The big marketing lie on flash drive performance

The big marketing lie on flash drive performance

Summary: When it comes to false advertising I used to think the Wireless LAN industry were the worst offenders, but the flash storage industry seems to be vying for this dubious distinction.  With the launch of Windows Vista at the end of this month, flash media performance will become a big issue for ordinary consumers because of Vista's ReadyBoost feature which will boost the performance of Vista.


When it comes to false advertising I used to think the Wireless LAN industry were the worst offenders, but the flash storage industry seems to be vying for this dubious distinction.  With the launch of Windows Vista at the end of this month, flash media performance will become a big issue for ordinary consumers because of Vista's ReadyBoost feature which will boost the performance of Vista.  But consumers are in for a rude awakening when they find out they have a flash drive that is incompatible with Vista ReadyBoost because of poor performance.  Making matters worse are the ridiculous speed ratings that companies claim on their flash products which really don't translate to actual performance.

Take the following two compact flash cards for example from A-Data and PQI.

I bought these cards looking at the "120x" or "Hi-Speed 100" rating in the advertisement thinking that these must perform much better than normal flash cards that use to have the 40x and 80x on them.  I took them home and both of them failed the ReadyBoost performance test which means they can't be used for Vista ReadyBoost.

I am aware of the fact that the 120x rating may be referring to large file transfers used in digital photography applications, but even that theory collapsed when I performed my own throughput and I/O tests showing bottom of the barrel performance.  Rob Galbraith runs a photography website and he has one of the most complete large-file sequential transfer rate databases on the Internet.  When we look at his chart for CF (Compact Flash) cards, we see that the ratings have little to do with actual performance.  Even the card reader has a lot of impact on the final performance.  My two particular cards were not in these databases but it's safe to say that they would place near the bottom of the pack based on my testing with the card reader I'm using.  I even tried an IDE to CF adapter and it was still terribly slow though the results were much better than going through the USB card reader (note that IDE to CF adapters are not supported in Vista though I really wish they would be).

Windows Vista ReadyBoost requires a different kind of performance that requires high IOPS (Input Output Per Second) on small file transfers.  If large file sequential transfer can be thought of as drag racing, IOPS can be thought of as street racing in the city with lots of tight turns.  AnandTech has this USB flash drive roundup that's gives you an idea how much of a difference there is in flash drive perform.  Vista ReadyBoost requires 3.5 MB/s (megabytes per second) on 4 KB random reads and 2.5 MB/s on 512 KB random writes.  That translates to a blazing 896 IOPS which is about ten times faster than a typical hard drive.  Both of my compact flash cards scored around 2 MB/s for the 4 KB random read test which means they're not even close to being fast enough for Vista ReadyBoost.

From a consumer standpoint it's almost impossible to determine what kind of flash drive they need until they plug it in to their Vista computer and wait for Vista to report ReadyBoost capable or not.  This website has a user-submitted database of ReadyBoost compatible devices which may help users decide what kind of flash memory to buy.  According to AnandTech's roundup, the Lexar JumpDrive Lighting and Kingston DataTraveler Elite USB drives seem to lead the pack for both read and write performance so it's probably a safe bet to go with those models if you're looking for a ReadyBoost flash drive though they are typically twice as expensive than the cheaper flash cards.  USB memory sticks are the safest bet because you don't need to worry about the performance of the card interface.  But a memory card isn't the best solution from a form factor perspective because it sticks out of the computer whereas a flash card can be hidden inside the PCMCIA slot.

It would really help if flash drives had a ReadyBoost logo on it guaranteeing ReadyBoost capable performance.  Ideally the device would have a ReadyBoost multiplier rating with a 1.0x exactly meeting the 3.5/2.5 read/write MB/s specification and a 2.0x representing double the 3.5/2.5 specification.  This would be a great way for flash manufacturers to differentiate themselves if they would advertise these specifications.  Unfortunately Vista doesn't report the actual raw data for its ReadyBoost test since it only gives you a pass/fail score making it as useless and confusing as the Vista performance index that's based on a scale of 0 to 6.  While it's not Microsoft's job to benchmark the hardware market, it's certainly a missed opportunity to make the user's life easier.  [Update 3:30 PM - Ed Bott has a nice follow up to this blog and he reveals how to get Vista to tell you the flash performance details]

Topic: Software Development

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • What's the point of Readyboost?

    Why not just buy more RAM memory and stick it in? USB devices already show up ok in most OSs (as a drive admittedly) but why would I want my "main" memory to be external to the box?

    The performance of flash memory will never be as good as "proper" RAM chips and the idea of plugging slower memory into a resource hungry OS such as Windows seems counterintuitive.
    • Exactly!

      It seems a beggar'w way of installing RAM for each session for people too lazy or too cheap to go buy and install REAL RAM. You can get 2GB flash drives for like $30 at some places, so it's fairly inexpensive substitute RAM (if Vista will accept it) -- BUT it sure seems a little jerry-rigged to depend upon it for your computer.

      Why some people are touting it is because it appeals to the geeks and nerds... but most people won't want to bother with it.
      • I can answer that


        I already checked and that flash memory will work.

        A savings of atleast $120 is definitely worth it to me if I want to run the latest and greatest games.

