840 EVO Specifications
At this time, the 840 EVO is a unique product thanks to the software layer. We suspect over time more companies will use DRAM to cache the storage system reads and writes, as well as take in all of the random reads and then output the data as sequential writes to the drive. This isn't anything new, we've seen the technology used on the enterprise side for some time now. It does make it difficult to show a list of specifications like the one below since your existing components play a role in determining perceived performance and benchmark results.
At the heart of the 840 EVO is a new controller, MEX. This is another tri-core controller like what Samsung used on the 840 and 840 Pro. We like to say one core is dedicated to readings, one to writing and one to arithmetic since each core has a separate function, the last, arithmetic part, refers to background activities, such as garbage collection. The MEX is 100MHz faster than the controller found in the 840.
Along with the new controller is support for AES-256 bit encryption. The 840 EVO supports eDRIVE, part of Windows 8 BitLocker and TCG Opal.
Samsung used new 19nm TLC NAND flash. This is second generation TLC for Samsung, the density has increased per die, while the lithography has shrunk. With those come a few concerns, mainly lower interleaving between die, a performance constraint on smaller capacity size SSDs.
Even though the Samsung 840 EVO has less interleaving between flash, compared to 840 products of the same capacity size, Samsung managed to increase performance. A pseudo-layer of SLC on each die, similar to SanDisk's nCACHE, increases write performance significantly. We'll talk more about the different cache layers on the next page.
There are five capacity sizes for launch day, 120GB, 250GB, 500GB, 750GB and finally the massive 1TB model. The two smallest capacity sizes use just two NAND packages, while the two largest capacity sizes use eight. The 500GB model is the one we are not sure about, but over time we'll have one in for review and then we'll know.
A few months back we started publishing 4K read and write IOPS performance at different queue depths. Many of our readers started falling down the rabbit hole of high queue depth IOPS performance. SSDs manufacturers were routinely publishing their high queue depth IOPS performance on their consumer SSDs, even though consumers rarely reach even half of the available queue depth levels possible by the SATA interface. I left the text above the chart so some of our new readers could better understand that low queue depth IOPS in a consumer or client environment (how you use your computer) plays a larger role in the SSD experience, than high queue depth IOPS performance. We are glad to see Samsung including this data and stepping away from the 100K IOPS at QD32 talk.
We never inquired about the techniques Samsung used to further reduce power consumption, but we suspect the new flash and refinements to the tri-core controller are responsible. The data provided represents the SSD alone and not the overall system power. In our testing with RAPID enabled, we found that battery life reduced due to DRAM using more power than NAND. The 840 EVO on its own though does use less power than the previous generation 840 and flagship 840 Pro.