Jump to content











Photo
- - - - -

The USB-Stick Benchmark Thread


  • Please log in to reply
47 replies to this topic

#1 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 June 2010 - 05:30 PM

USB-Sticks are a pain in the behind! :)

Not only, do most of them not feature any kind of useful technical data, on the shiping box, but the fact, that if one feature is in one stick better than in another, does not mean that this goes for all features.
So instead of simply extrapolating from one known value, an actual test suite is needed, to get any idea, if the stick is any good for the intended purpose.

Here's were benchmark programs enter the picture.

A good benchmark program should be easy to use and easy to understand, yet give all the necessary informations.

For this short review, i already eliminated all programs, which can test only a single feature or require expert knowledge to even set the test properly up.

The survivors are:
- h2testw
- Atto Disk Benchmark
- Crystal Disk Mark
- AS SSD Benchmark
- HD Tune

h2testw
----------
h2testw is a very simple program. It just tests sequentiell read and write and creates an error log for the tested area.
Values shown are usually a bit lower than what other tools show.
Very easy to use and good enough for all, who move only big files. Though, where big begins, differs from stick to stick.

H2testw.gif

Atto Disk Benchmark
-------------------------
Atto Disk Benchmark is a very nice tool, as it shows the read and write rates for any number of different file sizes and thereby gives one a good idea, what performance to expect from the drive in any situation.
Unfortunately all values show sequentiell operations. Not a single random access test. Also no display of accesstimes.
But if you use your stick solely for transporting files from her to there, this is the only tool you will ever need!

Atto.gif

Crystal Disk Mark
---------------------
A very nice and simple interface, which gives you not the in depth view of Atto, but instead an overview of the usefulness of the stick in various usage scenarios.
On top we see the top read and write speeds, which only big files will ever have.
Then we have the 4k and 512k random tests. Which show performance while reading or writing the usual small individual files, like pictures and documents.
And last a test, which is suppose to emulated the performance of the stick, when a program is directly run from it.

CrystalDM.jpg

AS SSD Benchmark
------------------------
In essence a Crystal Disk Mark on steroids, with some major drawbacks, i don't want to keep from you.
A:) This is a .Net application, so imo not usable in a PE
B:) There is no way to select the size to be tested. It always uses 1GB, which means easily hours for the 4k tests.

ASSSD.gif (Despite the picture, this comes also with an english interface!)

HD Tune
-----------
A classic. It gives top read speed and read access time. Equiped with those two and some experience on can easy guess the performance loss with smaller files. However the missing write tests, make it only useful for people, who hardly ever write to the stick, like for the use as a boot stick.
(Write tests are available in the buyware HD Tune Pro.)

HD_Tune.gif

Last, but not least two recommendations from me.

The SanDisk Cruzer Micro Skin, is without a doubt the most sturdy USB-Stick, i have ever seen.
What makes this stick so sturdy is, the genius constructed shell. Instead of the usual, where the connector is just soldered to PCB and the PCB then put in two plastic shells, here the outer connector and the shell holding the PCB are a single metal part wrapped in a thin silicon skin.
A friend of mine managed, by accident, to lift his whole PC of the table by this stick.
The HDD was toast afterwards, but the stick didn't even had a bend.
- top read speed 24MB/s
- top write speed 10MB/s
- price ~16Euro for the 8GB version

SanDisk.jpg

The Supertalent Pico-?, deliver the best boot performance for the buck, but also have a lot of problems.
For the Pico-C, SuperTalent seems to have reinvented the USB standard to their liking. Many people report, that the stick is about 0,5mm too thick and has to be forced into the port on first use. Once the port is widened, it works good.
To give you an idea, what forcing in means, take this: More than one person has reported, that he actually broke the port while trying to get the stick in! :)
The Pico-A line does not have this problem. The stick fits easily in the USB-Port and delivers even a bit better performance. (Since i only own this version, the swivel to be exact, i can only compare to what i've read about the C.)
HOWEVER the stick is very thin and made from plastic. I would really think twice, before putting it into a port, while being angry. :D
- top read speed 30MB/s
- top write speed 12MB/s
- price ~18Euro for the 8GB version

If you're looking for a stick with any higher write performance, you immediately double the price tag.

