You are here

Truth in Advertising benchmark scores on USB flash drives

9 posts / 0 new
Last post
EVBrown
Offline
Last seen: 12 years 6 months ago
Joined: 2006-01-10 10:42
Truth in Advertising benchmark scores on USB flash drives

All,

I have read some posts regarding which drives are good for Portable apps, and I wanted to share my experience. I run most of my apps as portable apps on my USB drive, so everything will run exactly the same at work and at home. I even have some old legacy Windows apps that I run entirely from my USB drive. I had a 2 year old 2 GB Lexar Lightning drive that I was happy with, but I needed more space, so I researched which drives were supposed to be fast.

My first purchase was an 8 GB Corsair Flash Voyager GT. It is advertised with 34 MB Read times and 28 MB Write times. Using the ATTO disk benchmark (available from www.hugesystems.com/supportspace/bench32.exe), this drive benchmarked out at 23924 Read and 21203 Write. Far below the advertised times. These scores are on my office PC, a Dell Optiplex GX-620, 3 GHz with 1 GB of RAM.

My next purchase was a 4 GB OCZ ATV-Turbo, with an advertised 33-35 MB Read time and 26-30 Write time. This drive benchmarked out at 31506 Read and 12543 Write. As you can see, the read times are close, but the write times are less than half the advertised rating.

Finally, I wised up and bought a 4 GB Lexar Lightning. It benchmarked out at 32656 Read and 24809 Write. These are better than the advertised 30 MB Read and 21 MB Write times.

Benchmarks aside, the drive FEELS fast. I have very little wait time for applications to pop up, giving me back my screen after file saves, etc.

Just wanted to share that the 4 GB Lexar Lightning was the right purchase for me. I got it for $69.24 from Amazon, shipping included.

8-) Ed

Aciago
Aciago's picture
Offline
Last seen: 8 months 2 days ago
Joined: 2007-01-24 14:23
Well

All that I know is that I have a 4Gb Titan flash drive, one of those Chinese brands I bought very cheap on eBay... I don't know the write/read speed but I know that is incredibly slow. At the beginning I thought that was because it was a 4Gb and of course it should be slower than a 1Gb, but two days ago, I had a failure (I removed the drive improperly and lost 13Mb in chks files) and have to load my last backup, but before that, I decided to format it to NTFS, now is much more faster, I can call it "comfortably faster" now.

Conclusion: The speed also depends on the Flash Drive format.

BTW, Is there any way to return those chk files into their original files? I lost important information that wasn't on any backup... Sad

If a packet hits a pocket on a socket on a port,
and the bus is interrupted as a very last resort,
and the address of the memory makes your floppy disk abort,
then the socket packet pocket has an error to report Biggrin

LOGAN-Portable
LOGAN-Portable's picture
Offline
Last seen: 11 years 4 months ago
Developer
Joined: 2007-09-11 12:24
Bit off topic, but cheap

Bit off topic, but cheap ebay drives should always be tested once by writing 4GB and then checking the hash of the files because many cheap ones do not contain the size of the drive reported by Windows.

rab040ma
Offline
Last seen: 2 months 19 hours ago
Joined: 2007-08-27 13:35
If you know what the data

If you know what the data looks like, you can load each chk file with a hex viewer to see if you can identify it. If you can match one of the chk files with a particular data file you are missing, copy it to a file with the name it should have, load the copy, and see if it works.

It may take some tweaking. Sometimes one file is split into several by the process. I think the files will also behave as if they include the entire allocated space, where the original probably used a portion of the last block.

You're not the first to have this happen, so there are probably better instructions around than just my recollections of having done it a while ago.

MC

Aciago
Aciago's picture
Offline
Last seen: 8 months 2 days ago
Joined: 2007-01-24 14:23
well

with text files this is long but possible, the problem is with images... I have a Doc file with an image in top, after the incident, the picture loads in pieces so I have to reinsert it...

My question is if there i an "automatic" way to do it, like an app that determines which chk files belongs to which document and reinsert it...

I suppose this doesn't exist yet... Sad

If a packet hits a pocket on a socket on a port,
and the bus is interrupted as a very last resort,
and the address of the memory makes your floppy disk abort,
then the socket packet pocket has an error to report Biggrin

Jimbo
Offline
Last seen: 4 years 4 months ago
Joined: 2007-12-17 05:43
You could use something like

the 'file' command in cygwin or linux which will look at the headers of each chk file and tell you what sort of data it is, e.g. jpg image, ms-office document, etc.

But, as for inserting the correct image back into the correct places in the document, that is, and always will be, beyond what can be done automatically. It involves knowing what you want the document to contain, and, although you told it that once, it didn't save correctly and that's how you've ended up with all the .chk files, so that's the best it can do for you.

53376
Offline
Last seen: 15 years 10 months ago
Joined: 2008-01-12 14:50
NTFS

Doesn't NTFS make the usb wear out considerably faster? Much more than lets say if it was Fat32. Shock

Jimbo
Offline
Last seen: 4 years 4 months ago
Joined: 2007-12-17 05:43
That depends on who you ask

and the more I research it, the less definite the answer gets.

Due to it's journal, NTFS write transactions involve more writes to disk

However, for very small files, they can be stored in the MFT itself, meaning that instead of writes to the FAT, and the directory, and the file, there are only writes to the MFT and directory.

Also, from testing that has been done (this is very poorly documented by Microsoft), it appears that for FAT formatted removable drives, the write cache is turned off, which means that if you write 30 files to a directory, there are 30 separate updates to the files, 30 to the directory entry, and 30 to the FAT.

Whereas for NTFS, the cache is on, meaning that for the same scenario, there could be as little as 30 writes to the files, 30 to the journal, and 1 each to the MFT and directory, which is a significant improvement.

For me, the big reasons against NTFS on a flash drive are

security - if you plug it into a different machine, there is a good chance that you will not be able to access your own files, because their ownership is messed up.

and

MS don't expect it - all of their NTFS caching code is expecting a fixed disk, and isn't designed with unpluggability in mind. This means that sometimes when you try to eject the disk, it refuses, telling you that it cannot be safely shut down at this time, for no reason that I can ever fathom, and you end up needing to either risk just pulling it out, or rebooting the PC.

I now stick to FAT on my 8GB drive, even though I have a truecrypt container on there that is limited to 4GB by the FAT file system.

rab040ma
Offline
Last seen: 2 months 19 hours ago
Joined: 2007-08-27 13:35
add to existing topic?

You might want to visit an existing topic and add your benchmarks there. I don't think that's bumping. They're trying to collect numbers made with the same test, so they are comparable.

MC

Log in or register to post comments