Recommendations on USB memory sticks?

[Stereodude]

Sandisk Extreme III SDHC (older non 30MB/sec one):

           Sequential Read :    19.996 MB/s
          Sequential Write :    19.755 MB/s
         Random Read 512KB :    19.788 MB/s
        Random Write 512KB :     3.379 MB/s
    Random Read 4KB (QD=1) :     3.808 MB/s [   929.6 IOPS]
   Random Write 4KB (QD=1) :     0.034 MB/s [     8.3 IOPS]
   Random Read 4KB (QD=32) :     3.866 MB/s [   943.8 IOPS]
  Random Write 4KB (QD=32) :     0.063 MB/s [    15.3 IOPS]

  Test : 50 MB [I: 0.0% (0.1/7782.5 MB)] (x4)

Interesting that the Lexar card is a much better performer with the 4kB files.

 

[mubs]

Performance aside, speaking of small capacities, I have 2 64MB flash drives (that I don't use) and one 256MB Lexar Jumpdrive that comes in handy because it has a write protect switch for when I need to take photos to print at a local store. Why have they stopped putting the switch on newer thumb drives?

 

[time]

This site gathers results from USB flash drive speed tests. You can contribute by downloading their simple standalone test app. After running the tests, it asks you if you want to upload the results.

You can view individual results or click through to see an average, both of which are graphed. There's often significant variation between test platforms and product samples.

The Corsair Voyager GT is a standout, and I've also just noticed the probably now obsolete OCZ Throttle (combined eSATA & USB 2.0) is quite exceptional.

I think one lesson from all of this is that it will usually pay to pack small files into an archive (even zip uncompressed will do) before writing them to a USB pen drive/stick.

 

[time]

Same 4GB Verbatim Store 'n Go I referred to earlier, only now with 50MB dataset:

           Sequential Read :    25.082 MB/s
          Sequential Write :    10.683 MB/s
         Random Read 512KB :    25.022 MB/s
        Random Write 512KB :     4.188 MB/s
    Random Read 4KB (QD=1) :     3.070 MB/s [   749.5 IOPS]
   Random Write 4KB (QD=1) :     0.625 MB/s [   152.5 IOPS]
   Random Read 4KB (QD=32) :     3.115 MB/s [   760.4 IOPS]
  Random Write 4KB (QD=32) :     0.045 MB/s [    11.0 IOPS]

  Test : 50 MB [G: 5.6% (212.8/3818.0 MB)] (x1)

I've been thrashing it, so writes are probably a lot slower than on a freshly initialized drive.

 

[Stereodude]

This site gathers results from USB flash drive speed tests. You can contribute by downloading their simple standalone test app. After running the tests, it asks you if you want to upload the results.

You can view individual results or click through to see an average, both of which are graphed. There's often significant variation between test platforms and product samples.

The Corsair Voyager GT is a standout, and I've also just noticed the probably now obsolete OCZ Throttle (combined eSATA & USB 2.0) is quite exceptional.

I think one lesson from all of this is that it will usually pay to pack small files into an archive (even zip uncompressed will do) before writing them to a USB pen drive/stick.

So, I ran the test and submitted it for my Buffalo RUF2-S 4GB stick (Japanese product link). It would land 6th for write speed and 8th for read speed out of all the drives they've had submitted to them. :rofl:

 

[LunarMist]

Sandisk Extreme III SDHC (older non 30MB/sec one):

           Sequential Read :    19.996 MB/s
          Sequential Write :    19.755 MB/s
         Random Read 512KB :    19.788 MB/s
        Random Write 512KB :     3.379 MB/s
    Random Read 4KB (QD=1) :     3.808 MB/s [   929.6 IOPS]
   Random Write 4KB (QD=1) :     0.034 MB/s [     8.3 IOPS]
   Random Read 4KB (QD=32) :     3.866 MB/s [   943.8 IOPS]
  Random Write 4KB (QD=32) :     0.063 MB/s [    15.3 IOPS]

  Test : 50 MB [I: 0.0% (0.1/7782.5 MB)] (x4)

Interesting that the Lexar card is a much better performer with the 4kB files.

I wonder how this card will do?

 

[Stereodude]

I wonder how this card will do?

I tell you what. You buy one, send it to me, and I'll let you know.

 

[LunarMist]

I tell you what. You buy one, send it to me, and I'll let you know.

Maybe I'll buy one and not send it to you. :lol:

 

[mubs]

I think one lesson from all of this is that it will usually pay to pack small files into an archive (even zip uncompressed will do) before writing them to a USB pen drive/stick.

I don't understand why. Isn't near-zero access time a key strength of these things?

 

[time]

Write cycle times can be very slow, especially in comparison to read cycle times. But this is relative, you should try putting your hard drive through those tests and see how it fares.

 

[Stereodude]

But this is relative, you should try putting your hard drive through those tests and see how it fares.

Good idea...

Samsung F4 2TB 5400RPM:

           Sequential Read :   105.556 MB/s
          Sequential Write :   105.540 MB/s
         Random Read 512KB :    71.049 MB/s
        Random Write 512KB :   161.962 MB/s
    Random Read 4KB (QD=1) :     1.026 MB/s [   250.4 IOPS]
   Random Write 4KB (QD=1) :     9.711 MB/s [  2371.0 IOPS]
   Random Read 4KB (QD=32) :     2.226 MB/s [   543.4 IOPS]
  Random Write 4KB (QD=32) :     9.155 MB/s [  2235.0 IOPS]

  Test : 50 MB [D: 61.1% (1139.0/1863.0 GB)] (x4)

And, yes I realize the 50MB test isn't necessarily going to give results accurate for the drive (because of caching), but since the flash drives were tested the exact same way it is a valid point of comparison.

