Sign in with
Sign up | Sign in

The SSD Power Consumption Hoax

The SSD Power Consumption Hoax
By

Flash-based solid state drives (SSDs) are considered to be the future of performance hard drives, and everyone seems to be jumping on the bandwagon. We are no exception, as we have been publishing many articles on flash-based SSDs during the last few months, emphasizing the performance gains and the potential power savings brought by flash memory. And there is nothing wrong with this, since SLC flash SSDs easily outperform conventional hard drives today (SLC = single level cell). However, we have discovered that the power savings aren’t there: in fact, battery runtimes actually decrease if you use a flash SSD.

Could Tom’s Hardware be Wrong?

No, our results are definitely correct. We’ve looked at almost a dozen different flash SSDs from seven vendors over the last few months, and measured acceptable or sometimes even disappointing power requirements with most flash SSDs. In an effort to determine the actual impact on notebook systems, we took four SSDs that we had available in our test lab, and ran a series of Mobilemark benchmark runs on a Dell Latitude D630 notebook. We found runtime differences of up to one hour (!) when using a flash SSD compared to a high-performance 7,200 RPM 2.5” notebook hard drive.

Will this Slow Down the Flash Memory Hype?

We don’t think so. Flash is cool, because it’s expensive, and because it does provide significant performance advantages (when SLC flash is used). But while our results are interesting for most users, they will be shocking to road warriors. You will see very clearly in the benchmark section that theory and real life are sometimes entirely different.

We still believe that flash-based drives will be the future for the performance segment, but they must not become a key component for energy-efficient notebooks and ultra-portables — where performance is secondary — as long as their average power consumption is higher than that of conventional 2.5” notebook hard drives. In fact, even a high-performance 7,200 RPM 2.5” drive provided better overall battery runtime than most of the flash SSDs we put through the Mobilemark test.

Ask a Category Expert

Create a new thread in the UK Article comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 10 comments.
This thread is closed for comments
  • 0 Hide
    Solitaire , 27 June 2008 22:22
    Nice article, gave me some lunchtime chuckles. Sure flash consumes less power at peak, but peak=!load... at least for a normal HDD. It sure is for flash though!

    My only complaint would be that it would have been much better to have also tested some comparable MLC units and compared those to a 5400rpm drive. Not all people who use laptops heavily on the go are stinking rich, so it would have been interesting to compare the more mainstream solutions in a similar face-off.
  • 1 Hide
    Anonymous , 1 July 2008 20:54
    Page 13 results and corresponding claims regarding flash sucking more power = total bullshit.

    Numbers were calculated badly and ain't work a broken penny.
    Take a better look at page 12 and you will see that Memoright flash based PC gets >2x the amount of work done befor going out of juice comparing to hdd based laptop.
    What more, it does it in shorter time. It goes out of juice quicker, because cpu and other laptop components are working hard because there are no stalls from hdd access time.

    Want to get apples to apples comparision? Try a benchmark where there's X to do and after that laptop goes into idle state for the rest of test. Flash based devices will win this one easily (and then again, they will finish the bloody task much faster and that's what most of us are looking for)
  • 0 Hide
    Anonymous , 1 July 2008 20:55
    My mistake above, should have read:

    Numbers were calculated badly and ain't worth a broken penny.
  • 0 Hide
    Anonymous , 1 July 2008 21:44
    It would have been nice if the idle battery life was also tested. Then people who use laptop for text editing etc light work could have got a better picture of the difference between battery life.

