ADATA has announced its first SSDs in a BGA form-factor. The IUSP33F drives are designed for devices that require a storage solution with a minimal footprint, such as thin-and-light laptops, convertible notebooks, or tablets. Meanwhile, the performance of ADATA’s BGA SSDs is higher than that of mainstream SATA SSDs.

ADATA’s IUSP33F drives offer capacities of 128 GB and 256 GB and come in an industry standard Type 1113-X (11.5×13 mm) package, which is further attached to the host via a PCIe 3.0 x2 interface. The SSDs are based on Silicon Motion’s SM2263XT DRAM-less controller with HMB support and are paired with Micron's 3D NAND memory. The solution fully supports the NVMe 1.3 protocol, SMI’s NANDXtend LDPC ECC engine, an end-to-end data path protection as well as embedded programmable RAID to further improve reliability and durability of SSDs.

Looking at performance specifications, we see up to 1195 MB/s sequential read speeds as well as up to 940 MB/s sequential write speeds, which is in line with inexpensive PCIe 3.0 x2 SSDs in M.2 form-factors. As for random performance, the IUSP33F can hit up to 140,000 read IOPS and up to 114,000 write IOPS.

ADATA did not disclose when it plans to make its IUSP33F BGA drives available in volume, but since such products are aimed at PC OEMs, it is likely that either the manufacturer has already delivered the first batch, or is actively searching for interested customers. As for pricing, it will depend on volumes and other factors.

ADATA IUSP33F Specifications
Capacity 128 GB 256 GB
Model Number ? ?
Controller Silicon Motion SM2263XT Controller
NAND Flash Micron B16A 64-layer 3D TLC NAND
Form-Factor, Interface BGA Type-1113-X, PCIe 3.0 x2
Sequential Read up to 1195 MB/s
Sequential Write up to 940 MB/s
Random Read IOPS up to 140K IOPS
Random Write IOPS up to 114K IOPS
Pseudo-SLC Caching Supported
DRAM Buffer No
Encryption The controller supports TCG OPAL 2.0, but it is unknown whether the technology is enabled by ADATA
Power Management Active: ?
Slumber: ?
Warranty Depends on terms
MTBF millions of hours

Related Reading:

Source: ADATA

Comments Locked


View All Comments

  • deil - Thursday, September 13, 2018 - link

    Can we get this in any phone ? please ?
  • Dragonstongue - Thursday, September 13, 2018 - link

    I personally would rather then not waste space on this type of memory/storage which likely takes up a good chunk of available space and instead be far far more concerned about making them tougher, lower cost and much larger battery then the "average" which seems to currently be in the 2500-3400 when they should be in the 3500-5200 these days ^.^

    IMO (for me) 64gb available storage is AMPLE for a damn phone..but then again, I do not use my phone for computer like stuff, it is just a phone that I can do some invoice stuff, take notes, listen to music, maybe occasional pictures with...why do they not "segrate" their line a bit more, for people like me that want a bit of everything (much larger battery) ones that could care less about "simple" phone things prioritize for great pictures and increased storage speed/size and battery life etc ^.^
  • rpg1966 - Friday, September 14, 2018 - link

    Cool story, bro.

    Anyway, what do you mean, "waste space"? Do you think existing storage has zero dimensions?
  • PeachNCream - Friday, September 14, 2018 - link

    Greater market segmentation would benefit consumers, assuming they spend the time to match their needs to an suitable phone. OEMs would have to soak up the cost of developing, manufacturing, and shipping additional SKUs. That would harm efficiency and hurt shareholder value for a potentially small benefit. As a stakeholder, I want the companies I invest in to put a priority on adding investment value and paying out dividends. Social responsibility is nice, but is a secondary concern.

    Phones are good enough with a microSD slot. If you only require 64GB, then you ought to be served well by internal memory only. For those of us that need more, we can easily add up to 512GB by purchasing a suitable storage device. However, I do like the idea of additional internal storage for ever growing programs that can't make use of a removable memory card. A single BGA chip and the traces needed to connect it to the SoC would not really speak for very much battery capacity anyhow. What you'd benefit from is a thicker handset that offers more internal volume for a larger battery and adding thickness isn't appealing to a lot of customers right now.
  • surt - Saturday, September 15, 2018 - link

    Marketing folks at giants like Samsung are not completely incompetent. They don't make a phone like that because there aren't enough buyers to pay for the r&d&m.
  • altair_auditore - Thursday, September 13, 2018 - link

    How are these DRAM-less SSDs?
    ADATA sells the 250GB SU650 with a 140TBW rating (just below the 150TBW of the 860 EVO which has DRAM).
    Do the drives chew through the TBW a lot quicker due to the lack of RAM, and how's the performance?
  • Death666Angel - Thursday, September 13, 2018 - link

    The HP EX900 has the DRAMless Silicon Motion SM2263XT NVME controller, there are a few tests of it around the web. It uses the Host Memory Buffer (HMB) feature to utilize the system memory instead of onboard RAM like other SSDs. Considering that the RAM on SSDs does not hold user data, just overview tables, I don't see how it should eat into P/E cycles. Performance is low compared to other NVME devices and high compared to normal SATA based ones. I wouldn't buy one at this moment, because SATA is still much cheaper (140€ for the HP NVME DRAMless one and 90€ for the SATA WD Blue 3D). And if I go for NVME, a 960 EVO is the same price (500GB in both comparisons, Germany).
  • abufrejoval - Monday, September 17, 2018 - link

    FusionIO was the highest-performance Flash Memory market leader a couple of years ago, pioneering PCIe connectivity and a "native API" to avoid all the overhead of a block storage abstraction and an OS that mentally was a DOS (disk based operating system): Their designs put all the logic and buffering into OS drivers, for much better speed and also because SSD controller chips at the time were non-existant or very primitive. The FPGA controller on the card really just supported the "analog" art of trapping and counting selected numbers of electrons in cells that aged and changed over time, temperature and with usage and translate that in to a binary or ternary digital illusion.

    So in terms of performance these "primitive" SSDs are actually hard to beat, as long as your CPU is faster and has RAM left over for the mapping tables. And with mobile devices, the UPS is built right in and DRAM nominally operated non-volatile with it. Of course a bug in any other kernel driver could rain right into your precious mapping tables, so it's probably a bad idea to rely on DRAM based write-back caches, even if they'd be so much faster than an SLC mode cache on the device.

    I guess eMMC offers SATA abstractions these days, but early mobile flash was direct mapped, I believe, and it was the SoC responsibility to do manage endurance with a translation layer, just like here.

Log in

Don't have an account? Sign up now