The SSD market is currently going through the biggest change in its history. Ever since consumer SSDs started appearing around 2006, SATA has been the dominant interface but now that is about to change. With PCIe and NVMe making their way to the consumer space, SATA will sooner than later become obsolete, at least for high performance SSDs.

But right now, it is way too early to bury SATA. The transition to PCIe has taken its first baby steps but is far from taking over SATA. I believe 2015 will be the year of PCIe SSDs, because by then we will have several PCIe SSD platforms ready from SandForce, JMicron, and others; then the OEMs can start shipping drives in volume. However, we are still six months away from 2015 and even then, SATA is not going anywhere. The market and drives will continue to evolve and the latest proof comes from SanDisk with a humble name: the Extreme Pro.

The Extreme Pro takes the proven recipe of the Extreme II. The controller is unchanged but the firmware has been optimized even further for consistency and the NAND has been changed to SanDisk's own second generation 64Gbit 19nm MLC (same as Toshiba's A19nm). SanDisk's target markets for the Extreme Pro are gamers and professionals, which are user groups that traditionally crave performance and consistency.

The Extreme II had excellent IO consistency to begin with and was one of the most consistent client drives we have tested, so the bar is set high for the Extreme Pro. SanDisk is so confident about the reliability of the Extreme Pro that it is giving it a 10-year warranty, which is twice of what any consumer-grade SSD has offered so far.

SanDisk dumps the 120GB model from the Extreme Pro lineup but adds a 960GB one. This move makes sense because 120GB cannot provide the same level of performance due to the lack of NAND but also falling NAND prices have made 1TB-class SSDs affordable for consumers. I am a bit surprised that SanDisk has decided to use 64Gbit parts in all models because usually 1TB-class SSDs have required the use of bigger 128Gbit dies. I guess it helps that the page size is 16KB for the 64Gbit part too because the main problem with high capacity SSDs is that when you double the capacity, the amount of pages you need to track doubles as well. With both 64Gbit and 128Gbit parts having the same page size, the issue is pretty much neutralized as long as the controller and firmware can talk to that many NAND die (this is where having your own firmware team helps). 

Photography by Juha Kokkonen

The PCB in the Extreme Pro is single-sided in all models. There are eight NAND packages, which means that the 960GB drive must be using 16-die packages to achieve such high capacity. 16-die packages are actually nothing new because SanDisk and others have used them, and even 32-die packages, in memory cards for quite some time.

SanDisk's 16-die package used in 128GB iSSD

However, the yields are lower because it takes more precision to stack 16 dies and connect all the wires, which is why most manufacturers only use 8-die packages in SSDs. For memory cards one NAND package is enough because there is no space for more, but SSDs need several packages. Yields do get better over time though, and SanDisk's 16-die packaging is apparently economical enough to be used in SSDs now. There are some latency obstacles as well because taller stacks require longer wires, but it looks like the drawback is not enough to prevent their use in high performance SSDs.

The endurance remains unchanged from the Extreme II despite the doubled warranty. While 80TB should be enough for most users, I still would have liked to see higher endurance with the longer warranty. I think that the 10-year warranty is there just to differentiate the drive for marketing reasons, because if the Extreme Pro is used in (for example) professional video editing for ten years, the 80TB rating, or about 22GB a day, will not be sufficient. It appears SanDisk is betting that its target users will either replace the drive within a few years or the endurance limit will be hit after the warranty runs out, meaning that the 10-year warranty may not actually add any concrete value to the end-user. Of course it can still offer peace of mind for replacement drives, but I would give more value to the endurance rating when comparing high-end SSDs. 

The Extreme Pro features a more advanced version of SanDisk's nCache SLC caching technology, called nCache Pro. The difference between the two is that the original nCache was mainly for the firmware and NAND mapping table caching for data integrity reasons (it's faster to write to the SLC portion in case of a sudden power loss) but the nCache Pro is more optimized for user data caching. SanDisk would not reveal the exact size of the SLC buffer but SanDisk did tell us that it is less than 1GB, which is quite small compared to Samsung's TurboWrite with up to 12GB of SLC cache. That makes me think that the purpose of nCache Pro is to act more as a DRAM-like cache instead of a performance buffer, because ultimately you want to store as little data as possible in the DRAM due to its volatility.

