Date: May 3, 2018
Author(s): Robert Tanner
While ADATA has a wide variety of budget SSDs on offer, it also has a growing number of more premium offerings. Under review today is one such drive, the SATA-based SU900, an MLC drive with a five-year warranty. Let’s see how it performs compared to other drives in the same budget range.
Solid-states drives have been around for over a decade, which is why it is amazing to see so many PCs still shipping with spinning platter drives as the only option. It is true SSDs have a long ways to go for replacing spinners for basic data storage, but as the OS drive for desktops and laptops, SSDs are an essential piece of kit.
Nothing bogs down Windows (or your OS of choice) more than waiting on random disk access queries, or the system sitting idle for seconds at a time while a large office document is found and loaded.
Today we are back to take a look at ADATA’s SU900 SSD. We have already reviewed ADATA’s SU800 and budget drives such as the discontinued SP550 previously, but the SU900 is ADATA’s current high-end model for the SATA 6Gbps (SATA 3.0) market, only surpassed by the XPG branded SX950.
Solid-state drives themselves have long ago grown so fast as to be bottlenecked by the SATA 3.0 interface which is why M.2 drives have become the de facto choice for high performance models. But as a daily driver, SATA 3.0 drives generally remain more economical and offer most of the benefits of SSDs. Random access latency generally remains the same whether it is a SATA 3.0 or PCIe x4 capable M.2 drive, the only major differences are peak performance, raw throughput potential, and for premium drives, sustained performance over long duration heavy write workloads.
The SX950 is ADATA’s premier SATA interface offering, though in comparison to the SU900 its only differences are a longer warranty and increased over-provisioning of the flash. The next rung up the performance ladder is a big jump to ADATA’s XPG GAMMIX S10 entry-level M.2 drive, which we will be reviewing next. As the SU900 shares the same Silicon Motion SM2258 controller as the SU800 we can expect similar levels of performance.
The main difference is the SU900 upgrades the 3D NAND from TLC to MLC which should offer better sustained performance and appreciably longer flash endurance. For typical consumer use, a modern MLC SSD will likely outlast the system it is running within, which is probably why ADATA raises the SU900’s warranty to five years.
That isn’t the only thing raised, the SU900 commands a much higher price premium over its SU800 sibling, as well placing it fairly close to budget M.2 drive prices. It is important to highlight that the 128GB model will suffer from reduced performance as there are insufficient NAND chips to fully populate the SM2258’s four channels.
Saving a few bucks to buy a 128GB model SU900 won’t end well, consumers will see much better performance by choosing a 240GB SU800 than a 128GB SU900 given the similar pricing. Unfortunately, this is simply due to the ever increasing capacity of NAND chips and is true of pretty much all 120/128GB SSDs made today regardless of brand. It is also the same reason why 64GB drives haven’t been offered in several years.
|Form Factor||2.5″ 7mm|
|Controller||Silicon Motion SM2258|
|DRAM (Nanya DDR3)||256MB?||256MB?||512MB||512GB?||1GB?|
|Flash NAND||Micron 3D MLC|
|SLC Write Cache||Yes (dynamically allocated)|
|Sequential Read||560 MB/s|
|Sequential Write||500 MB/s||520 MB/s||525 MB/s||520 MB/s|
|4K Random Read||40K IOPS||80K IOPS||85K IOPS||90K IOPS||85K IOPS|
|4K Random Write||85K IOPS||90K IOPS||85K IOPS||80K IOPS|
The SU900 offers the same pseudo SLC fast-write caching as the SU800. Instead of a fixed allocation, the drive will dynamically adjust the SLC fast-cache size based on overrun of the SSD’s DRAM cache, though the maximum size of the fast-write cache is not given. The DRAM cache size itself is also not listed, although by opening the drive we can confirm the 512MB model houses 512MB of Naya DDR3 cache. It is likely smaller capacity drives that utilize a 256MB DRAM buffer, but that is simply conjecture at this point.
