Sign in with
Sign up | Sign in

Memory Corp F4 MLC SSD, 32 GB

Six New SSDs: Can Intel Be Dethroned?
By

Memory Corp provided us with two flash SSDs to review: the F4 single-level cell (SLC) drive, which is an industrial high-performance drive (see the next page) and the F4 MLC for business customers. The F4 MLC is based on cheaper MLC flash memory and is available from 8 GB to 128 GB capacities. We looked at a 32 GB model.

The data sheet talks about 150 MB/s read performance and 90 MB/s write throughput. Our testing revealed considerably different results: 110 MB/s sequential read throughput and 39 MB/s sequential write transfer rate. We also measured the interface throughput to be sure that we weren’t creating a bottleneck for the SATA/300 device: 175.4 MB/s interface bandwidth was available. Clearly, the controller and/or flash memory aren’t capable of sustaining constant, high throughput.

I/O Performance Nightmare

While most flash SSDs show excellent access time of less than 1 ms—sometimes even less than 0.1 ms (100 ns)—this does not automatically mean that they also deliver high I/O performance. Have a look at our workstation I/O test or our IOmeter runs. With the exception of the webserver I/O benchmark pattern (which requests blocks small enough to fit into a single request) all other I/O benchmarks attest to the horrible I/O performance of this drive. Even conventional hard drives do better in these benchmarks!

Acceptable Application Benchmarks and Power Consumption

The PCMark application benchmarks suffer from the low I/O performance, but they still show that the drive is rather quick; at least, quicker than conventional hard drives. The 0.7 to 1.9 W power consumption was very acceptable as well. Check out the power consumption for DVD playback in the charts section as well, where you’ll find that this drive consumed the least power to deliver this defined stream of data.

Display all 38 comments.
This thread is closed for comments
  • -4 Hide
    ava__ , January 28, 2009 5:13 AM
    Quote:
    Customers, such as Fujitsu-Siemens Computers (FSC), have been trying to validate flash SSDs for enterprise applications by looping hardcore I/O loads, and they all failed with write errors after only a few months


    citation, please?
  • 0 Hide
    montyuk , January 28, 2009 8:26 AM
    can you add the runcore ssd to your list of ones to do next?
  • 6 Hide
    hankscorpio1 , January 28, 2009 8:29 AM
    What is the comparison against conventional drives?
  • 0 Hide
    marraco , January 28, 2009 10:48 AM
    Good article. Its sad than the intel have such a bad write performance.

    I would like to make an XY chart.
    Please, can you add orientative prices?
  • 1 Hide
    runmymouth , January 28, 2009 12:11 PM
    To be honest I personally don't mind the write performance being a little slow, if it can read that fast. Too bad they are still so $$$ for intel's SSD.
  • -4 Hide
    zedx , January 28, 2009 12:11 PM
    I personally don't think that SSDs would be any competition for HDD's for a loooooong time ahead. The previous generation (150 mbps / 100mbps ) should at least be 0.25 $ per GB and last a lot longer to make sense for the general public and even to must high end desktop buyers (at least to me). Now even if the high end ssd's saturate sata 3 gbps by the year end and the speeds increase about 30 - 50 mbps per year the reduction in price would mostly be about 75 - 100 %. So by 2012 - 2013 people who use less space (about 250GB) might get them instead of HDD's. Of course some people would be buying say 80GB for $30 along with xTB HDD's and these things might make great gifts. By 2014 - 15 I think these would begin to replace HDD's in general computers(If were alive) as most people don't require
    excess on 500GB (depends though on that age's media). But for the real thing (read write around 300mbps) to get mainstream we might have to wait till about 2016. By then we should be having 12 core CPU's if they are able to further shrinken it...
  • 2 Hide
    one-shot , January 28, 2009 2:47 PM
    For a comparison of conventional drives, even the mighty WD 640AAKS and Intel's SSD's, visit this link http://techreport.com/articles.x/16291.
  • 3 Hide
    Anonymous , January 28, 2009 2:55 PM
    No x25-E?
  • 2 Hide
    Anonymous , January 28, 2009 3:10 PM
    It`s a good review though the drives are not really representative of what we are going to see this year. They perform so similarly to each other that it makes me think they are different rebrands of the same outdated components. Even the problematic OCZ Core would handily outperform those drives, not to mention the promising Vertex.
  • 0 Hide
    descendency , January 28, 2009 3:10 PM
    joejoe55No x25-E?

