USB Flash Drive Speed

What works, and doesn't, for you. Be specific, and please include Puppy version.
Message
Author
User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

USB Flash Drive Speed

#1 Post by smokey01 »

For some time now I have been looking for a reliable flash drive that is a respectable size at a cost that won't break the bank. I probably have about 100 flash drives but many are the same brand and size as I tend to buy them in batches for giveaways.

Recently I was reading Barry's blog.
http://web.archive.org/web/201609132126 ... iled=00402
I had a few questions about this drive so I bought one from MSY today for $AUD57.00. Using this test method "hdparm -t --direct /dev/sde" it actually performed quite well. The manufacturer claims read speeds of around 245MB/sec and I actually received 196MB/sec, not too bad. I only tested it formatted as vfat, the default. I often format USB flash drives as Ext3 or f2fs.

I assumed the above command "hdparm -t --direct /dev/sde" is showing read speeds and not write speeds. If it's the write speed then that's even more impressive.

The reason I raise this is because it seems USB flash drives read faster formatted as vfat than Ext?
To prove this I tested two 64GB SanDisk Ultrafits, one formatted as vfat, the other as Ext3. The vfat came in at 109MB/sec while the Ext3 could only manage 67.40MB/sec which is quite a difference.

I rather like the little Ultrafits as they are small and only protrude a little from my laptop. One issue I have noticed, they become quite hot.

Recently I was reading an article that stated they can get so hot they can damage the USB port. To get around this problem I bought a 0.5M extension cable. This means the heat is away from the actual port. I'm not sure this will completely solve the problem but sadly I've lost my small footprint. I also discovered one of the main differences between USB2 and USB3 cables is the numbers of connectors. USB2 only has four and can only transmit or send, not both at the same time (half duplex). USB3 has nine connectors and is full duplex, no wonder it's faster. USB3 flash drives and cables are identified by blue fittings. Inside the plugs are blue. You will also notice the cables are thicker as they have more wires to accommodate.

I would be interested in hearing other views on fast USB3 drives and how well they perform.

Is the hdparm command above the best way to test USB drives or is there something better?
Last edited by smokey01 on Mon 06 Jul 2020, 11:47, edited 1 time in total.

User avatar
drunkjedi
Posts: 882
Joined: Mon 25 May 2015, 02:50

#2 Post by drunkjedi »

Hi smokey, use F3 utility,
you can find it on our forum.
It works like h2wtest for Windows.

It will give you read and right speed and also test your drive for capacity and other defects.

I use it to check newly purchased drives.

And manufacturer quoted speed is never achieved by me.

LateAdopter
Posts: 361
Joined: Fri 27 May 2011, 17:21
Location: Reading UK

#3 Post by LateAdopter »

Hello smokey01

The power-off data retention time for MLC flash can be quite short, according to Intel data, if they are stored at a higher temperature than they were written. This is unlikely to be significant in the UK but could be a factor for you and BarryK in Australia.

MLC flash is used in the highest speed USB sticks and in consumer SSDs.

I use a Transcend Premium 300x microSD card, which is not MLC in a Transcend USB3 card reader. The command that BarryK used gives a read speed of 55MB/s when formated FAT32 with 64kB clusters. The write speed is about 25MB/s when copying a 2GB MPEG file.

The same card gave a read speed of 90MB/s in a review on a chinese website, so I suspect that the linux test is not fully sequential,

TyroBGinner
Posts: 236
Joined: Wed 30 Mar 2016, 20:18

#4 Post by TyroBGinner »

This is a subject that interested me quite a bit at one time. It has been discussed in this forum, but not too recently, I think.

I swear that I have seen, surprisingly, what appear to be 3.0 connections without the characteristic blue color of the plastic inner tab. Has anyone else encountered this? I found such a case, for example, on a new Dell Optiplex. The commands lsusb -t and lsusb -v reported a 3.0 connection and I read some 1 gig files off a SanDisk Ultra drive at about 100 MB/s. I don't think that the hdparm command gives a good sense of the transfer speeds. I prefer direct file transfers - and keep in mind that write caching can skew results.

The SanDisk Extreme drives appear to be well-received and at good prices. A few months ago, they were less than a dollar a gig. Forum member April mentioned some USB SSD flash drives at $40. Which brings up what I think is the main issue - the use of SSD-style controllers in the truly fast drives. The type of flash memory is important, and so is the number of dies, and so on...but the controller is the biggest factor. For example, Mushkin has put out SSD-type flash drives and are very fast, but I saw reports of rapid overheating (and resulting automatic throttling) of such devices on sites like Anandtech. This seems to be in keeping with what Smokey said. It seems suprising that it would get hot enough to damage the receptacle on the computer. Maybe the heat and price issues will improve with time. Just look around at review sites (Anandtech is a good example) and who is selling what products.

A old review, but a good starting point, is here.

Also try http://usb.userbenchmark.com/.