It is worth noting that while a 2TB model is listed on the specifications, it does not appear to have been released to market at this time. ADATA includes both a small 7mm to 9.5mm height shim, as well as an aluminum 2.5″ to 3.5″ bracket to allow the SU900 to fit into practically any laptop or desktop.
|SSD Test System|
|Processor||Intel Core i7-4771 @ 3.7GHz|
|Motherboard||ASUS Z97-A – BIOS 2801|
|Memory||Crucial Ballistix 8GBx4 DDR3-1600|
|Graphics||EVGA GeForce GTX 750 Ti SC|
|Storage||ADATA SP550 480GB (SATA)|
ADATA Ultimate SU800 512GB (SATA)
ADATA Ultimate SU900 512GB (SATA)
Crucial BX200 240GB (SATA)
Crucial m4 (C400) 256GB (SATA)
Crucial MX300 525GB (SATA)
Intel 520 240GB (SATA)
Intel 730 240GB (SATA)
Kingston Predator 240GB (PCIe)
Samsung 950 Pro 512GB (NVMe PCIe)
Transcend SSD360S (SATA)
|Power Supply||BitFenix Fury 650W|
|Chassis||Cooler Master HAF 932|
|Et cetera||Windows 10 Pro x64|
To ensure all drives have fair representation regardless of capacity in our testing, we use IOMeter to conduct continuous file writes beyond the rated capacity of each SSD. This will make sure every block has been written to at least once, including the unaddressable spare area present on all solid-state drives, ensuring all drives begin testing in a normalized “dirty” state.
SATA drives are tested on an Intel 6Gbps SATA port, while PCIe cards will use the PCIe Gen 3 x8 slot to the processor. M.2 drives will make use of an NVMe compatible PCIe adapter card in this same slot to bypass the M.2 x2 interface bottleneck (some motherboards only offer an x2 lane M.2 port), allowing drives to utilize the maximum bandwidth (PCIe Gen 3 x4) that the M.2 interface supports. Where applicable, we will ensure M.2 drives are using the NVM Express protocol.
The test system’s processor has had Intel’s Speedstep and Turbo Boost functionality disabled, and all cores are locked to the same 3.7GHz frequency in order to ensure maximum consistency between test runs.
Futuremark PCMark 08: Normal settings are used including the default setting of three runs. This gives us an extended test that takes roughly 45 minutes to complete, resulting in a sustained load that will thermally challenge drives. The free (demo) version is available on Steam for general use.
AS SSD: This handy little utility was custom-written for solid-state drives. We selected this program as by default it utilizes incompressible data and bypasses the Windows cache. This is another free program anyone can use to quickly test or verify normal performance of an SSD, available here.
ATTO: The bar graphs generated by this utility provide a huge amount of data in a condensed format. It is extremely handy for detecting problematic file sizes, performance bottlenecks, or simply inconsistent performance that can be hidden when viewing final performance averages. Though this is significantly less of a problem today as solid-state drives have matured, it remains an essential performance verification tool of any benchmark kit.
RoboCopy: This is Windows’ command-line utility included in most versions of Windows and provides for features beyond the default file copy handling of Windows Explorer. More information is available here. Windows Defender is also disabled to prevent bottlenecking the folder copy. We utilize a 9.81GB Movie file, and separately a 1.6GB Folder with 11,511 files across 449 subfolders and measure the performance to copying the data to and from a RAMdrive.
RAMdrive: Primo Software’s Ramdisk utility is a handy, convenient way to create a 12GB direct-IO ramdisk, meaning no indexing or caching is utilized and removes any bottlenecks associated with utilizing another storage device. File performance of the RAMdrive varies by I/O size but exceeds 6GB/s above 16KB IOPS with an overall peak of 12GB/s.
dBpoweramp R15.3: We transcode 500 FLAC encoded music files (14.4GB) into .M4A with the Apple Lossless Audio Codec (ALAC) for playback on Apple devices. As no lossy compression is taking place, this minimizes the CPU workload while maximizing the amount of data written to the destination drive. Unfortunately this test is bottlenecked by our quad-core processor and would show more variance in results for systems with greater than four cores as dBpoweramp can fully load almost any size processor.
IOmeter – The most powerful piece of storage testing software freely available, IOmeter allows the creation of practically any kind of test workload. We first utilize it to dirty every new SSD before we test the four basics of every solid-state drive. We measure random 4K read and writes at 1QD and 32QD, as well as 2MB sequential read and writes with a queue depth of one. Each one of the six tests is run with a ten minute duration.
Additionally we have added Database, File Server, and Workstation scenarios back into our regular testing. We have updated and modified the original three workload scenarios to bring them more in line with changes in workloads, access patterns, and disk IO sizes. These results are not comparable to the test scripts originally created and included with IOmeter nearly two decades ago.
