Xiao Biao Note: Happy New Year! Good Wen wanted to send you a Taiwan iPhone X for the New Year! # # New Commodities Warfare Call for Papers is in full swing. Share your experience in purchasing new products, your New Year shopping list, and your New Year product reviews. Participate in the event to win an iPhone X prize, along with multiple gifts such as Kindles, JD gift cards, and more! Check out the event announcement and learn more.
You come over! Run a minute!
Because of its excellent disk performance, SSD hard drives are now commonly installed. Even if the budget is small, it's recommended to reserve at least 128GB of SSD storage. After all, using an SSD as a system or software drive offers a much better experience compared to traditional mechanical hard drives. With many manufacturers entering the market, some test software is used to promote their products. If certain results are high, vendors will highlight them. As consumers, we must understand the significance of these results, which can be very valuable. However, not every performance feature is worth paying extra for.
This article will introduce the details of SSD testing software. Since I'm not a professional, I won't expect too much expertise, just a casual look. Before introducing each software, I'll first test the SSD that has just been purchased. After the test, I'll also share my findings!
---------------------
HP S700 PRO Series 512G Solid State Drive Unboxing
SATA SSDs are certainly not as fast as M.2 SSDs. However, since some older platforms do not have M.2 interfaces (think of using PCIE riser cards), SATA SSDs are still widely used. This time, I'm starting with a HP product: the HP S700 PRO series 512G SSD.
I've always been curious about how PC machines and notebook manufacturers design their SSDs. The HP S700 PRO series is available in 128GB, 256GB, and 512GB sizes (JD page also lists 1TB, but it hasn’t been sold yet). It uses 3D NAND TLC chips, and I’ve chosen the expensive 512GB version.
â–¼ The packaging is quite stylish, with a blue and white background. The front shows images of the SSD, with a 512GB tag next to it. Ads on the package mention high-speed read and write, durability, silent operation, and low power consumption.
[Image]
â–¼ On the back, there's basic information and the serial number of the SSD. There's no stated speed, but the JD promotion page mentions sequential read speed of 565MB/s and sequential write speed of 520MB/s.
[Image]
â–¼ The side of the package features the HP logo.
[Image]
â–¼ After opening the package, the SSD is placed in a plastic bracket, likely for shock resistance. Along with the manual and warranty card, there are only 4 screws. PS: That screwdriver is mine.
[Image]
▼ The SSD shell is made of aluminum, with a fine sanding texture. The front has a white "HP" logo. The official weight is 45g ± 5g, and the thickness is 6.7mm.
[Image]
â–¼ The back of the SSD mainly displays the model, capacity, S/N, and P/N code.
[Image]
â–¼ Side view of the SSD.
[Image]
▼ The SSD shell doesn’t use screws; it’s snap-type. It should be possible to open it with a screwdriver, but there's a warranty sticker in that position. Be careful when opening it.
[Image]
▼ Since I didn’t open it, I can’t see the internal components. The promotional page shows the HP logo on the controller. The model is H6028 G AB, which supports 4-channel, NANDXtend ECC technology, and DRAM Cache. These are similar to features found in the Huirong SM2258 controller. As for the flash memory, it's known as 3D NAND Flash. We’ll focus more on the software aspects and avoid going too deep into hardware.
[Image]
----------------------------
Commonly Used SSD Test Software
CrystalDiskInfo
Abbreviated as CDI, this tool is primarily for health diagnosis. Once started, you can check the health status of the hard drive. If the temperature is too high, look at the number of power cycles and the time. If it's too long, it may indicate that the drive isn’t brand new. After installing and formatting the hard drive, you usually check it first.
[Image]
AS SSD Benchmark
Compared to other testing tools, AS SSD Benchmark is a standard program designed specifically for SSD testing. It not only tests read/write speeds but also checks the installed SSD driver status and whether it is 4K-aligned.
After selecting the hard drive to test, first look at the green text in the upper left corner.
Storahci is for Windows 8/10 Microsoft drivers (MSAHCI is for Windows 7), typically used on AMD platforms, though it may not be installed on Intel platforms.
Iastora is the Intel driver, naturally for Intel platforms.
Stornvme is the NVMe driver.
1024-ok; 2048-ok; 4096-ok, all 4K-aligned, differences between them aren't significant. I think the difference isn't large.
[Image]
Press start to run the test. Speed scores can be expressed in units of "IOPS" in addition to MB/s.
“IOPS†refers to the number of read/write operations per second of the SSD, which is an indicator of throughput. The higher the value, the better the performance of the SSD in actual usage.