        I think if the Compact Flash world gets their head out of their collective arses, this would be a great thing.
    • Because of bang for the buck

      1 GB RAM + 2 GB Flash is cheaper and better than just 2 GB RAM in general for system sluggishness and the former is cheaper. 2 GB RAM is better for other situations.
      • ReadyBoost retains its cache even when power ...

        ... is interrupted. This speeds the next boot. So readyBoost not only speeds disk access during normal operations by providing a larger cache it also accelerates reboots. ReadyBoost was intended for the new generation of hard disks that will contain the cache inside the drive. The ability to use it with external cahe is a stop gap until the industry gets there.
        • Not really.

          [b]ReadyBoost retains its cache even when power ...
          ... is interrupted. This speeds the next boot.[/b]

          Vista (at least RC1, assuming the Gold version works the same way) wipes the contents of the flash drive at boot time and creates a fresh cache file at boot time. So, even if you experience a power failure, Vista will wipe out the cache file - there's no boost in boot speed.

          If you really think it through, that IS pretty much the only way it CAN work. Why would anyone want to read the contents of a potentially corrupted memory cache file anyhow?
          • I know that is the way it works

            There is error checking built into the cache and Windows can determine its state.
          • Wrong!

            See here:


            "...I?m the PM who owns the feature and wanted to give a bit more info about the ReadyBoost feature.

            Although ReadyBoost offers several benefits, accelerating boot is not one of them. We rebuild the cache after every state transition. ...

            I?ve put together a small Q & A, here:"
    • If all you have is a 512MB PC ...

      ... and it won't take any more memory (as is the case with my six-year-old Dell Dimension 4100), ReadyBoost can dramatically improve performance.

      Windows Vista makes heavy use of disk-caching technology to wring out as much performance as it can from whatever RAM that you have. If you've ever paid much attention to the Windows Task Manager, you will see that the memory 'footprint' with Vista can be as low as about 370MB on a machine with 512MB of RAM to as high as 1024MB on a system with 2GB of RAM.

      But, whether you have 512MB RAM or 2GB RAM, though, the amount of physical memory AVAILABLE is often under 20MB. This is because the rest of that available RAM is used as a hard-drive cache in order to improve overall system performance. Once you exceed 2GB of RAM, the performance gains come from free RAM for program execution, not from more high-speed cache.

      While USB memory cannot be used like RAM for program execution, if it is fast enough, it can act as a high-speed cache for your hard drives. This frees of physical RAM for program execution without sacrificing overall system performance.

      To be sure, ReadyBoost is a stop-gap. It offers a low-cost way to improve Vista performance on legacy hardware with a 512MB limit on RAM capacity. Nevertheless, it is not a substitute for RAM and if your system can support 1GB of RAM or more, that is the preferred upgrade path.

      By the way, while my lame Dell Dimension 4100 take about five miutes to boot up Vista, once booted up, Vista runs fine -- even without ReadyBoost.
      M Wagner
      • That's a good one.

        There is no way Vista would ever run smoothly on your 6 year old system with 512MB RAM.
  • A simple explanation of flash and ReadyBoost.

    There are two types of flash technologies, MLC and SLC. In order for ReadyBoost to work the device must be made from SLC. The "X" rating on memory is defined as follows X= 150,000KB where KB = 1024 bytes. The device with the rating of 120 should have a speed equal to 120 times that. The deceptive practice comes down to many manufacturers designating the read speed only which is generally a lot faster then the write speed.
    • But they failed both read and write, by a mile

      It wasn't even close on read or write performance to those X ratings.
      • As always go with manufacturers you can trust.

        SanDisk, Lexar and Rocketfish (Canada only) high speed memory are honestly rated and I have found they work with ReadyBoost.
    • You sure about that typo there?

      "The "X" rating on memory is defined as follows X= 150,000KB where KB = 1024 bytes"

      First of all, X should be a rate as in KB/s. x = 150 KB/s I can believe but 150,000 KB/s would almost mean 150 megabytes/sec which is crazy speed.
      • Sorry!

        Of course you are correct. I meant to say 150KB/second and not 150,000KB. I must be getting old. Performance will vary based on the device and OS. They use specialized equipment, (IMI 7100) to run the tests and determine the speed.
        • Well they must be lying then about 120x

          There was no way it was 18 MB/s, it barely pulled in 3 MB/s even when using the CF to IDE adapter and the CF to IDE adapter performed much better than the USB card reader.
          • False advertising is subject to civil liabilities.

            If you believe the claim is false you should take it to the attorney general of your state. We have a whole department that verifies such claims before we go to market for that very reason.
          • Are you sure the internal card reader is using ...

            ... the IDE buss. Most use the USB buss. The IDE buss has almost gone away. At best with most new mothernoards their is only one chip allowing you to have one master and one slave device. Once there is price parity for Sata optical drives that one too will go the way of the dodo.
          • It's a 40 pin IDE to CF adapter

            It's a 40 pin IDE to CF adapter I tried in addition to a USB 2.0 card reader. The IDE adapter had nearly double the IO performance for small reads compared to the USB 2.0 card reader. I'm going to go ahead and get one of those high-performance card readers in Rob's database and try it again with one of those.
          • Lite-on already has a 16x SATA DVD burner for $35

            Lite-on already has a 16x SATA DVD burner for $35. Once the 40 pin adapter disapears, I'm sure they will start selling SATA to CF adapters.