SanDisk still does DRAM caching too, but writes smaller than 4KB will be cached to the SLC cache. Most writes from the host are 4KB in size but those are also tricky for the SSD because page sizes are much larger than that (currently 16KB), so caching to the faster and more durable SLC portion makes sense for write-combining (i.e. wait for more small IOs to come in and then write them as 16KB to the MLC part). Traditionally write-combining is done in the DRAM but it is much safer to do that in the SLC buffer, because the data will still be safe after a power loss whereas with DRAM it would be gone. (Of course, a super capacitor design could protect data in the event of a power failure.)

Sadly there is no encryption support. I asked why in our meeting at Computex and SanDisk told me that because TCG Opal 2.0 is still relatively new, SanDisk wanted to roll it out with the lower volume X300s first and see how the market reacts. SanDisk argued that it did not see the need for that in the Extreme Pro because they view desktop users as the main market, although I disagreed because the popularity of mobile workstations and laptop-as-desktop setups is constantly increasing, such that good IO performance should not only be limited to desktop users. In addition, I think desktop users, especially in the enthusiast and professional crowd, would appreciate encryption support because the data they handle can be sensitive (such as in home clouds). Hopefully we will see proper encryption support in the next generation drive (or even better: through a firmware update) because it is a quite significant drawback in the Extreme Pro.

Test System

Thanks to G.Skill for the RipjawsX 32GB DDR3 DRAM kit

Performance Consistency

Performance consistency tells us a lot about the architecture of these SSDs and how they handle internal defragmentation. The reason we don’t have consistent IO latency with SSD is because inevitably all controllers have to do some amount of defragmentation or garbage collection in order to continue operating at high speeds. When and how an SSD decides to run its defrag or cleanup routines directly impacts the user experience as inconsistent performance results in application slowdowns.

To test IO consistency, we fill a secure erased SSD with sequential data to ensure that all user accessible LBAs have data associated with them. Next we kick off a 4KB random write workload across all LBAs at a queue depth of 32 using incompressible data. The test is run for just over half an hour and we record instantaneous IOPS every second.

We are also testing drives with added over-provisioning by limiting the LBA range. This gives us a look into the drive’s behavior with varying levels of empty space, which is frankly a more realistic approach for client workloads.

Each of the three graphs has its own purpose. The first one is of the whole duration of the test in log scale. The second and third one zoom into the beginning of steady-state operation (t=1400s) but on different scales: the second one uses log scale for easy comparison whereas the third one uses linear scale for better visualization of differences between drives. Click the buttons below each graph to switch the source data.

For more detailed description of the test and why performance consistency matters, read our original Intel SSD DC S3700 article.

	SanDisk Extreme Pro	SanDisk Extreme II	Intel SSD 730	Intel SSD 530	OCZ Vector 150
Default
25% Spare Area

Similar to the Extreme II, the IO consistency is just awesome. SanDisk's firmware design is unique in the sense that instead of pushing high IOPS at the beginning, the performance drops close to 10K IOPS at first and then rises to over 50K and stays there for a period of time. The higher the capacity, the longer the high IOPS period: the 960GB Extreme Pro takes ~800 seconds before the IOPS drops to 10K (i.e. the drive reaches steady-state). I do not know why SanDisk's behavior is so different (maybe it has something to do with nCache?) but it definitely works well. Furthermore, SanDisk seems to be the only manufacturer that has really nailed IO consistency with a Marvell controller because Crucial/Micron and Plextor have had some difficulties and their performance is not even close to SanDisk.

However, I would not say that the Extreme Pro is unique. Both Intel SSD 730 and OCZ Vector 150 provide the same or even better performance at steady-state, and with added over-provisioning the difference is even more significant. That is not to say that the Extreme Pro is inconsistent, not at all, but for a pure 4KB random write workload there are drives that offer (slightly) better performance.