    If they compared the x25-e, the others would be butt hurt from the rape.

    Price per performance might be bad for the x25-e, but the performance is top for SSDs.
  • 5 Hide
    Siffy , January 28, 2009 3:38 PM
    Geez, how long does it take articles to make it onto the website from conception to print? "hard drives have gotten really cheap. Terabyte hard drives are well below $150, and you can get 500 GB drives for as little as $75." You can find TB drives for $100 now and I just recently picked up a latest generation 500GB for $60. As for the SSDs compared, they're all from companies that aren't talked about often when it comes to SSDs. Why were the current G.Skill and OCZ drives not even mentioned? The OCZ that was mentioned is an insanely expensive SLC drive that's been out for a year. You can get an OCZ Solid 250GB for less than the OCZ SATA II SLC 64GB. As for "Can Intel Be Dethroned?" No, it can't when you don't actually test it against competing products.
  • 2 Hide
    deiceman , January 28, 2009 3:47 PM
    For most users the write time is almost insignificant. But read time greatly enhances boot-up and application launch. I might read 100 GB per day but write no more than 10 MB per day. I have Vista Ultimate on 32 GB OCZ SSD and the boot up is remarkable. I will buy another 32 GB SSD for applications. All data is written to RAID-1 hard drive array.
  • 2 Hide
    Anonymous , January 28, 2009 4:59 PM
    well I don't think price per GB matter that much. Because I have a cheap OCZ coreV2 30gb ssd (got it for 50 bucks) which is more than enough for OS and programs and I use traditional HDD for storage. After all you don't need SSD speed for picture, mp3, and movie and other document. Just do some research online to fix the random writing issue, the speed is unbelievable.
  • 2 Hide
    Wixman , January 28, 2009 5:42 PM
    This article is worthless. The tested SSDs aren't even the current models. The new SSDs for 2009 are here, and none of them were tested.
  • 1 Hide
    marraco , January 28, 2009 8:12 PM
    Ok. There is the price-Performance chart, based on read performance:

    and here is the relative chart
    It shows extra dollars payed over the cheaper unit, in Y, graphicated as function of the performance percentage gain over the cheaper unit (in X).

    This chart may have wrong data, because I take the data from e-bay, Amazon, and Google Shopping.
    But I needed to get data from somewhere. This is the best I made.
    Also, the comparison is not fair, because we are comparing different capacities of disks.
    Units with lower storage capacity may have the same speed, but lower price. Anyway, this is the only data we have.

    I omitted this units because I got no price data:
    Memory Corp F4 SLC
    Silicon Power SLC
    Chaintech Apogee
    Memory Corp F4 MLC
    Soliware
    Hama Highspeed Flash SSD 2,05"
    Hama Highspeed Flash SSD 3,5"
    Hama Flash 2,5"
    Hama Flash 1,8"


    Each point represent a Price-Performance point for a SSD drive.
    Only the units united by the blue line are worth of a buy.
    Any other SSD disk is more expensive or slower.
    The Samsung SATA 2 64 GB is the cheaper, so all the others are compared with it, to see how much performance can be gained by investing more money. If you are strict on budget, you should buy it.
    Paying 24 U$S for the Super talent MX give 20% extra read speed over the Samsung, with 4 Gb less.
    The green point is the Super Talent OX. It is not on the blue line, but is near enough to be a good choice.
    For 85 $, the Crucial gives 37% extra read speed over the Samsung, but have half the capacity (32 Gb less than the Samsung).
    At first, the Intel X25-M is the performance winner. It cost 200U$S more than the Samsung, giving 120% more performance. And his cost does not growth exponentially as is common in premium performance hardware. Better, it haves 80 Gb, so it not only givers performance, but also storage capacity.

    After all, The Intel may not be a smart choice.
    SSD units have little performance penalty when assembled in RAID 0, so we need to compare the Intel against a RAID 0 of the cheaper units.

    The red points represent faked price and performance data. They have two times the price and performance of the cheaper units on the blue line.