Pico_A.jpg


Reviews and recommendationss for other good or even better benchmark programs and USB-Sticks are welcome. :)


:D

#2 steve6375

steve6375

    Platinum Member

  • Developer
  • 7566 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films
  •  
    United Kingdom

Posted 08 June 2010 - 07:36 PM

Hi
May be useful for some people - RMPrepUSB now has a speed test function - it does a 60MiB read starting from Block 0 followed by a 60MiB write back. Takes just a few seconds and appends results to a .csv file so you can build up a results file. The test uses raw block writes and so the filesystem is irrelevant. This was done so I could compare UFDs just for raw speed.

a .csv file is shown below:

USB MODEL SIZE (MiB) Read/Write SPEED (MiB/S)
Verbatim STORE N GO USB Device 7639 READ 29.8
Verbatim STORE N GO USB Device 7639 WRITE 9.2
HP v135w USB Device 15423 READ 26.2
HP v135w USB Device 15423 WRITE 10.3
USB DISK Pro USB Device 990 READ 9.8
USB DISK Pro USB Device 990 WRITE 8.5
TakeMS MEM-Drive Smart USB Device 7711 READ 33.1
TakeMS MEM-Drive Smart USB Device 7711 WRITE 6.1
LEXAR JD SECURE USB Device 489 READ 15.9
LEXAR JD SECURE USB Device 489 WRITE 3.8
SanDisk Cruzer USB Device 3835 READ 18.6
SanDisk Cruzer USB Device 3835 WRITE 4.6
Kingston DataTraveler 2.0 USB Device 953 READ 14.6
Kingston DataTraveler 2.0 USB Device 953 WRITE 5.8
Generic USB Flash Disk USB Device 124 READ 5
Generic USB Flash Disk USB Device 124 WRITE 0.9
Lexar JD FireFly USB Device 7647 READ 21.7
Lexar JD FireFly USB Device 7647 WRITE 7.1
Patriot Memory USB Device 7643 READ 31.9
Patriot Memory USB Device 7643 WRITE 7.7
Lexar JD FireFly USB Device 15295 READ 22
Lexar JD FireFly USB Device 15295 WRITE 10.1
HP v135w USB Device 3853 READ 22.2
HP v135w USB Device 3853 WRITE 4.7
Netac OnlyDisk USB Device 956 READ 14.7
Netac OnlyDisk USB Device 956 WRITE 3.3
USB 2.0 USB Device 16383 READ 18.6
USB 2.0 USB Device 16383 WRITE 3.5
Intenso Rainbow USB Device 15423 READ 32.9
Intenso Rainbow USB Device 15423 WRITE 6.1




Comparing to CrystalMark for a few USB sticks...
e.g. RMPrepUSB result in () in MB (converted from MiB)

Unknown make 1 Seq Rd= 9.69(10.27) Seq Wr=8.52(8.91)
USB Disk Pro Seq Rd=19.32(19.39) Seq Wr=3.41(3.57)
TakeMS MEM-Drive Smart USB Device Seq Rd=34.24(34.7) Seq Wr=6.26(6.40)

The results are reasonably comparable to Crystal Mark.





#3 patsch

patsch

    Silver Member

  • Advanced user
  • 785 posts
  •  
    Germany

Posted 08 June 2010 - 08:59 PM

@MedEvil: AS SSD Benchmark works in PE (win7 pe se and win7rescuepe) as seen in this thread:
http://www.boot-land...?...ost&p=99997

#4 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 June 2010 - 09:04 PM

Steve are you sure your testing algorythm is working right?
Some of the read speeds seem too high.

USB MODEL SIZE (MiB) Read/Write SPEED (MiB/S)
Verbatim STORE N GO USB Device 7639 READ 29.8
Verbatim STORE N GO USB Device 7639 WRITE 9.2

TakeMS MEM-Drive Smart USB Device 7711 READ 33.1
TakeMS MEM-Drive Smart USB Device 7711 WRITE 6.1

Intenso Rainbow USB Device 15423 READ 32.9
Intenso Rainbow USB Device 15423 WRITE 6.1

I found the Intenso 8GB Rainbow listed with r:28 w6,5 at Intensos site, would be a first to market a stick slower than it actually is.
And the Verbatim should have the speed of r:10 w:4 acording to Verbatim.


:)

#5 steve6375

steve6375

    Platinum Member

  • Developer
  • 7566 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films
  •  
    United Kingdom

Posted 08 June 2010 - 10:08 PM

CrystalMark Sequential results in MB (with mine in (MB) )


verbatim 26.1(29.8) 7.5(9.2) (AS SSD Benchmark rd=25.5 wr=8.95)
Intenso 22.1 (32.9) 4.36(6.1) (AS SSD Benchmark rd=25.7 wr=5.3)

I use API raw writes with no OS filesystem overhead and read/write an entire chunk of 60MB with one API call.