	SanDisk Extreme Pro	SanDisk Extreme II	Intel SSD 730	Intel SSD 530	OCZ Vector 150
Default
25% Spare Area

	SanDisk Extreme Pro	SanDisk Extreme II	Intel SSD 730	Intel SSD 530	OCZ Vector 150
Default
25% Spare Area

TRIM Validation

To test TRIM, I filled the drive with sequential data and proceeded with 60 minutes of 4KB random writes at queue depth of 32. I measured performance with HD Tach after issuing a single TRIM pass to the drive.

TRIM works for sure as the write speed is at steady 400MB/s.

.

AnandTech Storage Bench 2013

Our Storage Bench 2013 focuses on worst-case multitasking and IO consistency. Similar to our earlier Storage Benches, the test is still application trace based - we record all IO requests made to a test system and play them back on the drive we are testing and run statistical analysis on the drive's responses. There are 49.8 million IO operations in total with 1583.0GB of reads and 875.6GB of writes. I'm not including the full description of the test for better readability, so make sure to read our Storage Bench 2013 introduction for the full details.

We are reporting two primary metrics with the Destroyer: average data rate in MB/s and average service time in microseconds. The former gives you an idea of the throughput of the drive during the time that it was running the test workload. This can be a very good indication of overall performance. What average data rate doesn't do a good job of is taking into account response time of very bursty (read: high queue depth) IO. By reporting average service time we heavily weigh latency for queued IOs. You'll note that this is a metric we have been reporting in our enterprise benchmarks for a while now. With the client tests maturing, the time was right for a little convergence.

Given that the Extreme II was already dominating the Storage Bench 2013, it doesn't come as a surprise that the Extreme Pro is the new crownholder. Even the SSD 730 and Vector 150 can't challenge the Extreme Pro despite the fact that in terms of pure random write performance they are better. I think SanDisk's strength lies in mixed read/write performance because write performance alone does not yield good results in real world workloads, which tend to be a mix of reads and writes.

In fact, client workloads (like our Storage Benches) are usually more read-centric anyway and in the case of the Extreme Pro, the drive spent over three times longer processing read IOs than write IOs, which makes sense because there are nearly four times more read IOs than there are write IOs in the trace (even though in terms of gigabytes the difference is only twofold). 

AnandTech Storage Bench 2011

Back in 2011 (which seems like so long ago now!), we introduced our AnandTech Storage Bench, a suite of benchmarks that took traces of real OS/application usage and played them back in a repeatable manner. The MOASB, officially called AnandTech Storage Bench 2011 - Heavy Workload, mainly focuses on peak IO performance and basic garbage collection routines. There is a lot of downloading and application installing that happens during the course of this test. Our thinking was that it's during application installs, file copies, downloading and multitasking with all of this that you can really notice performance differences between drives. The full description of the Heavy test can be found here, while the Light workload details are here.

The Extreme Pro does well in our older Storage Benches as well, although as you can see we have gotten to a point where the difference between SATA drives is less than 10% in most cases. 

Random Read/Write Speed

The four corners of SSD performance are as follows: random read, random write, sequential read and sequential write speed. Random accesses are generally small in size, while sequential accesses tend to be larger and thus we have the four Iometer tests we use in all of our reviews.

Our first test writes 4KB in a completely random pattern over an 8GB space of the drive to simulate the sort of random access that you'd see on an OS drive (even this is more stressful than a normal desktop user would see). We perform three concurrent IOs and run the test for 3 minutes. The results reported are in average MB/s over the entire time.

The random read performance is great, which was expected when coming from the Extreme II. 

Random write performance, on the other hand, is only mediocre, although it's consistency that matters and not necessarily the peak performance. It seems that drives with high consistency do not have as good peak performance because the Extreme Pro, Vector 150 and SSD 730 are in the middle, whereas the 840 EVO and M550 lead the pack despite their relatively poor IO consistency. I would take consistency over ~10% better peak performance, though. 

Sequential Read/Write Speed

To measure sequential performance we run a 1 minute long 128KB sequential test over the entire span of the drive at a queue depth of 1. The results reported are in average MB/s over the entire test length.

