[W1zzard]

Introduction

Corsair is one of the innovation leaders in the memory and power supply market.

At Computex this year, the company announced that they would introduce a new SSD controller from the relatively unknown company Link A Media, also called LAMD, to their lineup. Their LM87800 SATA 6 Gbps flash controller promises high-performance and very high write-performance, especially with incompressible data.

The LM87800 supports eight NAND channels with high-speed ONFi 2.3 and Toggle 1 flash interfaces using MLC, eMLC, and SLC flash chips.

Corsair has not only added the Neutron GTX to their lineup, but also a non-GTX version just called "Neutron", which offers slightly less performance at reduced pricing. Both drives are available in capacities of 120 GB and 240 GB.

Specifications: Corsair Neutron GTX 240 GB
Brand:	Corsair
Model:	Neutron GTX 240 GB
Controller:	Link A Media LM87800
Flash Type:	Toshiba MLC, 24 nm, Toggle Mode TH58TEG8D2HBA8C
Form Factor	2.5"
Capacity	240 GB (223.6 GB usable) 16 GB overprovisioning
Interface:	SATA 6 Gbps
Firmware:	M206
TRIM supported:	Yes
NCQ supported:	Yes
Warranty:	5 Years

Packaging

You will receive:

• The SSD
• Documentation
• 3.5" mounting kit
• 2x screw package

The Drive

The drive conforms to the dimensions set forth by the 2.5" form factor. It is made from metal which improves its durability. Please note that its thickness is down to 7 mm from the regular 9 mm thickness of a drive. Thinner drives are a requirement by Intel's Ultrabook standard, so the Corsair Neutron GTX is ready for these notebooks.


Like most recent SSDs, the Neutron GTX uses the SATA 6 Gbps interface. It is compatible with any other SATA standard, but will, in such a case, work at reduced performance.


On the PCB you will find the SSD controller in the center surrounded by eight flash chips. The smaller chip in the top-right corner is DRAM memory for the controller.


Corsair is using the new LAMD 87800 flash controller.


The eight flash chips are produced by Toshiba on a 24 nanometer process.

Test Setup

Test System
CPU:	Intel Core i5 2500K @ 3.3 GHz (Sandy Bridge, 6192 KB Cache)
Motherboard:	ASUS P8H67-M LX Intel H67
Memory:	2x 2048 MB Kingston HyperX PC3-12800U DDR3 @ 1333 MHz 9-9-9-24
Graphics:	Intel Integrated
Software:	Windows 7 64-bit Service Pack 1
Drivers:	SATA: Microsoft AHCI Display: Intel 15.21.13

• After the initial configuration and installation, we created a disk image that will be used to test every drive.
• Automated updates were disabled for all programs. This ensures that for every review, each drive gets the same settings without any possible interference from previous testing.
• Our disk image consumes around 25 GB on disk. We resized the partition to fill all available space on the drive.
• Partitions were checked for alignment.
• In order to minimize random variation, each of the performance tests is run ten times with reboots in-between tests to minimize the impact of disk cache.
• We used the average of ten runs as the final score.
• All application benchmarks run the actual application and do not replay any disk traces.

Windows 7 Startup Time

In this test we measure the time it takes Windows 7 to boot from kernel start-up to the execution of start-up programs. An internal kernel timer keeps track of the time, which is, for example, more accurate than a stop watch.

Office 2010 Installation

We installed Microsoft Office 2010 using the standard installation without any customization. The installation source files were located in uncompressed form on the tested drive (2.4 GB transferred, 72% write).

ISO File Copy

The 3.1 GB ISO image of Windows 7 64-bit (en_windows_7_ultimate_x64_dvd_x15-65922.iso) was copied to a different folder on the same drive. This represents a typical large-file usage model (6.15 GB transferred, 50% write).

WinRAR Compression

This test measures the time it takes WinRAR to uncompress the Linux 2.6.34 Kernel bz2 archive to the tested drive (865 MB transferred, 86% write).

Avast Antivirus

Avast Antivirus 5 was used to check the C:\Windows\System32 folder of our installation. We enabled "Full file scans" and "Scan of all files" regardless of extension (2.14 GB transferred, 0.6% write).