    They show that by almost the cost of the crucial, 2x Samsung gain performance near of the Intel, and reach 128Gb, 48 more than Intel.
    The next Fake RAID 0 is the 2X Super Talent MX. It outperforms the Intel, for 19 Mbs, saving 42 U$S, and gaining 40 Gb of storage. Is clearly a winner.
    The next performance king could be the 2X Crucial, in RAID 0, but since RAID0 in SSD have little performance penalty, it may be outperformed by the red circle; it is a 3X Samsung in RAID0.
    It should cost 345 U$S; 25 $ more than Intel, but outperforming the Intel in 72 Mb/s, and reaching 192 Gb, 112 Gb more than Intel.

    So, the best choices probably are, in this order:
    Samsung SSD SATA2 64Gb, giving 90.6 Mb/s in read speed.
    Super Talent MX 60 Gb, giving 109 Mb/s in read speed.
    2x Samsung SSD SATA2 128Gb(=2x64Gb), giving 181 Mb/s in read speed.
    2x Super Talent MX 120Gb(=2x60Gb), giving 218 Mb/s in read speed.
    3x Samsung SSD SATA2 192Gb(=3x64Gb), giving 271 Mb/s in read speed. As a bonus, the 3x Samsung should give 3x83.7=251 Mb/s write speed, meanwhile the Intel will only give 74 Mb/s in writing.

    So, as conclusion:
    We need better data on prices, and a comparison of the Intel against real world RAID0 of cheaper units.

    Now, I list the prices I have found:
    Unit: Intel X25-M; Capacity: 80; read performance: 199; Price: 320
    Unit: Crucial 2,5"; Capacity: 32; read performance: 123,7; Price: 200
    Unit: Mtron Pro 7500; Capacity: 32; read performance: 116,9; Price: 499
    Unit: MemoRight MR25.2-032S; Capacity: 32; read performance: 115,7; Price: 529
    Unit: MemoRight MR25.2-064S; Capacity: 64; read performance: 115,6; Price: 699
    Unit: Super Talent Masterdrive OX 2.5"; Capacity: 64; read performance: 109,8; Price: 166
    Unit: Super Talent Masterdrive MX 2.5"; Capacity: 60; read performance: 109; Price: 139
    Unit: Mtron Flash; Capacity: 32; read performance: 94,6; Price: 230
    Unit: OCZ Sata II 2.5"; Capacity: 64; read performance: 90,9; Price: 158
    Unit: Samsung SSD Sata 3,0 Gbps; Capacity: 64; read performance: 90,6; Price: 115
    Unit: Sandisk SSD5000 2,5"; Capacity: 32; read performance: 68,1; Price: 150
    Unit: Silicon Power 2,5"; Capacity: 128; read performance: 57,1; Price: 286
  • 0 Hide
    marraco , January 28, 2009 8:15 PM
    Ups! I forget to add the names on the first chart. See the second char, the relative prices vs relative performance.
  • 0 Hide
    Shadow703793 , January 28, 2009 8:38 PM
    For people saying SSDs aren't going to replace HDDs soon, take a look at how many OEMs offer SSDs on their laptops. Don't forget the net tops either.
  • 2 Hide
    TechDicky , January 29, 2009 1:10 AM
    Pretty good article. Has its pluses and minuses. One thin I noticed as I looked the charts was that the best performer across the board seemed to be the MemoRight MR25s. While they may not have the Sequential Read performance of the Intel, they smoke Intel and outperform all other drives on Sequential Write performance They also outperform or on par with the Intel and all other drives in all of the I/O performance charts. However, the MemoRight drives seemed to be missing from all of the efficiency charts? I know they weren't the focus of this article, but it would have been nice to include comparison drives in ALL of the charts.

    On a side note: The line graphs are very difficult to read. For one thing the colors should be consistent from one graph to the next. Another thing is you need better differentiators than subtle color differences. Maybe it is these crappy unadjusted LCDs I'm on, but that is pretty typical. Either split them up so that there are no more than 10 components per line graph or use more distinguished charateristics (line thickness, dots, dashes, and other patterns as well as stark contrast colors.

    Regards...
  • 3 Hide
    Area51 , January 29, 2009 1:44 AM
    Can you please include the Intel x25-e on your next benchmark? Also what are the MTBF of these drives?
    I have been testing the Intel x25-m 80GB 2.5 and I have never been able to go above 1W. How did you do it? How did you monitor the power consumption?
  • 0 Hide
    Area51 , January 29, 2009 1:46 AM
    Is Passmark performance test a good way to mesure performance?
Display more comments