Sequential performance is definitely good. I think the high read speed is one of the reasons why the Extreme Pro did so well in the 2013 Storage Bench because both the SSD 730 and Vector 150 are noticeably slower.

AS-SSD Incompressible Sequential Read/Write Performance

The AS-SSD sequential benchmark uses incompressible data for all of its transfers. The result is a pretty big reduction in sequential write speed on SandForce based controllers.

Performance vs. Transfer Size

ATTO is a useful tool for quickly benchmarking performance across various transfer sizes. You can get the complete data set in Bench. The read performance scales excellently with transfer sizes and beats all the other high-end SSDs. In write speeds the differences are much more marginal but the Extreme Pro still does well. 

Click for full size

Power Consumption

The Extreme Pro supports both DevSLP and slumber power. We are still working on a way to measure DevSLP power but in the mean time, we have to rely on slumber power measurements. Slumber power consumption is quite average -- it's not the most efficient drive we have tested but 100mW is pretty typical for a high performance SSD. What is awesome, however, is the load power consumption, because the drive doesn't exceed 3W in any case and stays at around 2W when under random write workload. Out of the high performance SSDs, the Extreme Pro is definitely the most power efficient because it supports DevSLP and slumber states, yet the load power consumption is very low too.

Final Words

The Extreme Pro is without a doubt the fastest SATA 6Gbps SSD in the market. As Anand said in the Extreme II review, SanDisk has picked an excellent balance of IO consistency and peak performance, resulting in high performance regardless of the workload. Too often I have seen manufacturers focusing on just one or the other, with the end result being an unbalanced drive. Intel's SSD 730 is a prime example of this -- its random write consistency is one of the best we have ever seen, but the lack of peak performance makes it a middle-class performer in our Storage Benches. In the end, real world performance is a mix of consistency and peak performance because the drive is not constantly hammered with high queue depth write activity (which is why peak performance matters) but on the other hand it must also be able to handle a constant IO load consistently. 

My only real complaint is SanDisk's decision to not include TCG Opal 2.0 support in the the Extreme Pro. Right now there are no high performance client SSDs with proper encryption support, and I think that's a niche SanDisk should have taken onboard. Currently you can buy the Crucial MX100/M550 or Samsung 840 EVO to get proper encryption support, but that comes at the cost of performance. In my opinion a user should not have to decide between performance and encryption because both are vital. There is hope that the X300s fills the gap by providing Extreme Pro class performance with full encryption support, but I am afraid that the limited availability and the higher price would make it out of reach for most consumers. We will see when we get one in for testing.

There is no way that the Extreme Pro can compete with Crucial MX100 and Samsung 840 EVO in price but it is a totally different animal with substantially higher performance. When the Extreme Pro is compared against other high-end SSD, namely Intel's SSD 730, OCZ's Vector 150 and Samsung's 840 Pro, the prices are quite competitive. NewEgg does not have any stock at the time of writing, but I would not be surprised to see the prices falling a bit once the drive becomes available later this month. 

All in all, the Extreme Pro is the only no compromise high-end SSD in the market (aside from the Extreme II, of course). Its performance is unmatched by any other SATA 6Gbps drive and it is the only truly high performance SSD with proper power management, making it perfect for mobile use as well. It's also the first high-end SSD in 1TB-class capacity, so there is no longer a need to choose between performance and capacity. As long as you can live without hardware encryption support, I am comfortable with saying that the Extreme Pro is currently the best SATA 6Gbps SSD in the market for users who seek the highest performance with consistency.

SanDisk is turning out to be a very dangerous player in the client SSD space. With nearly perfect vertical integration model (they just lack client controller silicon and DRAM), SanDisk has the ability to put against Samsung and Intel who have traditionally held the performance crown. Before the Extreme II, SanDisk was fairly unnoticeable in the retail SSD market, but the Extreme II acted as a warning of SanDisk's skill and knowhow, and the Extreme Pro just further reinforces that. If SanDisk can keep their pace going with PCIe SSDs, others will have hard time keeping up with them.