Actual vs max performanceThat article is misleading. SDXC transfer bandwidth under SD 3.0 is rated at 832 Mbps
max. 2.4 Gbps (~300MB/s) is the stated
maximum under SD 4.0 spec. What you actually get out of your SD card is likely to be
much lower than that. considering that SD 2.0 defines transfer speeds up to 200× (30MB/s) yet we get much lower speeds than that with most SD cards, I doubt we’ll see that increased interface bandwidth put to much use. Likewise, that 2TB is
stated maximum capacity; it doesnt mean we’ll see 2TB SDXC cards at launch! (good luck on expecting 3D-fabbed or <10nm NAND flash in the consumer market within the next year) The 32GB limit of SDHC was an artificially imposed limit anyway, and one can find non-compliant SD cards with more storage than that; the sector addressing used in SDHC allows addressing up to 2TB of storage after all.
Multi-SD SATA devices in the marketUsing multiple SD cards as SSDs is
not a new idea. I make no comments on the prices of those devices, but keep in mind that a storage controller which can give you speeds commensurate with a proper SSD is going to cost you as well.
Capacity, longevity and performanceNAND prices aren't going to
plummet just because consumers expect them to, although going by industry forecasts we can expect a steady drop. 2TB of NAND flash is not going to be as cheap as you expect. Your average 32GB SDHC runs for about $1/GB, and we’re talking about a Class 6 device, not even anywhere near SSD performance if you throw 4 of them in an enclosure with a JBOD/RAID controller.
What kind of performance difference are we talking? Below are some benchmark charts tested over USB 2.0. (I was lazy to dig into my desktop for a SATA connection, and besides this gives you an idea of what each can achieve over the same interface. Note that in practical use, one would typically see no more than 35MB/s with removable storage over a USB 2.0 interface.)
SD Card (Sandisk 2GB ExtremeIII, pretty good by SD card standards)
SSD (OCZ 30GB Core V2 — Note that this is an old, shitty SSD without TRIM and does not represent typical performance of reviewed SSDs)
HDD (1TB Seagate Barracuda 7200.11)
The lower latency of the SSD is noteworthy. At the time of writing, a 32GB ExtremeIII SD Card costs ~$75, an Intel 40GB X25-V (which does much better than the Core V2 and has TRIM) goes for ~$90, and a 1TB 2.5" hard drive can be had for $100 or less (all US prices). The potential price of a stacked SD storage device does not seem very compelling for the form factor it is going to occupy or the performance it will display.
Add to this the fact that a good SSD comes with firmware that can
lower write amplification by optimising write operations and doing wear-levelling, thus increasing the longevity of the drive. Unless this SD-stacking device also has such firmware (i.e. most likely with a pricey controller), you might see those 5,000–10,000 write cycles quickly eaten away on one or more SD devices, instead of being evenly spread out.
Even if we assume similar write amplification on both devices, SD products typically use 3-bit-per-cell MLC technology, which though cheaper and denser (storage-wise) than 2-bit-per-cell MLC, also has
lower performance and
longevity. That is how we manage to cram 32GB into microSD form factor. Keep that in mind when you note the size of the NAND packages on SSDs. In any case, 3-bit-per-cell MLC is typically used for cheaper and smaller storage products such as USB flash drives, SD cards, CF cards and other similar flash storage devices, where one does not expect heavy, frequent or large writes. Cramming many such devices into a heavier-use storage device without wear-levelling and write-amplification-optimising algorithms is just asking for trouble.
What such a device might look likeMy prediction: By the time someone comes up with a “stacked micro-SDXC ultra-compact storage drive”, it’s going to look a lot like this:
… Guess what that is?
