single partition for the entire volume of the support, the NTFS file system, the size of the cluster – the “standard”, “file compression” turned off.
As the test software are used:
Crave browser fate of occupied serial testing SSD models. But at least it is hard at the one who is interested in SSDs on a serious basis, and not on the principle of “Aha, the brand!Wrap two! “. The problem is that manufacturers are using the low level of knowledge of some users, as well as the fact that the opaque storage enclosure and sealed can under the cover of the product to put anything. Yes, first there is the best, then, when a wave of reviews will take place and some typed a lot of positive feedback, the course starts to go something cheaper. And sometimes one and the same model originally comes in different variations.Someone from users is no difference, but someone – interested in the question, for which he paid the money?
Someone begins to test your freshly bought a device, and then compare the results with those that he sees in the reviews. And there may be legitimate question: “? Why my SSD shows the lower / higher productivity levels than in the review” Yes, the difference is the reason may lie in properly configured PC (for example, such as antivirus applications are running in the background), are not satisfied with the microcode the motherboard BIOS (the example above – test board Zotac) and initially low level of system performance. For example, SATA 6 Gbps controller / with AMD’s chipsets, even in the new A88X and A78 slightly, but weaker than in the not very “fresh» Intel Z77.
And then there’s the game producers with stuffing SSDs. Especially the issue of the difference concerns SandForce device platforms: its peculiarity is that it does not have one or two or three (and so on, that is a limited number) of the controller configuration and flash memory.The total number of configurations in this platform to date is that their numbers have overcome the value of 33 000 (not a typo, it was thirty-three thousand). Some companies and does not shy away from the full replacement of “stuffing” of the other. As a result, a drive name for a full comparison is not enough, you need to know the specific platform on which the sample is constructed.
Let us consider the example of the graphics.
Is shown in parentheses:
Memory Type, process technology, memory mode and its manufacturer (in the event that one manufacturer, packer and another, indicated “packer / producer”, for example, «Spectek / Micron»);
ID memory and controller configuration (actual for SandForce);
Firmware version, which were also tested, and sometimes date.
If some data is missing or there are doubts about the reliability (for example, do not understand the packer memory chips), is a question mark ( “?”). This means that they were not detected by me or were lost. Basically it concerns the identity SandForce – not even assumed that the accumulated statistics gradually grow to the scale of a few hundred models. And these data, we will never know, because to catch the same configuration is difficult, but after a year and a half – not impossible.
This test was included in our testing methodology, more recently, a detailed description is given in the relevant material ” Review and testing of SSD-drives: updating the methodology.” Unfortunately, we are unable to carry out a set of tests for all previously studied SSD-drives, so the range of solutions on the charts will be different from the rest of the graphs. Here we have to choose from what is.
This benchmark includes a set of specialized tests of the disk subsystem that reproduce the real situation at the various applications. Each test – a kind of script-track of a particular application, and reproduced are not “dumb” load, and the real work scheme when the application processes the data, and then writes it to disk, read something else, you need to work, processes, stopping any operations with the carrier, and then again starts the action of reading / writing.
The result of this test is a general performance index, calculated on the rather complicated formula, and specific speed performance, in megabytes per second. It must be remembered that the numerical figures take into account the above and pause, so the total value in megabytes per second will be a small figure.
This benchmark allows us to see the speed of file operations within a single carrier. We used 1.7.4739.38088. This test may be dependent on the amount of RAM in the system.
This is more a synthetic benchmark, which is useful because it allows you to test in two modes. The first – respond well to compression of the same type of data stream, the second – a stream of random data, virtually compressible. Accordingly, the final result in both cases is very close to the maximum possible performance of the test vehicle.
Operations with different types of files within a drive
There was a move not only to the new configuration of the test stand, and a new operating system. And with this move we have a problem in this set of tests: the previously used TeraCopy program in Windows 10 environment showed inadequate results. It was therefore decided to abandon it.
From now on tests copying groups of files will be performed forces the operating system itself. For this batch file has been written that automatically copies files and records the time spent on the operation.
Access time in operations of random read and write
Pretty important attribute is the speed of data access time. It is understood that modern SSD drives in this regard have already reached such values that this issue will be of more academic interest. Average access time when reading and writing operations was obtained as a result of testing AS SSD Benchmark version 1.7.4739.38088.
The first minute after completion of the test AIDA64 to the linear recording (as practice shows, some stores after the test is completed in a short period of time continue background operation “garbage collection”).
Please note: tested linear read and write. In fact, in practice, read and write operations very rarely linear, so consumption will “ride” in between “reading – search for data – entry.” But as a whole the ratio between the drives on the level of energy consumption will remain virtually unchanged. Therefore, the figures given in the table, it is possible to navigate.
But we should not forget about the speed characteristics: A drive at a speed of 40 MB / s write one megabyte of data when power consumption of 1 Watt is more economical than the drive B at a speed of 30 MB / s and 0.9 watts.
Fact: The power drive Lite-On in read mode equal to that meter when recording is SLC-mode, and there is no error not (come to mind Crucial models of some lines by 16-20 nm memory, where rates were also similar). The only difference is that read SSD at speeds of 480 MB / s, against the recording speed of 75 MB / s, so in fact equate reading and writing can not (read the specified amount of data will take several times less than its record).
The SLC-mode current consumption to SATA Power connector is 0.39 A, after passing the drive in direct recording to flash the value of the current mode is reduced to 0.30 A. After removing the load the device for approximately 30 seconds showing the internal activity is clearly conducting an operation to consolidate the data recorded SLC-in mode, and power consumption at the same time corresponds to the direct recording in the flash memory.
But in the mode DevSleep Lite-On MU-3 clears more than c: power consumption is so small that it is not just the meter registers, displaying on the display “0.00”. Similarly indicator, if I remember correctly, he could not boast of any SSD tested in the laboratory. Thus CDI, as in the previously discussed SmartBuy S11-2280T , does not show the presence of the support DevSleep.
In short, in the case of Lite-On big surprise was that the major manufacturer in the end took complete solution from another company and in fact confined himself to a simple “branding.”Fortunately, Lite-On has chosen at least not the worst option, and Lite-On MU-3 as the budget drive, get a relatively stout.
Another thing is that Lite-On, it seems, takes as its “mediation” is too big a percentage, and MU-3 as a whole is not only more expensive than the opponents on the same flash memory, but the controller of a higher rank Phison S10 (SmartBuy Revival, Patriot Blast), but also a number of decisions on MLC NAND.
In the Moscow retail, for example, a few less money at the moment, you can purchase ADATA Premier SP600, SmartBuy Ignition 4 and Transcend SSD370. They will be able to offer in no way worse, and even better performance, especially with respect to measurable amounts of data write operations (more than a few gigabytes at a time). And for a similar cost can watch Kingston SSDNow V300 (MLC NAND, though with a SandForce SF-2281) and OCZ Trion 150 (similar to Lite-On MU-3 flash memory, but a more efficient controller Phison S10), which are provided have a Russian guarantee Unlike yet unclear in this regard Lite-on.
Short summary: at current prices is no little interest, and yet do not run to the store, and stifle the money in my hand, and we continue to monitor developments.