The Ultra II Gets TLC With A Dash Of SLC - SanDisk's First TLC NAND SSDWhen Samsung released their 840 Series drive back near the end of 2012, it was the first consumer drive to feature TLC (Triple Level Cell) NAND which took everyone by surprise as up until then, MLC drives ruled the consumer roost. The drive proved to be quite popular with decent performance and stellar power usage along with a very modest price tag. SanDisk has decided to drop some TLC NAND into their second iteration of the Ultra series drives which should provide some solid performance numbers along with budget level pricing. Without peeking at the specifications, we surmised that performance would be above average with low power consumption and we weren't wrong. SanDisk has really been coming on strong in the SSD market and should be considered one of the premiere drive makers given their expertise and access to raw materials and manufacturing. As such, it didn't come as a total surprise to see them out a TLC based drive but there are many in the industry don't believe TLC is a viable long term component for client SSDs. Enter the SanDisk Ultra II Solid-State Drive. Available in capacities of 120GB, 240GB, 480GB and 960GB, it covers all the bases while offering a splendid read/write rating of up to 550MB/s and 500MB/s respectively. IOPS specifications are solid too with reads and writes all above 80K and then some depending on the model. So far, it's only available as 2.5" drive although they could easily make an M.2 or mSATA drive given the size of the PCB in the review sample we received (check out page 2). If you don't recall, TLC is technically a form of MLC (Multi-Level Cell) and contains three bits per cell rather than the two in MLC and the one in SLC. The diagram below shows, at a high level, how this works. With more data per cell, TLC NAND can cram in more dies per package and is generally a little cheaper to manufacture than MLC and a lot cheaper than SLC. The downside is, it's generally slower and because of increased write-amplification its overall endurance is typically less than the others. Having some SLC on board to cache the data before flushing to the TLC greatly reduces the write-amplification and therefore drive endurance is improved.
SanDisk Ultra II Specifications and Features
|MSRP/Retail:||$79.99 / $79.99||$114.99 / $109.99||$219.99 / $199.99||$429.99 / $379.99|
|Flash:||SanDisk 1Ynm, X3 ABL, Toggle|
|Interface:||SATA 6GB/s backward compatible|
|Seq Read (up to):||550MB/s||550MB/s||550MB/s||550MB/s|
|Seq Write (up to):||500MB/s||500MB/s||500MB/s||500MB/s|
|Rnd Read (up to):||81K IOPS||91K IOPS||98K IOPS||99K IOPS|
|Rnd Write (up to):||80K IOPS||83K IOPS||83K IOPS||83K IOPS|
|Max Read Operating:||2.5 W||2.7 W||2.7 W||2.9 W|
|Max Write Operating:||3.3 W||4.5 W||4.5 W||4.6 W|
|MTBF:||Telcordia, Stress Part||1.75 Million Hours|
|Weight:||Individual Product:||58 g||58 g||58 g||61 g|
|Size:||2.5" SFF-8223 & -8201:||70mm x 69.85mm x 100.5mm|
|Environmental:||Operating Temp:||0C to 70C|
|Non-Operating Temp:||-55C to 85C|
|Warranty:||3 Year limited warranty|
- SATA 8Gb/s Compliant
- ATA-ACS v6
- NCQ support upo to queue depth = 32
- Support for TRIM
- S.M.A.R.T. feature supported
- Advanced Flash Management:
- nCache 2.0 - Non Volatile Write Cache
- Multi-Page Recovery
- XOR Recovery
- Dynamic and Static Wear-leveling
- Bad Block Management
- Backgrounf Garbage Collection
- Advanced features:
- Tiered caching - Volatile and non-volatile cache
- Supports multi stream - improves user experience in multitasking systems
- Minimal write amplification - increase endurance and performance
- Support For Thermal Throttling:
- Performance will be throttled in the event junction temperature of critical components is measured to be exceeding the maximum allowable for the product
A Closer Look at the SanDisk Ultra II:Opening the drive consisted of the removal of four screws, each of which were located under the sticker on the back of the drive making removal without detection near impossible. A thermal pad resides on the back plate which covers all of the components on the controller side of the PCB. The PCB is tiny in comparison to the overall drive size. It's like buying a bag of potato chips (or crisps for you folks across the pond). One one side of the PCB we find just two NAND modules. A closer look at the NAND reveals SanDisk branding and part number 4286DYN01A. This is TLC 19nm toggle NAND which we spoke about in more detail on the opening page. There are only four total packages on the board, each with a density of 64GB. Flipping the board over we see more of the NAND, the cache and the controller. Apologies for the lint on the chips, it was left there in the attempt to wipe off the residue left behind by the thermal pad. The SK Hynix DDR3 SDRAM cache is labeled as H5TC2G63FFR and is 256MB in capacity. The Marvell 88SS9190 controller is new and the info we have on it is not very robust. From what we've been able to discern, it's roughly the same as the Marvell 88SS9189 controller except is has four channels, not eight. This makes sense since there are only four NAND packages on board and we are told the higher capacity drives use the Marvell 88SS9189 controller as they can leverage more of the channels. The controller handles all the usual duties and runs a SanDisk custom firmware that does not support encryption or the newer DevSleep function although neither will probably be missed - especially in a budget friendly drive.
