The Drives

Corsair surprised us all at Computex this year when they announced their Neutron and Neutron GTX SSDs based on a so-far relatively unknown Link A Media Devices (LAMD) controller. While LAMD was new to the consumer-grade SSD controller market, the company had been making controllers for hard drives and enterprise SSDs for years, so the LM87800 found inside Neutron SSDs wasn't LAMD's first attempt in the storage world. The experience LAMD has gathered in the enterprise world is definitely visible in the LM8700 because even though its LAMD's first consumer-grade controller, it was not dwarfed by SandForce or Marvell.

Overall we were pleasantly surprised with the Neutron when we first reviewed it. There's always room for more competition because not only will it lead to lower prices, it will also provide the market with better products. Lately the consumer SSD market has been flooded with SandForce based SSDs, so it was very refreshing to see something new for a change.

Corsair only sampled the 240GB Neutron and Neutron GTX at the time of the release, which is a common practice among manufacturers. Stock is obviously limited and manufacturers want to get as many units as possible to retailers because ultimately that is what generates them revenue and profit. 240/256GB is usually the sweet-spot in terms of price and performance; smaller capacities perform worse due to less NAND die and higher capacities don't bring any improvements because the bottlenecks are elsewhere (*cough* SATA 6Gbps).

Now, three months later, Corsair finally sent us the rest of the available capacities. Testing different capacities is crucial, especially if the drive has a new controller, because controllers and firmware are designed differently and some design choices result in bigger differences between smaller and larger capacities. For example, the 128GB Vertex 4 is almost as fast as its 256GB and 512GB siblings, whereas the difference between 120GB and 240GB Vertex 3 ranges from negligible to as much as 50-75% (e.g. in write-heavy scenarios). Here's the quick rundown of the various drives.

	Neutron		Neutron GTX
Capacity	120GB	240GB	120GB	240GB	480GB
Raw NAND Capacity	128GiB	256GiB	128GiB	256GiB	512GiB
NAND	Micron 25nm synchronous MLC		Toshiba 24nm Toggle-Mode 2.0 NAND
Cache	128MB	128MB	128MB	128MB	256MB
Sequential Read	555MB/s	555MB/s	555MB/s	555MB/s	555MB/s
Sequential Write	211MB/s	370MB/s	330MB/s	511MB/s	511MB/s
4KB Random Read	50K IOPS	85K IOPS	80K IOPS	85K IOPS	90K IOPS
4KB Random Write	86K IOPS	87K IOPS	90K IOPS	90K IOPS	85K IOPS
Warranty	5 years

Toshiba's Toggle-Mode NAND is quite a bit faster than Micron's synchronous ONFi NAND as you can see in the table above. The 120GB Neutron is substantially slower than the 120GB Neutron GTX and there is also a noticeable difference at the capacity of 240GB. Corsair isn't offering a 480GB Neutron so if you're looking for a drive bigger than 240GB, you'll have to pay the premium for the Neutron GTX. Quite a few manufacturers, such as Intel and Plextor, have similar product schemes where only the high-end model offers the highest capacities (Intel SSD 330 tops out at 240GB and Plextor M5S at 256GB).

Comparison of NAND Interfaces
	ONFi				Toggle-Mode
Manufacturers	IMFT (Intel, Micron, Spectec), Hynix				Toshiba/SanDisk, Samsung
Version	1.0	2.0	2.x	3.0	1.0	2.0
Max Bandwidth	50MB/s	133MB/s	200MB/s	400MB/s	166MB/s	400MB/s

The Toggle-Mode interface can currently provide up to twice the bandwidth of ONFi. The interface alone isn't the reason for the difference in performance; Toshiba's NAND is also considerably faster when it comes to program times based on the data I've seen, which would explain why only write speeds are slower on the Neutron.

