USB vs SATA II transfer rate

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

A20_user

Distinguished
Sep 30, 2008
1
0
18,510
Regarding the maximum transfer rate on USB2.0:
The fastest read rates I have seen so far are in the range of 29 to 33 MByte/s from Harddisks or Memory Sticks. But ... read ...

People are talking about USB2.0 protocol limits and overhead. Does anybody know what is really going on here?

At 33 MByte/s, the USB2.0 bus is clearly NOT saturated, even accounting for protocol overhead.

I connected two Mass Storage Devices (Memory Stick or HDD) to the same port via a Hub, both reading at 31 MByte/s alone. Then I tested reading from both in parallel, everyone giving approx 22, the sum was 43.8 MByte/s.

I added a third Mass Storage Device (a little slower) and got a sum of 47.0 MByte/s.

That test clearly shows that the USB2.0 resources are not completely used by 1 device alone. The results are quite the same for Windows (XP, Vista) and Linux. I still do not see the point in USB protocol which causes this limit for a single device.
 
G

Guest

Guest
It really sounds like you have an involuntary (probably sexual) attraction to this guy, you want him to be the bottom to your improbable top. There you are the wee guy with the small stick and no CS degree ( which obviously makes you feel inferior as per the thread contents) sliding it in slow to his welcoming brown door. Trawling the web looking for evidence that you are right, so concerned about what mutual friends will say/think.

Why dont you just get past the innuendo and kiss the guy, mutter slowly into his soft eyes about your irrestible need to be one with his sphincter, and you guys can lie back exhausted later and play with a macbook pro as your mutual friends sms you incessantly with "SO HOW WAS IT?"

that's the truth as I see it, sorry for being honest.

 

coredatarecovery

Distinguished
Mar 14, 2009
1
0
18,510
I'm running a real world test using linux to give you guys some numbers.

I'm doing a sector for sector copy of a 1 Terabyte SATA II western digital drive to another 1 Terabyte SATA II Western digital thru usb.

I'll post the results on my blog.



For the record I'm using a quad core Intel chip, Biostar board, Thermaltake USB docks (2 of them)

Right now it looks like drive to drive copy using:
ddrescue -n /dev/sdb /dev/sdc

it looks like it's going to take about 16 hours of transfer time (estimated at this point) to copy between the drives.

I'm going to try 2 different platforms and drives and give you guys some real world numbers here.

I use SATA II External Thermaltake drives for data transfer of striped array members for their ease of use and hook them up with ESATA as a rule.

It will be interesting to see the performance difference for the copy, of course copying from drive to drive amplifies difference in speed.

(My blog is at http://www.coredatarecovery.com)



 
G

Guest

Guest
Regardless of what anyone says, eSATA is extremely quicker in transfer rates than USB or Firewire.

I got fed up with trying to back up my 300GB hard drive through a USB port because it took 1 hour and 25 minutes. I bought an eSATA capable Western Digital Backup Drive and when I did the exact same back up through the eSATA connection it took 28 minutes!!! That's nearly 66% faster.
 

linkolng35

Distinguished
Sep 26, 2009
2
0
18,510


USB 1.1 = Slowww
USB 2.0 = Faster, but still not THAT fast
IDE = Rated slower than Firewire
SATA = WAY Faster than IDE
SATA II = Even Faster still
eSATA = faster than USB2.0 or IDE

Why they don't include a power source for eSATA ports I don't know...
 

linkolng35

Distinguished
Sep 26, 2009
2
0
18,510



What allocation table did you use on the drives? That could make a significant difference since FAT32 and NTFS have much different cluster allocation sizes (so a text document with 1 letter in FAT32 might take up 64kb but only 4kb in NTFS). Granted this shouldn't affect the transfer speed, but if you are transferring files the files might take longer to transfer to a FAT32 formatted drive than an NTFS formatted drive.

Food for thought...
 

keith_leitch

Distinguished
Jan 5, 2010
2
0
18,510
I see this hasn't been revisited for awhile, so I hope somebody out there is watching. I seem to be getting, at best, 10MB/s transfer speeds from one internal SATA drive to another (both running 7200 rpm). I have tried a backup program, 7-zip, and just straight Windows copying. This is creating 10-hour sessions to back up 350 GB, and seems ridiculously slow compared to what you have written here. I have checked that write cacheing is enabled, and that the drives are in Ultra DMA 6. Can anyone suggest something I may be missing?

As an example: I just copied about 7 Gig from one 7200rpm internal SATA drive to another. The total time for the copy was 9 minutes and 29 seconds; this makes the transfer rate 12.75 MB/S (which is fast compared to what I have been getting).
 

keith_leitch

Distinguished
Jan 5, 2010
2
0
18,510
It's a bit early yet to be sure, but I found and ran the VBScript duplicated below. The author claims that Windows sometimes deactivated UltraDMA mode due to write errors, but that this deactivation may not show in system properties. After running this utility and rebooting, I copied a 6.2 GB file between the same two drives as before in 97 seconds. This is 63 MB/second, which is actually slightly quicker than the averages I have been researching.

