03 January, 2015

Transcend® JetFlash®710 pico-review

I bought a 64G Transcend® JetFlash®710 drive from eBay recently as my previous "tiny keyring drive" (A Sandisk Cruzer Force) had passed-away due to some kind of mechanical failure (For once not the loop that secures it to your keychain/lanyard, but the plastic holding the flash-module inside the shell).
In the past you always had to be wary of flash storage bought through eBay (The brand-name forgeries get more and more convincing as time goes on). So this article is covers my analysis of what I received.

TL;DR: Looks genuine, but the transfer speed doesn't really benefit from the USB3 interface (Which is not unusual for most USB3 thumb drives today). It's still faster than any other USB flash drive I have.

I bought this model of flash drive as it looked like a pretty robust enclosure but I'm not sure about the innards (which may fail like the Sandisk drive I mentioned before). It also has a USB3 interface, so I was looking forward to the increased access speed.
To this end, I want to verify as accurately as possible whether this is a genuine item, that it has the stated capacity and determine what its read/write speeds might me.

The tests

Write speed

I generated a pseudo-random data file of about 50G using the following command:
# dd if=/dev/urandom of=data.bin bs=1024000 count=50000

I then copied this to the drive using rsync (So I could get a running transfer-speed display). It surged at the start (Presumably being cached... somewhere?) - in the end it averaged 27.83MB/s (Which is in the 20-30MB/s range suggested for this model):
# rsync --progress --inplace --partial data.bin /mnt
51200000000 100% 27.83MB/s 0:29:14 (xfer#1, to-check=0/1)

sent 51206250070 bytes received 31 bytes 29185665.49 bytes/sec
total size is 51200000000 speedup is 1.00

Capacity + read speed

A lot of the forged thumb drives you saw years ago would present themselves to the host OS as their stated capacity, but were actually populated with much less flash ram. The net result was that the knock-offs could hold some data, but if you wrote more than the populated flash, your files would get corrupted as the address lines wrapped-around. So the main "Is it real or is is it Memorex?" test is to read the written data and see if it matches the original. I thought I'd use dd to do the data-pumping so I could simultaneously test the read-speed (Which will give a much better indication of the USB3 utilisation). Ultimately I piped the data through md5sum and compared the result with the md5sum of the original file:

# dd if=/mnt/data.bin bs=10240 |md5sum
5000000+0 records in
5000000+0 records out
51200000000 bytes (51 GB) copied, 1175.85 s, 43.5 MB/s
089cb596737b0a7be21ea96a489719f4  -

# md5sum data.bin
089cb596737b0a7be21ea96a489719f4  data.bin

Compared to other Flash and/or USB3 drives

So that looks good: Holds ~50G well enough and can read at 43.5 MB/s.
I thought I'd compare the write-speeds of some other USB3 media just for kicks, so I made a 1G random file like the 50G one above (So I could run the tests a bit more quickly and used much smaller drives) and wrote these to the various drives. I used dd to do the transfer as rsync would quickly dump the data into a cache somewhere and then just sit in a zombie state for a minute or so until synced. Using dd means I get a complete transfer rate from start to flush. Here's a class-2 MicroSD card connected through a USB3 card reader.
# dd if=data2.bin of=/mnt/data2.bin bs=10240
100000+0 records in
100000+0 records out
1024000000 bytes (1.0 GB) copied, 157.203 s, 6.5 MB/s
Seems pretty reasonable. Here's a generic USB3 flash drive purchased through Kogan:

# dd if=data2.bin of=/mnt/data2.bin bs=10240
100000+0 records in
100000+0 records out
1024000000 bytes (1.0 GB) copied, 119.611 s, 8.6 MB/s
Slightly faster than the SD card, but not by much and significantly slower than the Transcend drive. Here's a 2G portable WD Elements drive I bought a couple of years ago:

# dd if=data2.bin of=/mnt/data2.bin bs=10240
100000+0 records in
100000+0 records out
1024000000 bytes (1.0 GB) copied, 18.8853 s, 54.2 MB/s
Faster than anything else (even the Transcend read speed). But this is not surprising as this drive has much more involved cache, it's a single sequential write and the 50:1 file size difference is probably a magnifying factor in all that.


All in all, this is a fine thumb drive, it's not expensive, it has decent capacity and seems pretty rugged. Nevertheless the USB3 interface is not utilised well (Which is the same for most USB3 flash drives today), so it'd be nice to see improvements in this area.

No comments:

Post a Comment