Price Comparison (12/18/2012)
	120/128GB	240/256GB	480/512GB
Corsair Neutron GTX	$130	$225	$455
Corsair Neutron	$120	$190	N/A
Corsair Force GS	N/A	$200	$410
Corsair Force GT	$130	$220	$440
Plextor M5 Pro	$125	$200	$460
Plextor M5S	$105	$200	N/A
Crucial m4	$110	$190	$383
Intel SSD 520	$129	$220	$445
Intel SSD 330	$113	$179	N/A
Samsung SSD 840 Pro	$139	$260	$577
Samsung SSD 840	$100	$169	$315
OCZ Vector	$150	$270	$555
OCZ Vertex 4	$108	$210	$458

Neither the Neutron or the Neutron GTX is the cheapest SSD available, but the pricing is still fair. Compared to the OCZ Vector or Samsung SSD 840 Pro, the Neutron GTX is actually quite a bit cheaper, although performance is at a different level as well. As always, prices fluctuate a lot so it's hard to draw any conclusions based on one day's pricing—it can be totally different in less than a week.

In fact, just to show you how prices fluctuate, here's the same table as above, only with prices from just over two weeks ago; I've italicized and bolded the differences, which you can see is about 80% of the SSDs we've listed. Sure, it's the Christmas season and so we're more prone than ever to price fluctuations, but in particular some of the newer drives (like the Samsung 840 500GB) as well as some of the older outgoing models (e.g. OCZ Vertex 4 and Intel 520) have seen some pretty heft price cuts. Be sure to shop around!

Price Comparison (12/1/2012)
	120/128GB	240/256GB	480/512GB
Corsair Neutron GTX	$140	$230	$470
Corsair Neutron	$120	$205	N/A
Corsair Force GS	N/A	$220	$450
Corsair Force GT	$130	$220	$440
Plextor M5 Pro	$130	$250	$460
Plextor M5S	$110	$200	N/A
Crucial m4	$110	$200	$385
Intel SSD 520	$130	$250	$490
Intel SSD 330	$115	$200	N/A
Samsung SSD 840 Pro	$194	$270	$600
Samsung SSD 840	$105	$179	$392
OCZ Vector	$160	$290	$570
OCZ Vertex 4	$140	$236	$475

 

Inside the Drives

	NAND Type	NAND Configuration	NAND Part Number
Neutron
120GB	Micron 25nm synchronous MLC	16x 8GiB packages	29F64G08CBAAB
240GB		16x 2x8GiB packages	29F128G08CFAAB
Neutron GTX
120GB	Toshiba 24nm Toggle-Mode MLC	8x 2x8GiB packages	TH58TEG7D2HBA4C
240GB		8x 4x8GiB packages	TH58TEG8D2HBA4C
480GB		8x 8x8GiB packages	TH58TEG9D2HBA4C

Neutron GTX 120GB and 480GB

In general, the only difference between the various capacities is the amount of cache and NAND per die. The 480GB model is the only one with 256MB (two 128MB chips) of cache; other models have 128MB. Both DRAM caches are DDR2-800 like in the 240GB Neutron GTX, but Corsair has opted for Micron's DRAM in the 480GB model whereas the 120GB and 240GB models use DRAM from Samsung.

LM87800—The heart of Corsair's Neutron series

Micron's ONFI NAND in 120GB Neutron

Test System

CPU	Intel Core i5-2500K running at 3.3GHz (Turbo and EIST enabled)
Motherboard	AsRock Z68 Pro3
Chipset	Intel Z68
Chipset Drivers	Intel 9.1.1.1015 + Intel RST 10.2
Memory	G.Skill RipjawsX DDR3-1600 2 x 4GB (9-9-9-24)
Video Card	XFX AMD Radeon HD 6850 XXX (800MHz core clock; 4.2GHz GDDR5 effective)
Video Drivers	AMD Catalyst 10.1
Desktop Resolution	1920 x 1080
OS	Windows 7 x64

 

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). I perform three concurrent IOs and run the test for 3 minutes. The results reported are in average MB/s over the entire time. We use both standard pseudo randomly generated data for each write as well as fully random data to show you both the maximum and minimum performance offered by SandForce based drives in these tests. The average performance of SF drives will likely be somewhere in between the two values for each drive you see in the graphs. For an understanding of why this matters, read our original SandForce article.

Random read speeds at low queue depths are usually not bound by the NAND because reading from NAND is a lot simpler process than writing, so you get no performance benefit from using faster NAND. There is no performance gain from more NAND die either as all capacities perform nearly equally. The gap between the fastest and slowest of the Neutron drives is 4%, which is hardly worth thinking about.