The AS SSD Benchmark has four running scores (read and write), and the test data can be set to 1G, 3G, 5G, 10G, with scores given:
Seq: Continuous read and write speeds. AS SSD will first use 16MB as a unit, continue writing to the tested partition to generate a file up to 1GB in size (default), then read the file in the same unit size, and finally calculate the average score and give the result. Upon completion of the test, the test file will be deleted immediately.
4K: 4K random read and write speeds. This is where SSDs truly excel over mechanical hard drives. Small files like DLLs are frequently called when software is running, especially during startup. The advantage of SSDs is that these small files can be read and written quickly, improving disk performance without stalling.
A 1 GB test file is generated at a test size of 512 KB, and then written and read at a random 4 KB unit size within its address range (LBA). Until this range is reached, the average result is calculated and the result is given. Due to the generation step, this test will generate a total of 2 GB of data to the hard disk. After this test is completed, the test file will be temporarily retained.
4K-64Thrd: A random 64-row deep read/write test. The test mode is multi-threaded concurrent read/write. The software will generate 64 16MB test files (a total of 1GB), and then write and read tests in these 64 files at a unit size of 4KB. Finally, the results are averaged. This step also produces 2 GB of data writes. Upon completion of this test, the test file will be deleted immediately.
This multi-threaded concurrent performance test can be used to measure the NCQ (Native Command Queuing) gap. Generally speaking, the AHCI mode installation opens NCQ. At this time, the gap between 4k and 4K-64Thrd is very large. If it is the IDE mode, it is not much different from the ordinary 4KB random, so this test is more to judge server-level SSD performance, many people think that the reference value for individual users is not big. In the next test software introduction, I will elaborate on this again.
Acc.Time: Average access time, seek time. Read is to test randomly addressed 4KB file (full LBA area), write is to test randomly addressed 512B file (full LBA area), test write addressing will generate a 1GB test sample file, and finally averaged Results give results. If you use a mechanical hard drive to run this test, you may have a pain in the egg. It takes about 1 hour to run.
Score: Comprehensive scores, scores give a very intuitive impression, running points is higher or lower, but the rationality of the score is always a bit controversial.
AS SSD's scoring standard formula:
Reading score = (continuous reading × 0.1) + 4KB random + 4KB 64Thrd
Write score = (continuous write × 0.1) + 4KB random + 4KB 64Thrd
Total score = (continuous writing speed × 0.15) + (continuous reading speed × 0.1) + (random 4KB writing speed × 2) + 4KB random reading speed + 4KB 64Thrd Random Read Speed + (4KB 64Thrd Random Write Speed × 1.5)
Intel driver run results (now the latest AS SSD Benchmark version should be 2.X, I did not update)
[Image]
Microsoft Runs Results (AMD RYZEN Platform)
[Image]
It can be seen that the performance of AMD RYZEN and Intel platform does not differ greatly in terms of disk performance. Even RYZEN has a slight advantage in continuous read and write speeds, but it is slightly less than 4K random read/write speed on the Intel platform, especially in terms of writing. The gap, of course, is much different in the home experience.
In addition, the AS SSD Benchmark default test model is completely random, so the SandForce series master's compression skills have not played well. Some SandForce master SSD test scores in the software are very low.
CrystalDiskMark
CrystalDiskMark (hereinafter referred to as CDM) is a test software developed by the Japanese and does not have an Acc.Time test project of AS SSD Benchmark, so if you want to test a mechanical hard disk, you can barely, but the time will not be short; personal feeling if you test SSD The time will be longer than the AS SSD Benchmark (5 runs by default).
The default model of the test software is also random, but the test data options in the settings can be changed to 0x00, that is, fill in the blank with 0, this will affect the hit rate of the write-back cache for the general disk, but for the use of For SandForce-hosted SSDs, the impact on test scores is very large. Because SandForce's Durawrite technology is compressible, when the test data model is filled with 0, the compression capacity will be close to infinity, and the score will be skyrocketing. With the random data model, the test results will be difficult to read, and of course new. The SSD is not controlled by SandForce, so no matter which performance you choose, it will not be much worse. You can see that the HP S700 PRO 512G is not much different in the two modes.
In addition, the test software runs 5 times by default; each time you test 1GB (up to 32GB) of data, you can of course change it.
[Image]
[Image]
In the 5.X version, the test project processed Seq and 4K with AS SSD Benchmark. Different Seq Q32T1 and 4K Q32T1 appeared, where Q refers to the queue depth (Queue Depth); T refers to the thread, multi-threaded The test is the same as the 64Thrd of the AS SSD Benchmark, but the CDM is more flexible, and the queue depth and multithreading can be set by itself.