People here often mention reformatting flash drives and changing partition structure. There is reason to believe that this can alter the speed of the device. It has the potential to change boundary alignment between the address space of the drive and the page boundaries of the physical flash memory. A utility, called flashbench, was developed to look at read/write timing of flash devices to help reveal where to place partitions on those devices. It might also be the case that the details of a particular filesystem, such as cluster and sector sizes, are important here as well. Writing to a region on the drive could require excessive erase operations, for example. The whole subject is pretty tricky, and I don't claim to have a solid understanding of it, but it is worth being aware of the issue.

These articles are old also, but interesting: here and here. The second item has a comment by Joe Zbiciak, developer of the jzintv emulator for Intellivision games - it's a small world.

Please note: If you want to use the flashbench program, it will erase data structures, such as file tables, on your flash drive!!! Use the program on blank devices.

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#5 Post by smokey01 »

@drunkjedi
I remember looking at f3 in the past. Is there a way to tell it to write a single 1024M file instead of the entire available space?

It could take a while to test a 64GB drive. I think writing and reading a 1GB would give suffice to test the speed.

@TyroBGinner
Some interesting links, thanks.

@LateAdopter
Thanks

ASD

#6 Post by ASD »

Hi smokey01

Re: f3 - I think you can kill writing using Ctrl+c.

User avatar
drunkjedi
Posts: 882
Joined: Mon 25 May 2015, 02:50

#7 Post by drunkjedi »

Yes you can kill f3write from writing and f3read will test whatever is available but I never tried that as I mostly use it to test newly bought drives, I let it run for whole drive.

Edit: I have never used f3probe, f3fix, and f3brew

You can read about them here...
http://oss.digirati.com.br/f3/
Last edited by drunkjedi on Mon 12 Sep 2016, 12:31, edited 1 time in total.

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#8 Post by smokey01 »

It appears it's possible to write file but I having trouble working out the help.

Usage: f3write [--start-at=NUM] [--end-at=NUM] <PATH>

f3write --start-at=0 --end-at=1024000000 /mnt/sde1 :oops:

Doesn't work.
Last edited by smokey01 on Mon 12 Sep 2016, 13:35, edited 1 time in total.

User avatar
drunkjedi
Posts: 882
Joined: Mon 25 May 2015, 02:50

#9 Post by drunkjedi »

try f3write --end-at=2

This should write 2 files only.

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#10 Post by smokey01 »

I guess that would write 2 x 1 GB files.

f3write --end-at=1

Would only write 1 x 1GB file.

User avatar
drunkjedi
Posts: 882
Joined: Mon 25 May 2015, 02:50

#11 Post by drunkjedi »

Yes I believe.

I can't test as I am at work.

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#12 Post by smokey01 »

It worked. I guess we have it sussed.

User avatar
drunkjedi
Posts: 882
Joined: Mon 25 May 2015, 02:50

#13 Post by drunkjedi »

Nice, so what write and read speed you got?

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#14 Post by smokey01 »

I will have to bootup the lappy as this old clunker only has USB2.

I will report back soon.

I tried to write one file but it wouldn't work so I CTRL-C after 7 files and took notice of the speed. Write was 133.80 and Read was 182.87.

This was for a SandDisk Extreme 64GB.

Not too bad.
Attachments
f3.tar.gz
These were built in Fatdog64-710b
(11.28 KiB) Downloaded 397 times

User avatar
mavrothal
Posts: 3096
Joined: Mon 24 Aug 2009, 18:23

#15 Post by mavrothal »

If you want a stick to house an OS the random read/write speed and the number of simultaneous reads/writes are the crucial factors.
Flashbench is a tool that may be useful in this case. Has also some good documentation and links (some of them mentioned above by TyroBGinner)

BTW the 64GB SanDisk Extreme is an excellent stick. Booting/running an OS is consistently at 30-40% of the performance of an internal high quality SSD
== [url=http://www.catb.org/esr/faqs/smart-questions.html]Here is how to solve your[/url] [url=https://www.chiark.greenend.org.uk/~sgtatham/bugs.html]Linux problems fast[/url] ==

gcmartin

#16 Post by gcmartin »

Before @BarryK highlighted the hdparm command, I had always tested using the time command as a wrapper to the command to read or write data.

Code: Select all

# time cp  write-RAM-file    to_some_partition
# time cp  read_some_partition-file    to_RAM-file
I would compare the outputs for various drives to gain a perspective on system I/O performance and in most cases, where I should place my SWAP partition.

I have about a dozen+ of both USB2 and USB3 sticks with a few SD/microSD/microSDSX cards. Additional I have 5 USB enclosures which are USB2 and a couple USB3 for ATA and SATA drives.

At the moment, I only have desktops which have USB2 connections and a NUC which has USB3s plus 1 USB2.

I find that USB sticks are easy for transporting information and was the primary reason I began acquiring them.