If you are still reading our benchmark profile, thank you! It is not easy to devise tests that can tax a modern solid-state drive so if you have any suggestions or perhaps know of any applicable software that you would like to see added to our test regimen, please let us know. We are especially eager to add disk I/O trace recording and playback to replace the batch file testing we utilized in past storage reviews. We are still looking into how to make this possible, so if you know of a piece of software that doesn’t cost a few grand please drop us a note!
As we expected the SU900 offers 256MB/s in bandwidth, higher than the SU800 which achieved 239MB/s yet Futuremark still gave the SU800 a higher overall score. While 31 points is negligible it is still perplexing as the results were repeatable. A difference in how the two drives allocated their respective pseudo fast-write SLC caches may be in play. 256MB/s ties it with the BX300 and places it just behind the Transcend, both of which also utilize MLC NAND.
One thing to keep in mind is Iometer creates a test file that spans the entire size of the SSD. So it should come as no surprise that this mostly negates the SLC caching performance of the BX300 and other SLC-caching drives. While this allows us to see the raw performance of the NAND, ten minute saturation workloads are not even close to representative of the sort of work consumer drives would see in normal consumer use.
With that caveat out-of-the-way the S900 delivers strong performance. The Random 4K Read IOPS is in on par with the stronger drives, but when increasing the queue depth to 32 the drive is able to take the top spot in the SATA drives we have tested, only just below the M.2 drives that aren’t limited by the SATA interface bottleneck.
Performance continues to be strong in every test save the sequential read result, though for sequential writes the SU900 is able to inch out just ahead of the SU800 as we would expect. Iometer is not a consumer-drive friendly program, but the strong QD32 results the SU900 delivers makes it pretty clear the drive is quite capable of handling the heaviest multitasking workloads consumers might have.
Our new Iometer scenarios are more appropriate for enterprise and data-center SSDs, but due to the dearth of consumer programs that will stress modern SSDs, we have recreated the original Iometer usage scenarios with some modern, updated settings.
Each of the three scenarios runs for twenty minutes each with minimal time given between runs for the SSD to perform maintenance tasks like garbage cleanup or clearing any SLC fast-caches (if applicable). These tests are a worst-case scenario intended to highlight the underlying differences between solid-state drives that are well-hidden by new technologies such as fast-write SLC caching.
The SU900 is the first ADATA drive to endure these three scenarios and once again puts up a strong showing, taking second spot in the Workstation scenarios, though dropping to third in the file server results. Sustained low queue depth writes will eventually fill up any SLC cache, making it moot, which is what happened here. We can’t imagine most consumers buying these drives would use them as file servers anyway. The SU900 recovers with a very good finish in the Database scenario, with results close to the BX300.
The seven-year-old Crucial m4 SSD does not have an SLC fast-cache and even for its time did not have the best random IOPS performance. As far as intensive workloads are concerned, users on some of the early generation SSDs may see tangible benefits to upgrading to a modern solid-state drive.
|AS SSD – Latencies||Read||Write|
|ADATA SP550 (480GB)||0.035ms||0.034ms|
|ADATA SU800 (512GB)||0.032ms||0.035ms|
|ADATA SU900 (512GB)||0.032ms||0.037ms|
|Crucial BX300 (240GB)||0.031ms||0.029ms|
|Crucial MX300 (525GB)||0.043ms||0.527ms|
|Crucial m4 (Micron C400) (256GB)||0.037ms||0.049ms|
|Intel 520 (240GB)||0.110ms||0.173ms|
|Intel 730 (240GB)||0.032ms||0.032ms|
|Kingston Predator (240GB)||0.041ms||0.047ms|
|Samsung 950 PRO (Microsoft NVMe)||0.027ms||2.670ms|
|Samsung 950 PRO (Samsung NVMe)||0.023ms||0.021ms|
|Transcend SSD370S (512GB)||0.032ms||0.035ms|
The upgrade to MLC NAND makes a stark difference with AS SSD, with the SU900 pulling far ahead of the SU800 in practically every result. Though it is unable to overcome the SLC caching of the Crucial drives in the final tally, it is able to score higher than Intel’s 730 drive and all other SATA drives in our tests. AS SSD is coded to ignore Windows caching, though it defaults to a 1GB test file which fits easily within most SLC fast-caches today. As expected, given the identical controllers, the SU900 posts identical latency results to its cheaper SU800 sibling.