So these figures are best possible case. You would be hard pressed to achieve these results through an OS filesystem and clusters.

#6 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 08 June 2010 - 10:44 PM

@MedEvil: AS SSD Benchmark works in PE (win7 pe se and win7rescuepe) as seen in this thread:

Didn't meant 'unusable' as in 'it will not work' but as in 'makes no sense sizewise'. But for those, who have .net included in their PE anyway, surely a nice program.

:)

#7 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 13 June 2010 - 06:02 PM

Had today to find out about the influence of the hostmachine on the Benchmarks. :)

While visiting a friend, i insisted on a retest of his brand new stick, since the values he told me, were way beyond the ones the manufacturers gave.

Yet every tool confirmed the values and cache was indeed turned off.
So we tested mine.
On my computer it gives me: 12MB/s write- and 30MB/s readspeed.
On his computer it gave: 7MB/s write- and 35MB/s readspeed.

Now curious, we went to his sons computer and my stick gave: 14MB/s write- and 34MB/s readspeed.

7MB difference writespeed that's almost twice what any nonamestick can deliver! :)

:)

#8 steve6375

steve6375

    Platinum Member

  • Developer
  • 7566 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films
  •  
    United Kingdom

Posted 13 June 2010 - 06:21 PM

Which benchmarks and which OS's?

http://www.tomshardw...et,1940-19.html

#9 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 13 June 2010 - 08:31 PM

Atto and Crystal. On mine it was XP, on the one of my friend, it was Vista and on the one of his son, it was Win7.

:)

#10 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 13 June 2010 - 08:46 PM

Those Benchmark results look terrible. On my old P4, the USB HDD reached between 31 and 32 MB/s. (measured with HD-Tune)

:)

#11 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 13 June 2010 - 09:38 PM

Tested all my USB-Sticks on all my computers. (just with Atto)
Worst and best values for any transfer size for a stick, does not vary more than 500KB/s.

From this tests i would say, that neighter CPU performance nor bus speed nor onboard or PCi card solution make any difference. The only possible difference left would be the OS. (All my machines run XP.)
But seriously? :)


:)

#12 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 13 June 2010 - 09:55 PM

Little Warning!

Different HD-Tune versions measure different speed values on the same drive.
Version 2.53 (included in NaughtyPE) measures consistently ~4MB/s less then version 2.55, which is more in tune :) with results from other programs.

:)

#13 steve6375

steve6375

    Platinum Member

  • Developer
  • 7566 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films
  •  
    United Kingdom

Posted 15 June 2010 - 10:59 AM

Feedback from Atto about their benchmark test:

>>>Our Benchmark Tool uses the Windows CreateFile, ReadFile and WriteFile APIs. It creates a file on the drive, then goes through one pass to write the entire file to get the disk space allocated to the file, then it runs the tests. It does not do raw I/O because there is no raw I/O API that allows multiple outstanding requests at a time (queue depth).

So any results would be dependant on the OS, filesystem, chipset and driver.

#14 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 15 June 2010 - 12:27 PM

So any results would be dependant on the OS, filesystem, chipset and driver.

Since USB drivers are always Windows default, have you ever seen a comparison of the performance of the different Windows in this regard?

Maybe running the tests again on those computers, but this time from a XP-PE, Vista-PE and a Win7-PE would be interesting?


:thumbup:

#15 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 15 June 2010 - 06:01 PM

Did a test on my new computer.
Tested with Atto my little old stick, which i already used for the snapshots, because it is mostly empty these days.
Win 7 home premium 64Bit full install vs. NaughyPe 32Bit (XP)
NaughtyPE mostly brought the same speed, as my test on my WinbuilderPC. Was just between 500-800KB/s slower.
Win7 showed, though still on the same machine, even 1MB/s less. :thumbup:

I would have sworn, that with the right chipset - and processor drivers, that Win7 should have had an advantage. Appearantly i was wrong.

XP rulez!
;)

:cheers:

#16 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 16 June 2010 - 04:31 PM

Technology is a wondrous thing.

Raced my stone age USB1.1 Stick against a USB2.0 one in Atto, both on a USB1.1 and a USB2.0 port.
Guess which won! Wrong! :cheers:

Would you have guessed that a USB1.1-Stick can outperform a USB2.0-Stick?
And boy can it outperform!
On the USB1.1 port, up to 16KB it writes more than twice as fast, to then with 32KB go for 50% faster and then settle for being 25% faster for the rest of the tests.
Readspeeds are consistently, but not significantly, always 1 to 5KB/s higher.