The Queue Depth (QD) is the number of read/write instructions sent each time, increasing the io depth of the device, feeding more IO requests to the device at a time, giving the device the opportunity to arrange for consolidation and internal parallel processing, improving overall effectiveness. At the same time, you can use multiple threads to execute concurrently and fill the io depth. For the layman, we do not need to understand the concept of queues and threads, they are all performance of a SSD in a multi-threaded environment. For example, the meaning of Q8T8 is to test with 8 threads at the same time. Each thread executes 8 queue depths. It can be understood that the total depth of QD is 64. Here we have to say that the next SATA solid state drive supports up to QD32; and the nvme protocol SSD has a maximum depth of 64,000.
Of course, this kind of test is also full of controversy, because the high queue test can only exert its performance on applications with high requests such as database/website server. Many people think that the daily application of our system will also use QD3, the level of QD4. , QD1 is the level of the real daily use, so the QD32 (or higher queue depth) test results given by major manufacturers are some show muscles. But this is, after all, an important difference between SSDs and mechanical hard disks. For tasks with high queue depths, since there is only one magnetic head in the mechanical disk, only the tasks in front of the queue can be processed at the same time. The remaining tasks can only wait, even No response. Of course, if the SSDs are compared, the QD3 and QD4 levels can be measured. Higher queue testing may have little significance for civil experience.
The high queue depth score in the 5.X version defaults to Q32T1, and to the 6.x version, the Q8T8 appears by default, and the total QD depth reaches 64, which exceeds the QD32 supported by the SATA SSD, so it is easy to see that there is bias The NVME SSD test means something.
[Image]
ATTO Disk Benchmark
ATTO is a U.S. company that does data processing solutions and is headquartered in New York. Its main product lines are disk array cards, optical fiber or network bridges, 10 Gigabit Ethernet cards, fiber switches, and data management software. This ATTO Disk Benchmark software is what their company developed for customers to test the performance of their own products, of course, may be due to the versatility of disk testing software, and because software is a free relationship, a lot of users and manufacturers like to take this software To test the U disk's maximum data throughput bandwidth, SandForce master control master is also used by SSD manufacturers as a measure of official nominal value software.
The main settings that can be changed in the test software are:
Drive: The logical drive letter for this test.
Transfer Size: The test data block size, the default is from 0.5KB (512B) ~ 8192KB (8MB)
Total Length: By default, the test data length is 256MB, and the maximum can be set to 2GB.
Force Write Access: Force Write Access. When this feature is enabled, the write data is not optimized by the data cache on the array card. But actually run the actual disk performance (with disk cache and system cache optimization).
Direct I/O: Also known as synchronous I/O. When this function is enabled, it means that after the command is issued, it must wait until the hardware responds to calculate an IO. If it is not enabled (asynchronous IO or buffered IO), after the instruction is issued, it is not necessary to wait until the hardware responds to continue sending the next instruction. The previous instruction will sort and wait for processing, which indirectly equals the data cache speed on the array card. Or one of the file system cache speeds (depending on system cache settings).
If you don't use an array card, this will not be the same if you open it; if you don't have direct I/O turned on, whether it's a read or write test, it's all running in the system data cache. Improve the speed of small file transfers, not real disk performance data. It can be seen that the HP S700 PRO 512G has an abnormally high reading and writing performance when the direct I/O is turned off (1024k and 2048k are abnormally low).
[Image]
I/O Comparison: After the selection, the test pattern will be added. It can be selected randomly, or it can be filled in with 0.000000 or and so on, plus Increment, Decrement And, like the test model in the CDM, if the test data model is or the like, the compression capacity can be close to infinity and the results will be very beautiful. If it is a random model, it will be soft. In addition, there is an option to run continuously. This hooking means that the test continues until it finds an error. The same HP S700 PRO 512G performance difference between these two modes is not very great. As with CDM, the HP S700 PRO 512G is not much different in the two modes.
[Image]
Overlapped I/O: When this is selected, there will be Queue Depth options, which are also described in detail in the CDM. The ATTO version here is 3.2, and can only be set to 10, the default is 4 .
Neither: Selecting this means running under Queue Depth 1, and it is the closest thing to our daily operation.
It can be seen that the performance of the HP S700 PRO 512G with a QD of 4 is significantly better than "neither" (QD = 1), especially small files are much faster.