Random write performance is a lot more NAND dependent than random read. The 120GB Neutron simply can't keep up with the other models because it has slower NAND than the 120GB Neutron GTX and fewer die than the 240GB Neutron. The 120GB version of Neutron GTX does surprisingly well and is almost equivalent to its big brothers. While the gap between the higher performance Neutron drives is around 5% in the write tests, the 120GB Neutron is 20% (normal) and 40% (QD=32) behind the 120GB Neutron GTX.

Sequential Read/Write Speed

To measure sequential performance I ran 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.

Low queue depth sequential read speed is still an issue but given that Corsair has not provided a new firmware, this was expected. All the drives are running the same M206 firmware, which is what shipping units come with as well. Again, the gap between the fastest and slowest Neutron is around 6% here.

Sequential write speed shares the behavior of random write performance: having more and faster NAND increases throughput though there is essentially no benefit from more than 256GiB of NAND. Both 120GB drives fall behind, though the GTX is still close and performs around 60% faster than the Neutron.

AS-SSD Incompressible Sequential 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, while other drives continue to work at roughly the same speed as with compressible data.

Since there is no compression involved, incompressible sequential performance doesn't bring up any surprises. Write speed is heavily affected by the amount of NAND and its speed but this is typical to all SSDs.

Performance vs. Transfer Size

ATTO is handy tool for measuring sequential IO performance at varying transfer sizes. To keep the graphs somewhat readable, I didn't include all possible drives but for specific comparison you can always use our Bench tool. Sequential read performance at small transfer sizes is an area that many of the newer SSDs have been neglecting, the Neutron being one of them. Especially IO sizes of 4KB and 8KB are common in client workloads, so I'm hoping LAMD and Corsair can improve the performance at those transfer sizes in a future firmware update as the gap between the Neutron and other higher-end SSDs is fairly significant.

 

Click for full-size

Write performance isn't as bad, although there is definitely room for improvement at the smallest transfer sizes. Client workloads rarely see IO sizes below 4KB, meaning that there is little to no impact on real world performance unless we talk about some very specific workload. At the most important IO sizes, Neutron's write speed is on par with other drives.

Performance Consistency

If you read our Exploring the Relationship Between Spare Area and Performance Consistency in Modern SSDs article, you have seen the performance consistency data for Neutron already and the data here should be nothing new. However, because we didn't have any performance consistency results in our initial Neutron review, I wanted to focus on it here.

We started paying attention to performance consistency when Intel launched the DC S3700 because Intel specifically highlighted its low latencies, which translate to consistent performance. Consistency has been a big selling point in the enterprise market for a long time now, but that has not yet transferred to the consumer SSD market. Given LAMD's enterprise focused history, I was eager to find out how the LM87800 would do in this regard.

To generate the data below I took a freshly secure erased SSD and filled it with compressible sequential data (incompressible for Intel SSD 335) . This ensures that all user accessible LBAs have data associated with them. Next I kicked off a 4KB random write workload at a queue depth of 32 using incompressible data. I ran the test for just over half an hour, no where near what we run our steady state tests for but enough to give me a good look at drive behavior once all spare area fills up.

I recorded instantaneous IOPS every second for the duration of the test. I then plotted IOPS vs. time and generated the scatter plots below. Each set of graphs features the same scale. The first two sets use a log scale for easy comparison, while the last set of graphs uses a linear scale that tops out at 40K IOPS for better visualization of differences between drives.

The first set of graphs shows the performance data over the entire 2000 second test period. In these charts you'll notice an early period of very high performance followed by a sharp dropoff. What you're seeing in that case is the drive alllocating new blocks from its spare area, then eventually using up all free blocks and having to perform a read-modify-write for all subsequent writes (write amplification goes up, performance goes down).

The second set of graphs zooms in to the beginning of steady state operation for the drive (t=1400s). The third set also looks at the beginning of steady state operation but on a linear performance scale. Click the buttons below each graph to switch source data.

The Neutron is extremely consistent. After the first burst, the performance variation is very small compared to for example OCZ Vector and Samsung SSD 840 Pro, although it should be noted that the Neutron comes with 12% over-provisioning by default while Vector and 840 Pro have only 7% of the capacity reserved for OP. The amount of OP can have an enormous effect on consistency but if that extra consistency is only due to the extra OP Corsair and LAMD has set aside, then it's been a great decision. Out of the consumer-grade drives, only Intel SSD 335 is more consistent but SandForce has always performed well when it comes to steady-state performance.