Test System & Comparison DrivesThis is our new Z97 test bench and we thank ASUS again for their generosity in providing the Z97-A motherboard for us to use for our testing. All tests were performed on a fresh and up-to-date install of Windows 8.1 Pro x64 with no other applications running while using AHCI mode set through the BIOS. Synthetic Benchmarks were run with the OS loaded on a Crucial MX100 256GB SSD. In between every test, the test drive was secure erased using an instance of Parted Magic. As such, all results should be indicative of optimal performance with the exception being the PCMark 8 consistency test. Power saving modes, sleep and hibernation are all disabled and all components were set to their default/optimized speeds in the BIOS (1304) and are listed below.
Z97 Test Bench
|Intel LGA 1150 Test Platform|
|Intel Pentium G3258|
|ASUS Z97-A (BIOS v.1204)|
|Kingston HyperX KHX16C9B1RK28 8GB|
|Crucial MX100 256GB|
|Antec Basiq BP550W Plus-EC|
|Windows 8.1 Pro 64-Bit|
|OCZ ARC 100 240GB||Indilinx Barefoot 3 M10||SATA III|
|SanDisk Extreme PRO||Marvell 88SS9187||SATA III|
|Samsung 850 PRO 1TB||MEX S4LN045X01||SATA III|
|Crucial MX100 256GB & 512GB||Marvell 88SS9189||SATA III|
|ADATA Premier Pro SP920 512GB||Marvell 88SS9189||SATA III|
|Intel 730 Series 480GB||PC29AS21CA0||SATA III|
|Crucial M550 512GB||Marvell 88SS9189||SATA III|
|OCZ Vertex 460 240GB||Indilinx Barefoot 3 M10||SATA III|
|VisionTek PCIe 240GB SSD||SandForce SF-2200 (SF-2281) x2||PCIe|
|WD Black² Dual-Drive 120GB SSD + 1TB HDD||JMicron JMF667H||SATA III|
|OCZ Vector 150 240GB||Indilinx Barefoot 3 M00||SATA III|
|Corsair Force LS 240GB||Phison PS3108||SATA III|
|Samsung Evo 500GB||MEX S4LN045X01||SATA III|
|Seagate 600 240GB||LAMD LM87800||SATA III|
|OCZ Vertex 450 256GB||Indilinx Barefoot 3 M10||SATA III|
|Crucial M500 480GB||Marvell 88SS9187||SATA III|
|OCZ Vertex 3.20 240GB||SandForce SF-2200 (SF-2281)||SATA III|
|Samsung 840 Pro 240GB||Samsung MDX||SATA III|
|Sandisk Ultra Plus 256GB||Marvell 88SS9175||SATA III|
|Corsair Neutron GTX 240GB||LAMD LM87800||SATA III|
|Intel 520 Series 240GB||SandForce SF-2200 (SF-2281)||SATA III|
|OCZ Vector 256GB||Indilinx Barefoot 3 M00||SATA III|
|Kingston SSDNow V300 120GB||SandForce SF-2200 (SF-2281)||SATA III|
|Samsung 830 Series 256GB||Samsung S4LJ204X01-Y040||SATA III|
CrystalDiskInfo 6.1.12 Readout:For the SanDisk Ultra II 240GB drive, the readout on CrystalDiskInfo shows that both NCQ and S.M.A.R.T. are enabled, as well as TRIM. This is a great free tool to see lots of detailed information about the drive such as the firmware version for the drive is X3100RL. SanDisk also has a dashboard tool that will display similar information along with tools for partitioning and secure erasing the drive. Let's have a look at the performance with some synthetic benchmarks followed up by some real world tests.