If anyone can explain this, please don't hassitate! ;-)

OK, switch to the USB2.0 port. Stone age stick don't stand a chance here, right?
Guess what? Up to 16KB it again outperforms the faster stick in write speed, but then looses big time once the USB2.0 stick finally kicks into high gear.

Yet interesting to notice is, that the old stick still more than doubles it top write speed and more than triples its read speed, against the USB1.1 port test.

I guess only the gods of USB-Sticks can explain, why a cheap USB1.1 stick has the ability to perform way better than 1.1 speed, while a brand USB2.0 stick can't even perform close to 2.0 speeds. :thumbup:


:cheers:

#17 steve6375

steve6375

    Platinum Member

  • Developer
  • 7566 posts
  • Location:UK
  • Interests:computers, programming (masm,vb6,C,vbs), photography,TV,films
  •  
    United Kingdom

Posted 16 June 2010 - 07:32 PM

Both sound slow to me, should be in MB/s not KB/s ?? :thumbup:

#18 eboyhan

eboyhan
  • Members
  • 3 posts
  •  
    United States

Posted 16 June 2010 - 08:56 PM

I got an email today from you guys about USB flash drive performance and referencing this thread. I recently went shopping for a largish high performance UFD to serve as a multi-os boot key. The drive that I came up with was the Corsair Flash Voyager GTR 32GB. MFR claims 34GB read 28 GB write. Benchmarks (there is a good recent one at hardwarecanucks.com) show that for large block sequential IO 34GB read 20GB write is probably closer to the mark. Just thought I'd mention it here as I didn't see it mentioned in any of the tables. It is relatively new, and physically large (about the size of a Swiss army knife).

Cheers,
Ed Boyhan

#19 jfk001

jfk001
  • Members
  • 1 posts
  •  
    Germany

Posted 17 June 2010 - 07:02 AM

Hello!

OCZ THROTTLE 32GB:

-----------------------------------------------------------------------

CrystalDiskMark 3.0 x64 (C) 2007-2010 hiyohiyo

						   Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]



		   Sequential Read :	29.751 MB/s

		  Sequential Write :	24.400 MB/s

		 Random Read 512KB :	30.364 MB/s

		Random Write 512KB :	18.340 MB/s

	Random Read 4KB (QD=1) :	 6.270 MB/s [  1530.8 IOPS]

   Random Write 4KB (QD=1) :	 1.546 MB/s [   377.5 IOPS]

   Random Read 4KB (QD=32) :	 7.195 MB/s [  1756.7 IOPS]

  Random Write 4KB (QD=32) :	 1.504 MB/s [   367.2 IOPS]



  Test : 100 MB [Z: 28.1% (8.4/30.1 GB)] (x1)

  Date : 2010/06/17 8:47:33

	OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x64)

Corsair VoyagerGT 8GB:

-----------------------------------------------------------------------

CrystalDiskMark 3.0 x64 (C) 2007-2010 hiyohiyo

						   Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]



		   Sequential Read :	23.614 MB/s

		  Sequential Write :	17.833 MB/s

		 Random Read 512KB :	23.481 MB/s

		Random Write 512KB :	10.057 MB/s

	Random Read 4KB (QD=1) :	 5.170 MB/s [  1262.3 IOPS]

   Random Write 4KB (QD=1) :	 0.105 MB/s [	25.6 IOPS]

   Random Read 4KB (QD=32) :	 6.418 MB/s [  1566.9 IOPS]

  Random Write 4KB (QD=32) :	 0.116 MB/s [	28.2 IOPS]



  Test : 100 MB [X: 58.7% (4638.7/7904.0 MB)] (x1)

  Date : 2010/06/17 9:00:11

	OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x64)
Corsair Flash Voyager 32GB:

-----------------------------------------------------------------------

CrystalDiskMark 3.0 x64 (C) 2007-2010 hiyohiyo

						   Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]



		   Sequential Read :	23.289 MB/s

		  Sequential Write :	12.421 MB/s

		 Random Read 512KB :	23.217 MB/s

		Random Write 512KB :	 0.915 MB/s

	Random Read 4KB (QD=1) :	 5.048 MB/s [  1232.4 IOPS]

   Random Write 4KB (QD=1) :	 0.007 MB/s [	 1.7 IOPS]

   Random Read 4KB (QD=32) :	 5.389 MB/s [  1315.7 IOPS]

  Random Write 4KB (QD=32) :	 0.007 MB/s [	 1.7 IOPS]



  Test : 100 MB [Y: 22.0% (6.7/30.2 GB)] (x1)

  Date : 2010/06/17 8:54:45

	OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x64)


#20 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 17 June 2010 - 10:20 AM

Benchmarks (there is a good recent one at hardwarecanucks.com) show that for large block sequential IO 34GB read 20GB write is probably closer to the mark.