When focusing on the last 10 minutes of the test where most SSDs have reached steady state, the Neutron is exceptional. While all other SSDs in this test including the enterprise focused Intel S3700 have noticeably variation in IOPS, the Neutron does not. The variation is only around 2-3K IOPS, whereas for Intel SSD 335 it's up to 25K IOPS. The steady state performance in general is also great; Vector and 840 Pro have an average IOPS of ~5-8K while the Neutron manages to push 12-13K IOPS. Intel DC S3700 is obviously in its own class but again, keep in mind that it also has significantly more OP than the other SSDs in this test. When the Neutron is given about as much OP as the S3700, they perform about the same.

Neutron's ability to retain performance and consistency over time is definitely appreciated and it's the most consistent consumer-grade SSD at the default OP we have tested. SandForce based SSDs can push much higher peak IOPS but the variation is in the magnitude of tens of thousands of IOPS. While the SSD 335 has higher IOPS on average, it drops below 10K right after 30 minutes of 4KB random writes but the Neutron is able to sustain a throughput of 12-13K throughout the last ten minutes of the test.

AnandTech Storage Bench 2011

Last year 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. Anand assembled the traces out of frustration with the majority of what we have today in terms of SSD benchmarks.

Although the AnandTech Storage Bench tests did a good job of characterizing SSD performance, they weren't stressful enough. All of the tests performed less than 10GB of reads/writes and typically involved only 4GB of writes specifically. That's not even enough exceed the spare area on most SSDs. Most canned SSD benchmarks don't even come close to writing a single gigabyte of data, but that doesn't mean that simply writing 4GB is acceptable.

Originally we kept the benchmarks short enough that they wouldn't be a burden to run (~30 minutes) but long enough that they were representative of what a power user might do with their system. Later, however, we created what we refer to as the Mother of All SSD Benchmarks (MOASB). Rather than only writing 4GB of data to the drive, this benchmark writes 106.32GB. This represents the load you'd put on a drive after nearly two weeks of constant usage. And it takes a long time to run.

1) The MOASB, officially called AnandTech Storage Bench 2011—Heavy Workload, mainly focuses on the times when your I/O activity is the highest. 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.

2) We tried to cover as many bases as possible with the software incorporated into this test. There's a lot of photo editing in Photoshop, HTML editing in Dreamweaver, web browsing, game playing/level loading (Starcraft II and WoW are both a part of the test), as well as general use stuff (application installing, virus scanning). We included a large amount of email downloading, document creation, and editing as well. To top it all off we even use Visual Studio 2008 to build Chromium during the test.

The test has 2,168,893 read operations and 1,783,447 write operations. The IO breakdown is as follows:

AnandTech Storage Bench 2011—Heavy Workload IO Breakdown
IO Size	% of Total
4KB	28%
16KB	10%
32KB	10%
64KB	4%

Only 42% of all operations are sequential; the rest ranges from pseudo to fully random (with most falling in the pseudo-random category). Average queue depth is 4.625 IOs, with 59% of operations taking place in an IO queue of 1.

Many of you have asked for a better way to really characterize performance. Simply looking at IOPS doesn't really say much. As a result we're going to be presenting Storage Bench 2011 data in a slightly different way. We'll have performance represented as Average MB/s, with higher numbers being better. At the same time we'll be reporting how long the SSD was busy while running this test. These disk busy graphs will show you exactly how much time was shaved off by using a faster drive vs. a slower one during the course of this test. Finally, we will also break out performance into reads, writes, and combined. The reason we do this is to help balance out the fact that this test is unusually write intensive, which can often hide the benefits of a drive with good read performance.

There's also a new light workload for 2011. This is a far more reasonable, typical every day use case benchmark. It has lots of web browsing, photo editing (but with a greater focus on photo consumption), video playback, as well as some application installs and gaming. This test isn't nearly as write intensive as the MOASB but it's still multiple times more write intensive than what we were running last year.

We don't believe that these two benchmarks alone are enough to characterize the performance of a drive, but hopefully along with the rest of our tests they will help provide a better idea. The testbed for Storage Bench 2011 has changed as well. We're now using a Sandy Bridge platform with full 6Gbps support for these tests.

