Corsair Neutron GTX And Neutron
With so many companies selling SSDs based on similar controller hardware, firmware, and NAND flash, there's more pressure than ever to differentiate. Much of the time, this comes in the form of price-slashing. That's a great way to get us excited about affordable solid-state storage, but decidedly not an effective money-making strategy.
An SSD veteran compared to some of the other companies in today's story, Corsair set out to change the one performance variable it could: the controller. Its new Neutron and Neutron GTX drives utilize a new controller from Link_A_Media called the LM 87800, and are designed to succeed the company's SandForce-based Force drives. We don’t know much about the chip, as much of its logic is still mired in secret sauce. However, we do know the Link_A_Media controller isn't sensitive to the type of data you throw at it, and that it's based on an eight-channel architecture.
In an effort to promote the best performance possible, Corsair arms its highest-end Neutron GTX drives with 24 nm Toggle-mode DDR NAND from Toshiba, while the Neutron hits a more aggressive price point through the use of 25 nm synchronous ONFi-compatible flash.
Both the Neutron and Neutron GTX come bundled with a mounting bracket. Corsair also supports the two families with five-year warranty coverage, which we really like to see.
Performance-wise, the Neutron drives are pretty impressive. At a queue depth of one, 4 KB random reads start out at ~25 MB/s (typical of most SSDs that pass through our lab). However, once you scale up to 32 outstanding I/O requests, throughput jumps to nearly 350 MB/s, roughly corresponding to 90 000 IOPS. In random writes, the Neutrons are slower than competing SandForce-based SSDs. However, they also don't take a big hit when you work with incompressible data, either.
In most cases, the Neutron and Neutron GTX fare very similarly. Our 128 KB sequential write test is the notable exception. Thanks to its faster Toggle-mode memory interface, the Neutron GTX takes a nearly 100 MB/s lead at queue depths higher than two.