I had a look at the review and i wouldn't trust their tests. I have several SanDisk Cruzer Micro and not one performs like their test show. They usually give between 22 and 24 MB/s not over 29MB/s as their test shows.

When you get it, maybe you can run a test and post here.

:thumbup:

#21 stressfreesoul

stressfreesoul
  • Members
  • 1 posts
  •  
    United Kingdom

Posted 17 June 2010 - 10:25 AM

I use both the new and old Transcend Jetflash 8Gb, Samsung chips (supposed to be fast)

New - Write/8.4Mb Read/32.4Mb
Old - Write/12.5Mb Read/19.5Mb

Formatted in Fat32, both empty after a fresh format.
It would seem they both have their advantages...

On the subject of fast read/write, the Corsair Flash Voyager GTR 128Gb... are the chips arranged in RAID0 to attain such high writes?

#22 MedEvil

MedEvil

    Platinum Member

  • .script developer
  • 7771 posts

Posted 17 June 2010 - 10:49 AM

On the subject of fast read/write, the Corsair Flash Voyager GTR 128Gb... are the chips arranged in RAID0 to attain such high writes?

It uses a quad channel controller.

:thumbup:

#23 mateuszek

mateuszek

    Member

  • Members
  • 93 posts

Posted 17 June 2010 - 06:13 PM

Kingston Data Traveler 2gb

-----------------------------------------------------------------------
CrystalDiskMark 3.0 © 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 19.955 MB/s
Sequential Write : 8.065 MB/s
Random Read 512KB : 19.903 MB/s
Random Write 512KB : 2.337 MB/s
Random Read 4KB (QD=1) : 5.620 MB/s [ 1372.0 IOPS]
Random Write 4KB (QD=1) : 0.026 MB/s [ 6.4 IOPS]
Random Read 4KB (QD=32) : 6.034 MB/s [ 1473.1 IOPS]
Random Write 4KB (QD=32) : 0.033 MB/s [ 8.0 IOPS]

Test : 50 MB [N: 0.0% (0.0/1304.8 MB)] (x1)
Date : 2010/06/17 20:11:58
OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x86)

#24 mateuszek

mateuszek

    Member

  • Members
  • 93 posts

Posted 17 June 2010 - 06:17 PM

Sandisk Cruzer Titanium 8gb

-----------------------------------------------------------------------
CrystalDiskMark 3.0 © 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 28.831 MB/s
Sequential Write : 9.296 MB/s
Random Read 512KB : 28.807 MB/s
Random Write 512KB : 2.791 MB/s
Random Read 4KB (QD=1) : 5.555 MB/s [ 1356.2 IOPS]
Random Write 4KB (QD=1) : 0.033 MB/s [ 8.1 IOPS]
Random Read 4KB (QD=32) : 6.281 MB/s [ 1533.3 IOPS]
Random Write 4KB (QD=32) : 0.045 MB/s [ 11.1 IOPS]

Test : 50 MB [M: 17.8% (1364.7/7671.3 MB)] (x1)
Date : 2010/06/17 20:16:27
OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x86)

#25 mateuszek

mateuszek

    Member

  • Members
  • 93 posts

Posted 17 June 2010 - 06:21 PM

Lexar retrax 16GB

-----------------------------------------------------------------------
CrystalDiskMark 3.0 © 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 17.996 MB/s
Sequential Write : 8.375 MB/s
Random Read 512KB : 17.924 MB/s
Random Write 512KB : 1.514 MB/s
Random Read 4KB (QD=1) : 6.706 MB/s [ 1637.2 IOPS]
Random Write 4KB (QD=1) : 0.015 MB/s [ 3.7 IOPS]
Random Read 4KB (QD=32) : 7.985 MB/s [ 1949.5 IOPS]
Random Write 4KB (QD=32) : 0.020 MB/s [ 5.0 IOPS]

Test : 50 MB [M: 7.7% (1.0/13.4 GB)] (x1)
Date : 2010/06/17 20:21:07
OS : Windows 7 Enterprise Edition [6.1 Build 7600] (x86)




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users