Beware of this 'fast' USB 3.0 Flash drive!

I already own a 64GB SanDisk Extreme, but I was looking for a larger 128GB USB 3.0 Flash drive.

As you may know, I am a big fan of the SanDisk Extreme USB 3.0 drives. They are amongst the fastest USB 3.0 Flash memory drives that you can buy, but also represent extremely good value for money.

So, I consulted a few league tables such as UserBenchmark, thewirecutter  and everythingusb.

I was looking for a fast drive, so although the SanDisk Ultra Fit drive was good value for money, I did not consider it because it is slower than several others and also it is just too small - I do have one but I keep 'losing' it, and only find it a few weeks later 'hidden' in a USB port on one of my other devices.

So the main contenders were: (with advertised speed and links to Amazon.co.uk and Amazon.com):

SanDisk Extreme     128GB £48 or $61   245MB/s Rd, 190MB/s Wr
SanDisk Extreme Pro 128GB £67 or $60   260MB/s Rd, 240MB/s Wr 
Lexar P20           128GB £52 or $49   400MB/s Rd, 270MB/s Wr

On paper, the Lexar P20 seemed to fit the bill, so I ordered one of these (as I had some Amazon.com Gift Vouchers, I ordered one from the USA for $66 = £51 inc. postage+import fees).

Lexar P20 128GB

The P20 comes with a thin loop of string that you can attach to a key ring. In my experience however, the string tends to fray in the pocket when used on a key ring and eventually it has to be replaced by wire or thin paracord.

I read some comments from reviewers about the P20 being 'very large', however in fact it turns out to be only 1mm wider and 7mm shorter than the SanDisk Extreme.

The slide mechanism of the P20 had a more 'positive' latch than the SanDisk and also had a half-metal shell with a lovely smooth, weighty 'feel'.

Benchmarks

After using it a while for E2B, I wanted to perform some Crystal DiskMark benchmarks on the SanDisk Extreme 64GB and the Lexar P20 128GB drives in order to compare them, but I found that the results were very inconsistent, especially on the P20. 

Note that you should never perform benchmarks on brand new Flash media. This is because the memory will be 'clean' and writes to the media will initially be much faster. You should ideally fill the whole drive with data before testing it (BootIce - Utilities - Disk Sector Filling is useful for this), so that the flash memory controller will have to perform a read-erase-write cycle on every write operation. Although SSD drives will use the TRIM function, most non-SSD Flash drive controllers do not support TRIM (and the OS does not issue TRIM commands either).

The write results that I obtained from the benchmarks were extremely random for the P20 and could not be relied upon. However, I did notice that the 4K Random Writes times always looked really dire on the P20 (0.009MB/s). This may mean that writing 100's of small files may take a while...

I also noticed that when using the UserBenchmark downloadable system test, the USB data from all their users had a very wide spread of results for the P20 drives too.

This indicated to me that the benchmark tests could not be relied upon, even after I disabled my antivirus software (maybe the USB flash memory controller chip inside the P20 was doing something strange?).

However, I wrote a simple 'copy files' command script to test the large-file copy speed of both USB drives, and this gave consistent results which are shown below:

File: 4,238,336,000 bytes (4.238 Gb) NTFS

                              Lexar P20      SanDisk Extreme

Avg. Read from USB device:  12s (353MB/s)     17s (249MB/s)

Avg. Write to USB device:   16s (265MB/s)     25s (170MB/s)

So for large files, the Lexar P20 is faster than the SanDisk Extreme (as expected).

Copying small files is really slow!

I then tried using 7Zip to extract the files from an XP ISO file (which contains thousands of files) directly onto the two USB drives.

The XP files extracted onto the SanDisk Extreme 64GB drive in 75 seconds.

However, performing the same test on the Lexar P20 took over an incredible 45 minutes!!!

The DPMS version of E2B includes many small XP driver files. Copying all the E2B+DPMS files to the P20 took over 20 minutes (it took only 1 minute on the SanDisk).

Booting to grub4dos and E2B

I noticed that the Lexar took longer to boot to the E2B Main menu than the SanDisk - file enumeration was noticeably slower (using the same USB 3.0 port on the same notebook).

I then used the Utilities menu - Measure BIOS USB speed menu entry to measure the file access speed under grubdos (USB 3.0 port - Lenovo IdeaPad 300) with these results:

SanDisk Extreme 64GB 1000 loops = 18 seconds

Lexar Pro 128GB      1000 loops = 36 seconds

So the Lexar is twice as slow (on read directory accesses)  as the SanDisk. This was a surprising result because all the read benchmarks of the Lexar P20 were better than the SanDisk benchmarks.

I decided to try installing XP (from USB 2.0) and Windows 10 (from USB 3.0) from an ISO using E2B and compare each of the drives:

                               Lexar P20      Sandisk Extreme

XP file copy phase to HDD        330s               240s 

Windows 10 file copy phase       258s               254s

Loading a 590MB ISO (USB 3.0 port) into memory under grub4dos took over 60 seconds on the P20 compared to 30 seconds on a SanDisk Extreme.

Conclusion

The Lexar P20 is certainly faster for large files (both reading and writing) under Windows 10, however the extremely poor small files performance of the Lexar P20 is a real disappointment. Maybe that is why there is not a great deal of difference in the price?

I would NOT recommend the Lexar P20 drive for use with E2B or any other occasion where you want to writes lots of small files to the USB drive! Under grub4dos it is half the speed of the SanDisk Extreme on reads. I did not test it, but I suspect that for WinToGo use, the P20 will be far worse than the SanDisk Extreme too.

P.S. I have tested it now. It took over 1 hour to boot to WinToGo from a Win10 VHD (fresh boot with hardware detection on IdeaPad 300). So definitely no good for WinToGo!

The Kingston HyperX USB Flash drives also gives similar poor 4K Random write results, so I strongly suspect that will also suffer from the same problem.

Whatever USB Flash drive you are considering buying, check the 4K benchmark figures first, the 4K write figure should be around 10MB/s and definitely not less than 1MB/s!

The Voyager GTX is very fast (ideal for WinToGo) but is of the 'fixed disk' type and not the 'Removable' type.

The Patriot Supersonic Magnum 2 is the 'Removable' type and fast, but has a wide body which may block adjacent USB ports. Also, the small file (4K) write speed is quite slow.

I have now tested also the SanDisk Extreme Pro (see here for blog post)

For a general purpose E2B USB drive of reasonable cost and good all-round performance, I still recommend the SanDisk Extreme.