The Corsair Force LE200 240GB earns the coveted title of "The Best of the Worst," at least for now. DRAMless SSDs serve as hard disk drive replacements, and more are planned in the future, so competition will intensify.
Over time, several key technologies will increase DRAMless SSD performance. NVMe is one of the most obvious, and 3D NAND will also weigh in to improve both performance and endurance. We've already tested the Phison PS3111-S11 with both Micron and Toshiba 3D, and we are testing a Maxiotek DRAMless controller with 3D NAND.
DRAMless NVMe SSDs that use your system memory to cache the table map (Host Memory Buffer - HMB) will debut by the end of the year. The Windows 10 Anniversary update added support for the new technology, and products will follow in short order. HMB technology suffers a little more latency than direct-attached DRAM designs, but the cost savings, as well as access to a larger buffer, are beneficial. Both Toshiba and Marvell have demonstrated working HMB products. At Flash Memory Summit in 2016, Toshiba displayed an HMB SSD on a modified Asus motherboard, and at CES, Marvell displayed a prototype with an off-the-shelf motherboard.
I don't think our readers will consider the category a viable option for personal use until DRAMless NVMe SSDs with HMB technology and 3D NAND make it to market. Intel's Optane cache technology also provides increased system responsiveness, which brings another technique with similar low-cost attributes into the picture.
That leads us to another set of problems with the current DRAMless designs. The size of the flash translation table increases in tandem with the amount of data stored on the drive. We haven't heard of the controller companies increasing the SRAM capacity on the controller to accommodate higher capacity SSDs. We've already tested several 512GB DRAMless products and found that performance decreases as capacity increases. A 1TB or larger DRAMless SSD filled to 75% capacity could be slower than an HDD in some applications. That wouldn't be acceptable, and the user experience would suffer even if it were behind a high-speed next generation cache.
As far as our readers are concerned, I think the current crop of DRAMless SSDs, like the Corsair Force LE200 we tested, are the type of product you install in someone else's computer. Users that would normally install a hard disk drive are the primary target market. The Force LE200 delivers better performance than a hard drive. Many people are happy with a 1TB HDD and see the capacity as excessive because they won't fill the drive. An upgrade to a Force LE200 will increase the user experience by decreasing latency, which leads to a more responsive system.
The biggest problem we see for the Corsair Force LE200 is the price. The MSRP for the 240GB model we tested today is $99.99. You can already find the drives for $89.99 (a $10 savings), but I don't think it's enough. The Hyundai Sapphire we recently tested retails for $55 at Newegg. The Force LE200 is faster, but this isn't a performance race--it's a race to the best value. Price is high on the list of concerns when your real competitor is a hard drive that costs as low as $45.
In the past, we've seen Corsair release products at a high price point, and then lower it to an acceptable level within two months. I have no doubt we'll see the Force LE200 priced in the mid $70's before Computex in June. There are already several DRAMless SSDs from Asia on Newegg and Amazon at that price. Corsair can charge a little more for brand recognition, but in the race to the bottom, name recognition doesn't mean much.
MORE: Best SSDs
MORE: Latest Storage News
MORE: Storage in the Forums
the 70$ price by june might make it more reasonable?
Does the DRAM savings really make that much of a difference ?
How are these DRAMless devices getting so much throughput with a single channel while the high end controllers need eight channels ? Is the scaling on additional channels efficient ?
All NAND fabs combined can't make enough to satisfy the demand right now. They are all running at maximum capacity. Increasing yields (as mentioned in the article) will be the fastest way to increase supply. They can build more factories, but it takes a couple years and billions of $. They can switch to a better process technology, but that can take even longer and even more $.
When I see people still in denial about the Nand shortage, and the HDD factory flooding. I just have to wonder why are they so uninformed to the point I think they must be joking and are just baiting people.