I rarely have used USBs for Frugal use as I have found many years ago, that there is NO system performance difference whether I boot from DVD, USB or HDD/SSDs. The systems run in RAM and as is the case with many in the forum, systems are 2GB+. Lastly for my production use I rarely shut those PC down and when I do, I save to the DVD. So the ONLY benefit to me is transporting and this is where my DVDs and my USBs benefit me.

There is a lot of fondness for USBs. And, I see users benefiting.

So, I wonder if there is benefit if we adopted a standard measurement utility/approach for reporting and measuring USB behavior on PUPs.

This thread can be useful if it is leading to some measurement tool which can be used to present a consistent picture. And a consistent command use, as well.

Here is one report I just did 2 days ago for one of my systems that I booted Slacko7a3; data obtained using hdparm command:

Code: Select all

Linux Slacko7a3-onSLI 4.1.32 #1 SMP Sat Sep 10 22:20:07 AEST 2016 x86_64 AMD Athlon(tm) 64 X2 Dual Core Processor 4600+ AuthenticAMD GNU/Linux

40GB SATA MSDOS
/dev/sda:
Timing buffered disk reads:  153 MB in 3.01 seconds = 51946 kB/s

1.5TB SATA MSDOS
/dev/sdb:
Timing buffered disk reads:  294 MB in 3.00 seconds = 100263 kB/s

2GB USB GPT STICK
/dev/sdc:
Timing buffered disk reads:   54 MB in 3.03 seconds = 18230 kB/s

10GB USB MSDOS ATA
/dev/sdd:
Timing buffered disk reads:   75 MB in 3.04 seconds = 25197 kB/s
This is just a manual representation of something I was looking at back then.

Although I have never tested, I think @Smokey is right and there very well may be some differences in how the sticks/cards/drives are setup. The 3 primary issues are manufacturer and unit tested, the disk layout types (namely GPT or MSDOS), and the partition filesystem that is present. Now, which tool would be best to expose performance?

I dont believe hdparm commands cares and I am not sure how it actually is obtaining his data. What exactly is he reading? (of course this is not the primary question in determining performance)

Lastly, in looking over this site, it appears that Lexar's Jump has it over Sandisk Extreme. But both seemingly are best in what they provide.

Just a few ideas. FYI

User avatar
smokey01
Posts: 2813
Joined: Sat 30 Dec 2006, 23:15
Location: South Australia :-(
Contact:

#17 Post by smokey01 »

I think the f3 utility does a pretty decent job of determining write and read speed. It writes files in 1GB sizes which you can read back. By default it will fill the avaiable space on the drive with these files. Large drives can take some time to test. I think this will be my benchmark for the near future.

anikin
Posts: 994
Joined: Thu 10 May 2012, 06:16

#18 Post by anikin »

Never mind.
Last edited by anikin on Wed 14 Sep 2016, 18:37, edited 1 time in total.

User avatar
BarryK
Puppy Master
Posts: 9392
Joined: Mon 09 May 2005, 09:23
Location: Perth, Western Australia
Contact:

#19 Post by BarryK »

TyroBGinner wrote:I swear that I have seen, surprisingly, what appear to be 3.0 connections without the characteristic blue color of the plastic inner tab. Has anyone else encountered this?
Yes, you are right.

I have an Asus baby laptop that I purchased a few months ago. It has one USB3 socket, and the plastic inside it is black.

I asked the guy in the shop about this, and he said, yes, some laptops do not have blue in the USB3 sockets.
He pointed out that it has lettering beside it, "high speed" I think it said, something like that. Also the manufacturer's specs identified it as USB3.
[url]https://bkhome.org/news/[/url]

User avatar
BarryK
Puppy Master
Posts: 9392
Joined: Mon 09 May 2005, 09:23
Location: Perth, Western Australia
Contact:

Re: USB Flash Drive Speed

#20 Post by BarryK »

smokey01 wrote:Recently I was reading Barry's blog.
http://barryk.org/news/?viewDetailed=00402

I had a few questions about this drive so I bought one from MSY today for $AUD57.00.
Hey, you beat me to it!
I ordered mine from the Eastern States, and it is still on its way. I think that they have sent it surface mail, that's why the postage was so cheap.

Currently I am using my Emtec 64GB USB3 stick for building Quicky using woofQ and for compiling packages, such as SeaMonkey, on my Pi2.

Apart from the fact it is like watching grass grow, and I wonder how long this stick is going to last, getting flogged like this.

There was another post about a 128GB stick for AU$40. Yeah, that was a sale at Australia Post. Ended now. I bought my Emtect 64GB stick for AU$38 from Officeworks.

So I reckon the Sandisk Extreme for around the $50 mark is an excellent buy, considering the speed improvement.

Note, I was using a Lexar 8GB stick as a swap partition. It is near-new. Anyway, something has gone wrong with it, it now flashes continuously. Will do a reformat, see if that fixes it.
[url]https://bkhome.org/news/[/url]

Post Reply