AnandTech Storage Bench 2011—Heavy Workload

We'll start out by looking at average data rate throughout our new heavy workload test:

The 480GB Neutron GTX enjoys a small speed increase over the 240GB version but in general the overall performance follows the same pattern as our synthetic tests. The only truly significant difference in performance is between the 120GB models where the Neutron GTX beats the regular Neutron by 23%. Neutron GTX is actually a very high performer at 120GB although I have to note that we have not tested the 128GB version of Samsung SSD 840 Pro and OCZ Vector yet.

The next three charts just represent the same data, but in a different manner. Instead of looking at average data rate, we're looking at how long the disk was busy for during this entire test. Note that disk busy time excludes any and all idles, this is just how long the SSD was busy doing something:

AnandTech Storage Bench 2011—Light Workload

Our new light workload actually has more write operations than read operations. The split is as follows: 372,630 reads and 459,709 writes. The relatively close read/write ratio does better mimic a typical light workload (although even lighter workloads would be far more read centric).

The I/O breakdown is similar to the heavy workload at small IOs, however you'll notice that there are far fewer large IO transfers:

AnandTech Storage Bench 2011—Light Workload IO Breakdown
IO Size	% of Total
4KB	27%
16KB	8%
32KB	6%
64KB	5%

As our Light suite is more read centric, the benefit of more and faster NAND is smaller. All except the 120GB Neutron perform almost equally and even the 120GB Neutron is only ~16% behind whereas in our Heavy suite it was as much as 30% slower than the larger capacities.

Power Consumption

With more NAND, you get higher performance but there are also more dies that require power. The difference isn't substantial when compared to the performance increase, though, because for example the 480GB Neutron GTX draws only ~10% more power under load than the 120GB model. At idle there is essentially no difference between smaller and larger Neutron capacities. For some reason the 240GB Neutron stands out with its surprisingly high power constumption; it's noticeably higher even compared to the 480GB Neutron GTX. Power consumption in general is something I hope LAMD and Corsair can address thorugh a firmware update. Load power consumption is bearable but at idle I don't think any SATA SSD should draw more than 1W, preferably no more than 0.5W.

Final Words

The 480GB Neutron GTX didn't reveal any surprises as its performance is for the most part the same as what the 240GB model provides. It can't challenge Samsung's SSD 840 or OCZ's Vector but when compared with for instance Plextor M5 Pro and high-end SandForce SSDs, it's a competitive drive. Both Neutron drives are also among the more consistent performers in torture testing.

If we look at the pricing, both the Neutron and Neutron GTX are actually pretty compelling because the 256GB 840 Pro or Vector will cost you $35~$45 more and the difference can be over $100 as we move to higher capacities. The Neutron GTX is priced very similarly to Plextor's M5 Pro and they also share performance characteristics as well as 5-year warranty.

The 120GB Neutron GTX is a more interesting case. While it has half the NAND of the 240GB model, it can really hold up in the tests against the bigger capacities and currently it's the fastest 120/128GB SSD we have tested. I should note that we don't have 128GB samples of Samsung SSD 840 Pro or OCZ Vector yet, so that title may be only temporary but nevertheless its performance is impressive for such a small drive.

The same can't be said about the 120GB Neutron as its performance is more along the lines of other 120/128GB SSDs such as Crucial m4 and Corsair Force GT, but that was expected since it's aimed at the mainstream market and is also priced as such.

What the Neutron shows is that IO consistency does not have to be an enterprise-only feature—it's something that anyone can have if the manufacturer focuses on it. Only SandForce based SSDs manage to offer IO consistency anywhere near the Neutron and even then the Neutron takes the lead when ~30 minutes of 4KB random writes has been surpassed (though it's very unlikely for consumers to stress the SSD that much). Sure partially the reason for Neutron's IO consistency is its extra OP space compared to most of the other consumer SSD, but I don't believe anyone will leave out the Neutron because it offers slightly less space. We have always recommended keeping 20% or so of the SSD empty anyway and all that the extra OP really does is to make sure you can't fill the SSD enough to get into serious performance troubles. I think IO consistency is an area where manufacturers should focus more, even if it means assigning a bit more NAND for OP. Corsair and LAMD have shown that it's do-able and now others should take the hint and follow up.