 

[Stereodude]

A more accurate 500MB test:

           Sequential Read :   106.736 MB/s
          Sequential Write :   106.747 MB/s
         Random Read 512KB :    39.349 MB/s
        Random Write 512KB :    55.288 MB/s
    Random Read 4KB (QD=1) :     0.485 MB/s [   118.3 IOPS]
   Random Write 4KB (QD=1) :     1.190 MB/s [   290.6 IOPS]
   Random Read 4KB (QD=32) :     0.852 MB/s [   207.9 IOPS]
  Random Write 4KB (QD=32) :     1.231 MB/s [   300.5 IOPS]

  Test : 500 MB [D: 61.1% (1139.0/1863.0 GB)] (x4)

 

[ddrueding]

I don't understand why. Isn't near-zero access time a key strength of these things?

As Time suggested, doing a write is very slow, irrespective of the size of that write. Doing fewer, bigger writes is much, much faster than doing many smaller ones.

 

[ddrueding]

My now several year-old m84 512GB

Sequential Read :   428.340 MB/s
          Sequential Write :   487.829 MB/s
         Random Read 512KB :   460.070 MB/s
        Random Write 512KB :   462.049 MB/s
    Random Read 4KB (QD=1) :    22.524 MB/s [  5498.9 IOPS]
   Random Write 4KB (QD=1) :    42.193 MB/s [ 10301.1 IOPS]
   Random Read 4KB (QD=32) :   106.153 MB/s [ 25916.3 IOPS]
  Random Write 4KB (QD=32) :    46.159 MB/s [ 11269.2 IOPS]

  Test : 50 MB [C: 60.4% (285.5/473.0 GB)] (x4)

 

[BingBangBop]

I just got myself a 16 GB SanDisk Cruzer and a 16 GB Kingston R500 USB 2.0 drive which cost be about the same money. I can tell you that on a simple multiple GB file copy to the disks that the Kingston is a good 5x faster at 4.06MB/s vs 21.04MB/s on a simple q6600 machine. The Sandisk did come with its own security app that the Kingston didn't have.

I decided to play around with Readyboost and I will say that putting in the Kingston as a dedicated drive significantly speeds up the snappiness feeling of the computer when doing mundane tasks. I monitored the readyboost using the performance monitor and it isn't using that much space either (about 2.5GB) so a smaller drive would seem to be sufficient. I did reformat it with NTFS first so as not to be limited to 4GB.

It is a shame that readyboost won't run on 2008 server though. That is the machine that I would have liked to speed up.

 

[mubs]

I tried Readyboost on a RB ready flash drive, and didn't feel any improvement. Stopped using it.

 

[ddrueding]

If you have enough RAM it doesn't make any difference IMHO.

 

[mubs]

I guess I had enough for the purpose; 2GB & Win 7.

 

[BingBangBop]

On the machine I'm testing I have 4GB ram about half is either free (.8GB) or avail (1.9GB) with 1.2GB cached, but I still notice a difference. I'm also sure that the speed of the USB flash drive matters, not so much in transfer speed which is limited by the USB 2.0 interface, but like an SSD it is random access that matters. I also have a low speed green HD, and since this is a caching tech it is the speed difference between the USB and the HD that matters.

I'm not disagreeing with others when they test they may get different results. I'm also sure that a true RAM cache should be better than a USB thumb drive cache. So if you have enough RAM... However, there are too many variables. So, all I can say is that for my HW, I'm noticing a significant difference with readyboost.

The other thing is that I'm reporting the Kingston R500 is very much faster than the SanDisk Cruzer and they cost about the same at $31 for 16GB.

 

[ddrueding]

B3, I think you are right about the speed difference between the HDD and the caching drive. I don't have any of those "green" drives in operation anywhere.

 

[BingBangBop]

I know it isn't particularly linear but it may give a scale of the improvement. The Microsoft experience index for the disk drive went from 4.9 to 6.2 by adding Readyboost

 

[LiamC]

Before I moved to SSD's, I used a Maxtor Atlas II 15K and Velociraptor 15GB as boot drives. I still feel that the Velociraptor + ReadyBoost was snappier than without ReadyBoost, on Windows 7 + 4GB.

 

[LunarMist]

That must have been a much older WD (early Raptor?) to have such low capacity, but I thought it was 18GB.

 

[ddrueding]

The 18 formatted to around 15.

 

[LiamC]

Typo on my part. 150GB, not 15GB

 

[Adcadet]

Anybody have a small USB drive that they like? I'm looking for something I could clip to my ID to ensure that I always have it with me. Seems there is nothing like this in the world of USB3, although that's not a huge issue since >95% of the computers I use at work don't have USB3.

PNY Micro Swivel? Kingston DataTraveler 108? Super Talent Pico-C?

 

[CougTek]

I've used several SuperTalent Pico USB keys. While they are fast, they have a few important issues. They are so thin that they often don't fit well in USB ports and sometimes they are not recognised because of that. Also, the bottom plate has ended up detaching itself from the key on all the Pico USB keys I've used. You can glue it back or put a heat-shrink band around it, but the key in itself is not very durable as-is.

 

[MaxBurn]

Some of the micro SD card readers are very small, barely bigger than the card. Those are the smallest USB drives I have. I have on on an ID for work.

In other news I picked up a 32gb USB3 patriot supersonic. Don't so much need the space but all that memory and the internal parallelism sure gets some good speeds.

 

[Sol]

I have a SuperTalent Pico attached to my keys. I agree that it doesn't always fit in all USB ports (though it does in most) but I've had no problems with the durability. Mine's been banging around in my pocket for years and other than a few minor dents and scratches isn't any the worse for wear. Maybe I got lucky though, I've only ever had the one.