[Image]
The ATTO test is a continuous disk read/write performance under extreme conditions, and the test model used has high compressibility. However, virtually none of the program's startup and execution processes are continuously read and written. The representative of the compression algorithm is the SandForce master. When the master SSD software performs the ATTO test, the data is like an elephant being compressed into the size of a mosquito. Because the amount of data actually written into the flash memory is small, the write operation is completed quickly, and the externally measured write speed approaches the SATA 6Gbps interface bandwidth speed. When reading these completely compressed data, a mosquito is reduced to the size of an elephant, as if a large amount of data was read out in a flash. At this time, the externally measured reading speed is also close to the SATA 6Gbps interface. Bandwidth speed. Obviously, the value of this reference is not high. So when the SandForce master SSD was prevalent, this software was very much used, but now there is basically no such SSD product, so this ATTO is less and less used.
Anvil's Storage Utilities
Compared to the decline of ATTO, Anvil's Storage Utilities now uses more and more frequency, the test part is divided into six parts: Seq 4M, random 4K, random 4K QD4, random 4K QD16, random 32K, random 128K, There are only four entries in the write section: Seq 4M, Random 4K, Random 4K QD4, Random 4K QD16. You can see that reading is more important than writing. And the software also has scores, compared to the first three software, the software is more flexible, you can independently test separate projects. In the lower left corner of the software you can see the system and CPU, motherboard, memory information; the lower right corner can see the drive status and 4K alignment, enough to replace the function of the AS SSD Benchmark. In addition, the size of the test file can be set to 32 GB as well as the CDM.
[Image]
In the benchmarks option, you can perform independent IO read and write tests (as well as mixed modes), and you can specify the number of queues (read up to 128, write up to 64) and the size of the test file (from 1K to 512K , the default is 4K).
[Image]
In the setting interface, 100% can be compressed (filled with 0) to 100% incompressible for a total of 6 files, of course, this is to take care of the main control of the compression algorithm, generally select 46% (applications) on it.
[Image]
The software can also perform endurance tests and repeatedly erase and write SSDs. Of course, due to the damage to the life of SSDs, the general public should not play.
[Image]
The characteristics of this test software are summarized in two words:
TxBENCH
TxBENCH In addition to the basic test items, click on the task options in the lower right corner is also free to set the test block size, queue depth, and so on, see just introduced several software, the project here should be able to understand.
[Image]
The advanced benchmark is a full-write test (both FILE and RAW are supported).
[Image]
Data erasing is an erase function. There are four modes: standard Secure Erase (Erase FTL layer), Enhanced Secure Erase (Erase FTL and clear NAND flash memory), all SSD space in Trim, overwriting all sectors written to the entire disk, and manual Trim function.
[Image]
Drive information can see information more comprehensive than CrystalDiskInfo, but it's a little professional and it's all in English, so we didn't study it carefully.
[Image]
PCMark8
PCMark series is actually to test the overall performance of the PC, the operating system requirements are at least Win7, WIN8/10 performance will be better, of course, can also be used to individually measure the SSD.
PCMark 8 uses a sub-item test, which is Home, Creative, Work, Storge, and Applications. Among them, Storge mainly tests the storage performance and stability, selects the Storge option, and then selects the hard disk to be tested.
[Image]
PCMark8 test items include: "World of Warcraft" and "Battlefield 3" game loading test, Photoshop, Adobe InDesign, Adobe After Effects, Adobe lllustrator, Microsoft Office Word, Office Excel and Office PowerPoint and other ten tests. These tests are more capable of effectively simulating some of the user's behavioral needs in daily life compared with the theoretical tests of the previous software. PCMARK8 has a test data volume of about 50GB, and can test the master's "handling" capability and the actual writing of NAND particles. Speed can better reflect the overall performance of SSDs in the general home, games, and small SOHO environments.
[Image]
However, it takes more than one hour to perform the PCMARK8 test, so many people will not take such a long time to do the test. In addition there is the latest PCMARK10, and at this stage you have to pay for it.
summary
SSD is still a generation further than the next generation. The mastermind like SandForce has no market (let's think of the latest intel vulnerability). In the past, it was necessary to worry about the SandForce master's SSD shamelessly. It is necessary to pay attention to the setup problem of the test software, and now the test basically does not have to worry about it, so now the integrated read and write speed and 4K random read speed are the most important indicators. Pointing at large files pointing to small files is as simple as that. There are some differences in the performance of different software, especially continuous reading and writing (define the size of large files will be different), but in general will not be too much worse. Of course, there are some more specialized software, such as IOMeter, which has other criteria for judgment. I will not introduce it here (because I do not understand it).
Here are a few tips for getting high scores on SSDs.