There is negligible difference between the SU800 and SU900 in this test, both deliver 500MB/s or better writes at 128KB access sizes and above.
This program emphasizes better than most the hidden performance limitations that used to be common in the earliest generations of solid-state drives, as the m4 results differ drastically with the drive capped around 280MB/s writes regardless of access size.
Drives like the m4 are still preferable to spinning rust, but the performance differential between it and modern SSDs such as the SU900, particularly under heavy multitasking loads, may be enough to justify an upgrade.
Robocopy is one of the few real-world usage scenarios that can highlight differences between SSDs. The movie file is nearly 10GB which exceeds the 8GB fast write-cache of most drives, which is why the SU900 begins to pull away from its SU800 sibling in the write tests. The read tests have no such limitation and show a bit more variability, through the SU900’s MLC flash edges out again by 18MB/s for the movie transfer.
In an effort to find more real-world tests, we attempted lossless transcoding from FLAC to .M4A with the Apple Lossless Audio Codec (ALAC). Unfortunately, this test remains CPU bound with a Haswell quad-core processor, which goes to show why transcoding and encoding is still relegated to spinning platter drives. The SU900 isn’t able to replicate the 2-second quirk in results. That said, even if we removed the bottleneck by upgrading to a ten (or more) core CPU, it seems a smidge unlikely people would be pairing such processors with interface bottlenecked SATA drives.
ADATA’s SU900 is a solid-state drive that delivers strong performance using a proven Silicon Motion SM2258 controller and 3D MLC NAND. It also includes an optional aluminum 3.5″ adapter bracket and 7mm-to-9mm height shim which allow the SU900 to fit into practically any desktop or laptop system as a drop in upgrade.
While the SU900 is only a marginal performance upgrade over the SU800 series, it upgrades the warranty to five years and swaps the TLC NAND for MLC ,which should give the SSD a longer lifespan than the system it is being put inside. Unfortunately, these upgrades over the SU800 also come with a steep upgrade in price that just doesn’t seem justifiable over the SU800.
Unless you are looking for an SSD that is intended to be used in some uncommon, very high wear application where the appreciably longer endurance of MLC flash is a strong value add, the current $210 price for the 512GB model is just not justifiable. To give some perspective, the SU800 we reviewed currently costs $130.
Even ADATA itself currently offers not one, not two, but three NVMe M.2 class SSDs of the same capacity for less, namely the SX6000, SX7000, and the S10 Gammix that we have in the lab undergoing review! Needless to say it seems we are quickly reaching the point where consumers looking for a solid-state upgrade will begin paying extra for the legacy SATA interface than those buying a higher performing M.2 NVMe solid-state drive.
ADATA’s SU900 may be on a short timer as because as of this writing it was hard to find direct sellers other than ADATA itself via Amazon, the 2TB model never seemed to reach market, and the drives have already been superseded by the very similar SX950 XPG model, which one can also find on Amazon direct from ADATA for $150.
The SX950 retains the same 3D MLC flash and controller, but upgrades the warranty to six years. Undoubtedly buying the same drive as the SU900 at a $60 discount and with a better warranty is the way to go, so we will have to dutifully point consumers toward the SX950 over the SU900. Sadly we cannot end there, as ADATA makes a very similarly named “SX950U” drive as well. The “U” model appears to be an older drive from 2015 with a lower warranty and TLC NAND, so consumers should be careful not to confuse it with the “SX950” non-U variant.
Needless to say this alphabet soup of models available to purchase (often with very minor changes) is becoming a problem; in the process of researching drives it has become readily apparent that ADATA has the most models still offered of any SSD manufacturer in the market. The $150 price point of the SX950 seems like an excellent deal in comparison to the SU900, but that still leaves the SX950 priced equivalent to options like the Samsung 860 EVO or entry level NVMe M.2 drives such as ADATA’s own SX6000, SX7000, and S10 drives.
If it absolutely must be a SATA drive then the SX950 is certainly a capable and reasonably priced option to add to the shortlist. After seeing the SU900’s results we can affirm the SX950 will have strong performance and it is hard to argue against a six year warranty. Just don’t go paying $200 for one.
Copyright © 2017 SmartKevin