Today Samsung revealed what they claim are the first flash memory chips based on the eMMC 5.1 standard which was just approved by JEDEC, the group that oversees the development of many semiconductor industry standards. eMMC is an embedded version of the MultiMediaCard standard that was once popular among digital cameras and PDAs. It allows for the NAND and controller in a device to be put into a small package on the logic board, which is a necessity when working with small mobile devices. Essentially every smartphone and tablet on the market today uses the eMMC standard as its storage solution.

As mobile devices have become more complex. there has been a need for improved NAND performance. Streaming and recording higher resolution content requires faster speeds for reading and writing, and so the evolution of the eMMC standard has been accompanied by improvements in performance at each step. When Samsung introduced their first eMMC 5.0 flash memory in 2013, their 64GB chips were rated for maximum sequential reads and writes of 250MB/s and 90MB/s respectively. Samsung's eMMC 5.1 memory increases the maximum write performance to 125MB/s. Similarly, random read performance increases from 7,000 IOPS to 11,000 IOPS and random write performance increases from 7,000 IOPS to 13,000 IOPS.

eMMC 5.1 based memory also has new features that will enable further improved performance. The feature that Samsung is specifically making note of is command queuing. True to its name, command queuing provides an interface for queuing up commands to be executed. This has been a feature of storage solutions in the PC world for some time, but previous eMMC controller implementations would submit commands and wait for completion before sending the next command.

Source: Samsung Semiconductor

Comments Locked

42 Comments

View All Comments

  • Cellar Door - Tuesday, February 17, 2015 - link

    A nice evolutionary step. Basically trying to eliminate bottlenecks in overall operation.

    Still, with the mobile market moving forward so quickly, this seems more of a step to pump out small changes to fuel sales.
  • eanazag - Tuesday, February 17, 2015 - link

    It's both. This will allow Samsung to differentiate in the Android space. They can claim to have the fastest storage system in mobile.
  • WorldWithoutMadness - Tuesday, February 17, 2015 - link

    LOL, fastest storage system in mobile doesn't really translate with UI snappiness.
    They'd better redo their UI from scratch first.
    But oh well, it seems other oem can benefit from these.
  • Stochastic - Tuesday, February 17, 2015 - link

    What are the biggest bottlenecks affecting subjective snappiness/responsiveness of smartphones today? Are we mostly SoC limited or is storage a big bottleneck as it is with traditional PCs?
  • willis936 - Tuesday, February 17, 2015 - link

    Even moreso in mobile. Look at those IOPS numbers. Those only go one direction over time and getting grabage collection smoothed out all the way down the stack has been taking a while just like it did with AHCI except here there's no high performance controller pulling half a watt just to make the storage faster. When an app takes an extra second to load or things feel chuggy when swiping around it's the storage that's doing that, not the soc.
  • SleepyFE - Wednesday, February 18, 2015 - link

    Well that's not true. I have a cortex-A5 with no GPU and my father has a cortex-A9 and PowerVR sgx531. Mine is chuggy while his is not. Since eMMC 5.1 just came out he isn't benefiting from it so storage not such a big problem as you make it out to be.
  • jjj - Tuesday, February 17, 2015 - link

    Those are not small changes , random IOPS gets a major boost (57% and 86%) and command queuing can be a big enough deal if it's smart enough.
  • Lolimaster - Tuesday, February 17, 2015 - link

    Dunno why they still keep the random numbers measured in iops rather than MB/s. I know why, performance is so low that this iops in the "thousands" confuse the average customer.

    For 11-13K IOPS its basically 10MB/s
  • pronuncer - Tuesday, February 17, 2015 - link

    IOPS is IOs per second. It is not the same thing as data rate, which is measured in MB/s because each IO can have different bytes written/read.
  • mkozakewich - Wednesday, February 18, 2015 - link

    13,000 4KB IOPS would be a little more than 50 MB/s, which is actually really good. I have no idea if that's the usual transfer sizes on these, though.

Log in

Don't have an account? Sign up now