1. The performance of SSDs with different capacities differs. The difference is that the single-core capacity of flash memory particles is different from the number of flash memory channels. The larger the capacity of SSD, the larger the single-capacity of flash memory particles and the greater the number of channels, the greater the throughput data volume. Of course, the faster the data processing speed is, the more money will be spent, so strictly speaking, this is not a skill.
2. The power options are high performance, and even some of them turn off the power-saving of the motherboard and let the CPU work at the highest frequency. The SSD performance can be better.
3. It is much better to test empty data.
4. Some SSDs in the system should choose to close the write cache buffer refresh (a little risk), otherwise write run tests will be very low.
------------------------
High-level tips
Is there a difference between 10 to 20% in SSD performance? In fact, even if the gap is even higher, it may not be a huge difference in the experience. However, as an ordinary user, it is also necessary to pursue stability. For example, it is often heard that some SSDs are loaded with large amounts of data, or they experience slowdown after a period of time. The situation can also be tested in this area before the actual use of the software.
HD Tune Pro
HD Tune is a very old speed test software, suitable for mechanical hard disk, not suitable for ssd test, but through the test of reading and writing, will draw a speed curve, this speed value may be of little significance reference, but through Observing the flatness of this curve can have a simple understanding of the stability of ssd, representing the trend of performance degradation of continuous writing and reading. The more stable the course is, the more stable the HP S700 PRO 512G of this 3D NAND TLC particle is basically a straight line, and the SSD of my previous MLC particle could not have such a flat curve.
[Image]
Increase test data block size
Flash memory particles, if they are TLC, will generally use SLC Cache technology.
TCL SSD uses a part of TLC as SLC (or MLC) to enhance the experience of TLC SSD. This is the SLC Cache technology.
The virtual 1GB SLC requires 3GB of TLC. If it is a full-scale simulation, it requires three times the capacity and the cost is too high. Manufacturers can only simulate part of the TLC as SLC, and this part of the faster space as a full cache. Increasing the amount of test software data to exceed the size of the SLC Cache cache space requires relying on the master to timely clear the cached data to free up more cache space. If the data is continuously written, the master control capacity is exceeded. If you do, you can only write to the ordinary TLC space, and the speed will return to the true level of the TLC. Running the score will naturally be greatly reduced, and it will expose the real body of its TLC chip.
I use the CDM (others can adjust the data block software is also possible), set the test data block for 4G, 16G, 32G, but the results of running sub-basic no major changes, may be HP S700 PRO 512G cache space is relatively large, beyond In 32G, the deceleration will not be measured naturally; it may also be caused by the black technology of 3D NAND TLC, or the master's superior ability to recover waste. It is still very good to test 2D NAND TLC with this method, their SLC cache space will not exceed 10GB generally, but to HP S700 PRO 512G here, this method is invalid.
[Image]
Increase SSD data volume
Although using the above method, HP S700 PRO 512G did not measure the speed, but I was discouraged, thought of ways to increase the amount of data in the SSD disk, GAT5 game program (more small files, more able to show the actual situation), HP S700 PRO 512G, and keep replicating to make it less available space, and then test its running score
[Image]
Still use the CDM software, you can see if you use the test data block is 1GB, the hard disk occupied to 63%, 76%, 89%, HP S700 PRO still does not slow down; set the test data block is 32GB, occupy 76%, 4k reading slightly decreased, the other is still strong. When the occupancy rate reached 89%, the 4k readings continued to decline, and the written results fell sharply. Finally successfully tested the HP S700 PRO speed down situation. Of course, even if the occupancy rate is so high, continuous reading still does not slow down, and the magnitude of 4k reading speed reduction is not too large. If it is used as a system disk, I believe that it can still work normally (mainly reading work).
[Image]
--------------------------
to sum up
The test software is just like playing. There is no need to measure it often. Because the software test will generate a test file, the bigger the setting, the less the data buffer will interfere, and the result will reflect the true performance of the SSD, but the disadvantage is that it will be serious. Impact on the durability of SSD (write too much data to affect P / E), use all right and do not always run to play, those SSD's durability or to the gods to do it, we look like it!
HP S700 PRO 512G's stability is its characteristics, it is difficult to make it slow down, but the manufacturers are not doing charity, black technology SSD prices naturally will not be cheap, before 512G sold 1399 yuan, has now risen to 1699 yuan, The price/performance ratio is not very high. If stability is required, it can be started. Those who do not mind slowing down can start with more cost-effective products.
Fabric Bluetooth Speakers,Fashion Fabric Bluetooth Speaker,Portable Bluetooth Speaker,Tws Bluetooth Speaker
Comcn Electronics Limited , https://www.comencnspeaker.com