ATTO & AS-SSD Benchmarks
ATTO v2.47ATTO is one of the oldest drive benchmarks still being used today and is still very relevant in the SSD world. ATTO measures transfers across a specific volume length. It measures raw transfer rates for both reads and writes and places the data into graphs that can be very easily interpreted. The test was run with the default runs of 0.5KB through 8192KB transfer sizes with the total length being 256MB.
ATTO - Intel Z97 Platform:Benchmark Results: ATTO shows the best case performance scenario and we find that the SanDisk Ultra II puts out some strong numbers - especially on reads.
AS-SSD (1.7.4739.38088) Benchmark - Intel Z97 Platform:We have been running the AS-SSD Benchmark app for over some time now and found that it gives a broad result set. The programmer has worked very hard on this software and continues to make updates often so if you use it, show him some love and send him a donation. There are now three tests that are found within the tool and we'll show the results from two of them. Benchmark Results: Not much of a dropoff in performance coming from the ATTO benchmark which is always a good sign. Compared to OCZ's budget ARC drive, performance was better in nearly every metric. Benchmark Results: As expected, the level of data compressibility plays not part in performance as we see nice zero slope graph lines.
CrystalDiskMark & Anvil IOPSCrystalDiskMark is a small benchmark utility for drives and enables rapid measurement of sequential and random read/write speeds. Note that CDM only supports Native Command Queuing (NCQ) with a queue depth of 32 (as noted) for the last listed benchmark score. This can skew some results in favor of controllers that also do not support NCQ.
CrystalDiskMark 3.0.3 x64 - Intel Z97 PlatformBenchmark Results: The Ultra II actually did better on this benchmark than the AS-SSD which is unusual as they generally output similar scores. Overall, the numbers look better than what we saw on the OCZ ARC drive - especially in reads. There's a relatively new benchmark called Anvil Storage Utilities that is in beta but close to production. It's a very powerful tool that measures performance through a variety of tests which can be customized. Since some of the tests more or less duplicate what we get from other benchmarks we use already, we decided to use the IOPS (Input/Output Operations Per Second) testing on 4kb file sizes at a queue depth of 32. IOPS performance is something SSD makers tout quite a bit but we generally don't do a lot of IOPS testing because frankly a lot of users can't relate to IOPS metrics as well and it tends to be more meaningful to the enterprise/server crowd. Still, it is another performance indicator with relevance and while some drives post good MB/s numbers, their IOPS scores aren't always commensurate which this test will prove out. Benchmark Results: IOPS performance is very solid as well, outpacing the OCZ ARC drive with reads and writes being relatively equal which isn't always the case, especially in budget drives.
Real World Copy & Boot Tests
File Copy Times Via Teracopy 2.27:One of the most common operations performed on a PC is moving/copying files. Using a free application called Teracopy, we copied large numbers of two file types from one folder to another on the same drive. Teracopy allows us to objectively measure the time of transfer and using the same drive prevents other devices from tainting the outcome. The operation requires the drive to perform both sustained read and writes simultaneously. The first set of files is a 5GB collection of JPG's of variable size and compression levels with a few movie (.MOV) files thrown in for good measure since most cameras now take video as well as stills. The second is a collection of MP3 files of various sizes which totals 5GB collectively. These file types were chosen due to their wide use and mixture of file sizes and compression levels. Benchmark Results: The file copy times of the Ultra II didn't fare quite as well as the OCZ ARC drive which may be due to only having four NAND modules which restricts overall bandwidth as the NAND array is smaller.
Windows Boot Times Via BootRacer:Windows start up/shutdown time is always something people are interested in and we haven't done it in a while because there was little variation with the majority of the SSDs. We recently began using an application called BootRacer to objectively measure the start up times of the drives. All of the instances of Windows were identical and freshly installed with only the video driver installed. Benchmark Results: Boot times are the same as we saw for the ARC drive and since we've moved to the new test bench, much improved.