    Now I can't tell if it is better to get SSD or HDD laptop for school, where most of the time, the computer is idle as I listen to the teachers.
  • 1 Hide
    Anonymous , 1 July 2008 22:51
    I think they should MEASURE the power comsumption graph of the SSD/HDD alone.
    What really matters I think is the power comsumtion per let's say a 100MB of data written and read. SSD consume more power during writing operations to the flash cells as compared to the reading operations. HDDs consume a lot of power with the accelerating of the heads during random reads/writes.. (emE)
  • 1 Hide
    mastrom , 2 July 2008 02:51
    To Patrick:
    First of all it's totally unacceptable to have this title and making such accusations about a "Hoax" when all but one of your graphs talk about performance and only one measures the "battery life". If you want to prove a "hoax" you need to perform all the battery life tests you can and then more.
    Now to the real important stuff. Since I didn't know anything about your benchmark program I couldn't be sure how it works and I was unable to prove you wrong about your conclusion.
    Going to http://www.bapco.com/techdocs.html I found the white paper for Mobilemark 2007. Under paragraph 2.5.2 Battery life rating methodology I read: "The benchmark generates battery life ratings as its principal metric. The battery life rating in MobileMark 2007 is measured in minutes. This metric reflects the number of minutes the system can remain operational while executing a chosen module. Each module will produce a different battery life rating, reflecting differences in system loading."
    Everything ok until now. Each module should accommodate different usage patterns. And it continues: "The battery life is established by recording the start time of the benchmark, then repeatedly performing the workload. When the remaining battery capacity has fallen to 7% the benchmark records a timestamp once per minute. Once the battery has been depleted and the computer plugged in and rebooted, the benchmark compares the “start” timestamp and last recorded (“end”) timestamp. The battery life rating is the number of minutes between these timestamps."
    Did you notice the important detail? "REPEATEDLY performing the workload". Well my friend, this is the reason your conclusion is completely off base. It is proved and accepted the SSD drives are faster (on average) than conventional Hard Disk Drives. Even your graphs prove this fact. However when you test the battery life by repeating the same workload again and again you force the system to perform more cycles of the same workload when the SSD is used compared to the HDD because most tasks in the workload wait for the hard disk to finish a task to move to the next. So unless you tell us how many times the workload was repeated by each configuration you can't compare the battery life times...
    That’s why the DVD playback test is so popular in battery life tests. It makes sure the computer will perform the same work per x amount of time. So please explain to me why you didn't publish the results of the "DVD2007: Battery Life" and the "Reader 2007: Battery Life" modules which are part of the MobileMark2007 as I see at the paragraph "3.0 MobileMark 2007 Scoring Methodology" in the white paper.
    I suggest you change your title and publish more tests on the subject. Not only your graphs were wrong, your logic is flawed too.
  • 2 Hide
    Anonymous , 2 July 2008 04:56
    I was suspecting exactly about what mastrom said, even before I got to the comments.
    The only way to explain those results would be what he said.

    The principle is: even if a SSD uses more power during "active" operation, it does so for a fraction of the time required by a regular spinning disk.

    For example:
    SSD - writing 100MB, uses 3 Watts, takes 1 second, totaling 3 Watts.
    HD - writing 100MB, uses 2 Watts, takes 2 seconds, total = 4 watts.

    In that case, SSD still (and will almost always) win over rotating HD.
    (Not to mention the fact that under normal operations, and HD will be using a lot more power moving the heads back and forth, as multiple programs request their data - something that will be a lot less of an issue in SSDs and their "instant" readback.
  • 1 Hide
    Anonymous , 3 July 2008 01:32
    I agree with Mastrom, you've put out some pretty strong words and haven't backed them up with hard data.

    Do another test, run a CPU & Drive intensive task on a regular basis (every 5 mins or so) and them come up with some times.
  • 0 Hide
    sensorman , 18 July 2008 04:32
    someone wrote:
    For example:
    SSD - writing 100MB, uses 3 Watts, takes 1 second, totaling 3 Watts.
    HD - writing 100MB, uses 2 Watts, takes 2 seconds, total = 4 watts.
    ---

    the idea is good but the result is erroneous.

    Watts are a unit of power. Power has units of energy/time

    A watt is defined as a Joule/second

    so if you are dissipating 2 watts for one second, you consume 2 Joules of energy. If you dissipate 2 watts for two seconds you consume 4 Joules of energy.

    ---
    A few criticisms of the report:
    1) the statement of the conclusion was broader than supported by the data presented. A more accurate conclusion would be that the results reported were for the small sampling of drives tested.

    2) there is a wide range of performance and power that can be observed in SSDs. At the two most recent Intel Developer Forums, Intel presented data on several drives and showed how the power dissipation had large differences from drive to drive. Not surprisingly, Intel's drives showed the best performance and power profile.

    It would have been a more credible report had one of the Intel drives been tested since they claim it is best in class.

    I am not surprised to see big differences in performance from one vendor to the other, particularly when one considers that the technology is relatively immature. In my experience it often takes a few generations of design to really have optimal performance.

    Consumer affordable high performance SSDs based on NAND flash haven't been around for very long so I would submit the controller designs that essentially establish the performance of the drives, are at a lower level of maturity than are other elements in the system.

    I'd expect to see significant improvements in the power profile in the next few quarters/years.

  • 0 Hide
    Anonymous , 7 August 2008 10:21
    Okay is it just me or does this article contradict itself with it's own
    findings?! Look at the last graph under Energy Consumption [Watts], the
    conventional Hitachi HDD chews more power at idle and under load then
    2 of the flash based HDDs! Granted the Hitachi conventional HDD
    probably has better power saving features than the current flash based
    HDDs but in general flash based HDDs do have better power consumption,
    the last graph shows that! Usually the power saving features of a HDD are
    written in the driver or software that is available sold with the drive or easily downloaded from the manufacturer's website! Are the right drivers and software being used on these drives? Correct me if I'm wrong it is not stated what software/driver these drives are using! I have a feeling if the Dell laptop they were using was purchased with a flash HDD already installed and the apropriate software/driver pre-installed on the laptop the results would be a lot different! Being that they are new technology XP or Vista's standard drivers probably don't have any power saving features coded in the driver!