Photoshop CS5 Startup

We measured the time it took Photoshop CS5 to start the application, load a 21 MP photo, close the image, and exit the application (73 MB transferred, 1.4% write).

Photoshop CS5 Performance

Heavy Photoshop use can result in a large number of disk accesses when Photoshop is processing its scratch file. In this test, we measured the time it took Photoshop CS5 to open ten 21 MP images at the same time and, once done, to process each image, one by one. The operations performed on each image were crop, move, auto levels, resize to 1024x768, and save for the web (862 MB transferred, 79% write).

Crysis Level Loading

Crysis is well known for its long level-loading times. We disabled the rendering path of the engine to take the graphics card out of the equation and measured the time it took to load the level "Island" (188 MB transferred, 0.5% write).

Battlefield 2 Patch

Battlefield 2 uses a typical approach to game patching by integrating the new patch data into the existing game data - as opposed to just dumping a patch file into the game installation directory. This usage model results in a lot of disk activity and generally long patch-times. We applied the 500 MB BF2 patch 1.41 to our installation (18.3 GB transferred, 77% write).

Software Compilation

We compiled a large open source C# .NET project and measured the time it took to create the release build.

Windows Experience Index Disk

Windows Vista and Windows 7 have added a Windows Experience Index rating to all systems. It identifies possible weak points and gives users a quick idea of the performance they can expect. We ran the disk portion of the test that represents the "Primary Hard Disk" score listed (2.27 GB transferred, 81% write).

MySQL Enterprise Testing

We added a new test to our SSD testing suite, which is an enterprise oriented MySQL benchmark.

The benchmark's name is "TPC-C", and it's pretty much an industry standard for this kind of testing. You can find more info about it here. It simulates a company with multiple warehouses where several users put in orders, record payments, check order status, and monitor stock levels at the warehouse. The warehouses will also have to be refilled by moving inventory around. We use MySQL with the InnoDB storage engine and 32 warehouses. The concurrent user count is ramped up from 2 to 128 (indicated on the bottom of the graph). The vertical axis of the graph reflects the number of transactions per second; one transaction usually consists of several database queries.
We tested with a single drive, no RAID.

Performance Summary

We used the performance data from all our benchmarks to condense the results into a single relative performance score.

Performance per Dollar

In addition to the performance alone, we also present a performance-per-USD score that will be important if you want to maximize your investment. Please note that this score was normalized to exclude the capacity of the tested drive.

Price per GB

If you are only interested in getting maximum storage capacity for the least amount of money, this price-per-GB graph should be the one to look at.

Value and Conclusion

	The Corsair Neutron GTX 240 GB is available online for $249.
	Amazing sequential performance 7 mm thick, Ultrabook ready SATA 6 Gbps support Supports TRIM 5-year warranty
	Real-life performance lower than expected Relatively high price Needs a SATA 6 Gbps controller to perform best
9.2	Corsair's new Neutron GTX performs quite well in our testing. We see real-life performance that's almost on par with SandForce drives, and a few percent behind the Corsair Performance Pro, which is the fastest drive we ever tested. Our colleagues who use synthetic benchmarks see much higher performance from the Neutron GTX, especially with high queue depth. Unfortunately, our real-life testing does not reflect this, but I have to admit it is very consumer focused - for you! With this review, we introduced an enterprise-oriented database benchmark which simulates typical operations in a company with multiple warehouses, but here we don't see any revelations from the Neutron GTX either. Overall, the Neutron GTX is still a great SSD and I have to compliment Corsair for trying something new, beyond the usual SandForce and Marvell stuff that we see in most drives. I can also report that there were no bluescreen issues or similar problems with the drive, which will be good news for users who got burnt by SandForce. Corsair's five-year warranty also underlines their trust in the new product. Corsair was kind enough to send us a second drive for RAID-0 testing. Here, we see huge gains in sequential tests like ISO file copy, but more random access patterns don't benefit as much. Price-wise, Corsair is asking $250 for their Neutron GTX 240 GB, which does not seem unreasonable if you only consider synthetic benchmarks. With our real-life testing in mind, I am having a hard time justifying the extra cost though. If Corsair can get the price of their drive down to around $200, where most other SSDs sit in that category, then the drive will certainly take some market share away from the "big" names.