PCMark 8 Consistency TestSomething new we are starting to do is run the storage consistency test in the PCMark 8 application from Futuremark. In short, it looks at drive performance degradation as the drive becomes "dirty" and how it rebounds after given time to recover and run background routines like garbage collection and TRIM. This takes nearly a day to run through all of the iterations of test sets. Since most of our benchmarks show the drive performance in a clean state, this is a nice contrast to provide yet another view of performance. PCMark 8's storage benchmark test contains workload traces from Adobe Photoshop, Adobe Illustrator, Adobe InDesign, Adobe After Effects, Microsoft Word, Microsoft Excel, Microsoft Power Point, World of Warcraft and Battlefield 3. The PCMark 8 Consistency test has five phases: Precondition phase
- Write the drive sequentially through up to the reported capacity with random data.
- Write it through a second time (to take care of overprovisioning).
- Run writes of random size between 8*512 and 2048*512 bytes on random offsets for 10 minutes.
- Run performance test (one pass only). The result is stored in secondary results with name prefix degrade_result_X where X is a counter.
- Repeat 1 and 2 for 8 times and on each pass increase the duration of random writes by 5 minutes
- Run writes of random size between 8*512 and 2048*512 bytes on random offsets for final duration achieved in degradation phase.
- Run performance test (one pass only). The result is stored in secondary results with name prefix steady_result_X where X is a counter.
- Repeat 1 and 2 for 5 times.
- Idle for 5 minutes.
- Run performance test (one pass only). The result is stored in secondary result with name recovery_result_X where X is a counter.
- Repeat 1 and 2 for 5 times.
- Write the drive sequentially through up to the reported capacity with zero data.
Final Thoughts & ConclusionsLooking at the capacity of the SanDisk Ultra II 240GB drive, we saw 256GB (1GB byte = 1,000,000,000 bytes) on board in Windows reports the capacity accessible to the end user as 223GiB, 447GiB and 894Gib (1Gib = 1,073,741,824 bytes). This is similar to what we see on just about any consumer level drive these days. So at the end of the day, where do we stand here? Well, we end up with a very solid budget oriented drive that competes very well with its peers and priced very competitively as well. With the 120GB, 240GB, 480GB and 960GB drives retailing for $79.99, $109.99, $199.99 and $379.99 respectively, they are some of the most inexpensive drives you can get today with near enthusiast drive performance. The 960GB drive is priced under $0.40 per GB without and sale prices or rebates being involved. With the TLC NAND inside, power usage is pretty low. We were able to measure the idle draw at 0.502W and active to be 4.31W. Both are right about where SanDisk specifies. This power consumption is also pretty much what we saw on the TLC laden Samsung 840 drive. As such, those with mobile devices that run on batteries will find this choice appealing although there is no DevSleep support at this time (and maybe never) so it's something to keep in mind. At a high level, the things that separate enterprise level drives are endurance, performance and special features like encryption and power loss protection. With the TLC NAND, some may expect endurance to be an issue but with SanDisk's nCache 2.0 technology along with some other technical tricks, they've managed to really reduce write amplification so longevity should be closer to that of an MLC drive. They do offer a three year warranty which is about standard anymore. Performance as mentioned is very good at 550MB/s reads and 500MB/s writes. Not where we see enthusiast level drives but still not far off and really not enough to notice a difference for most users. However, it's very consistent and we saw no aberrations in our testing. Encryption is another feature absent but for many it's not even a thought in terms of a feature desired when shopping for a drive. The dashboard software is one of the best drive utilities we've seen to date and the bundled extras can be handy too although not free in the long term. On the few instances we spoke to SanDisk about this drive, they were very excited about it and understandably so. Other than Samsung, they are the only ones with a very solid consumer offering of a TLC based drive. It remains to be seen if others will follow and if long term they hold up as expected in comparison to MLC drives. SanDisk is really pushing hard to gain the market share they have in the memory card space and they are certainly on the right track for doing so.
Legit Bottom Line: The SanDisk Ultra II SSD has low pricing and power consumption with speedy and consistent performance so it earns our Legit Reviews recommended award!