I will see later if and how it affects the restore of the 350 GB copy I performed.

If this works, I intend to take out the dialog boxes and simply run this script from the batch file that HIP invokes upon suspend. It takes less than a second to run it.




' Visual Basic Script program to reset the DMA status of all ATA drives

' Copyright © 2006 Hans-Georg Michna

' Version 2007-04-04

' Works in Windows XP, probably also in Windows 2000 and NT.
' Does no harm if Windows version is incompatible.

If MsgBox("This program will now reset the DMA status of all ATA drives with Windows drivers." _
& vbNewline & "Windows will redetect the status after the next reboot, therefore this procedure" _
& vbNewline & "should be harmless.", _
vbOkCancel, "Program start message") _
= vbOk Then

RegPath = "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E96A-E325-11CE-BFC1-08002BE10318}\"
ValueName1Master = "MasterIdDataChecksum"
ValueName1Slave = "SlaveIdDataChecksum"
ValueName2Master = "UserMasterDeviceTimingModeAllowed"
ValueName2Slave = "UserSlaveDeviceTimingModeAllowed"
ValueName3 = "ResetErrorCountersOnSuccess"
MessageText = "The following ATA channels have been reset:"
MessageTextLen0 = Len(MessageText)
ConsecutiveMisses = 0
Set WshShell = WScript.CreateObject("WScript.Shell")

For i = 0 to 999
RegSubPath = Right("000" & i, 4) & "\"

' Master

Err.Clear
On Error Resume Next
WshShell.RegRead RegPath & RegSubPath & ValueName1Master
errMaster = Err.Number
On Error Goto 0
If errMaster = 0 Then
On Error Resume Next
WshShell.RegDelete RegPath & RegSubPath & ValueName1Master
WshShell.RegDelete RegPath & RegSubPath & ValueName2Master
On Error Goto 0
MessageText = MessageText & vbNewLine & "Master"
End If

' Slave

Err.Clear
On Error Resume Next
WshShell.RegRead RegPath & RegSubPath & ValueName1Slave
errSlave = Err.Number
On Error Goto 0
If errSlave = 0 Then
On Error Resume Next
WshShell.RegDelete RegPath & RegSubPath & ValueName1Slave
WshShell.RegDelete RegPath & RegSubPath & ValueName2Slave
On Error Goto 0
If errMaster = 0 Then
MessageText = MessageText & " and "
Else
MessageText = MessageText & vbNewLine
End If
MessageText = MessageText & "Slave"
End If

If errMaster = 0 Or errSlave = 0 Then
On Error Resume Next
WshShell.RegWrite RegPath & RegSubPath & ValueName3, 1, "REG_DWORD"
On Error Goto 0
ChannelName = "unnamed channel " & Left(RegSubPath, 4)
On Error Resume Next
ChannelName = WshShell.RegRead(RegPath & RegSubPath & "DriverDesc")
On Error Goto 0
MessageText = MessageText & " of " & ChannelName & ";"
ConsecutiveMisses = 0
Else
ConsecutiveMisses = ConsecutiveMisses + 1
If ConsecutiveMisses >= 32 Then Exit For ' Don't search unnecessarily long.
End If
Next ' i

If Len(MessageText) <= MessageTextLen0 Then
MessageText = "No resettable ATA channels with Windows drivers found. Nothing changed."
Else
MessageText = MessageText & vbNewline _
& "Please reboot now to reset and redetect the DMA status."
End If

MsgBox MessageText, vbOkOnly, "Program finished normally"

End If ' MsgBox(...) = vbOk

' End of Visual Basic Script program
 

drwatson

Distinguished
Jul 14, 2010
1
0
18,510
Thanks a lot !

after running the vbs, before running the script 4.2 Gb data copy took aprx. 10 Min and after appx 1 min !!

I've used it on a windows home server (w2k3 business edition)

Rogier
 

BlueGazoo

Distinguished
Jul 15, 2010
1
0
18,510
This is a nice thread. I had the same issue as the guy that started this thread, and it was difficult to find good benchmarking to convince people what should be bought.

So... in 2010, which Hard Drive transfers data the fastest? And is there anything new coming down the road?

I'm about to build a new comp and am putting together my list of parts. What hard drive should i get?
 

puppiesfromheaven

Distinguished
Aug 5, 2010
1
0
18,510
Whew!!!!!!!! can't thank enough everyone who responded, your time, commitments to helping, knowledge, I appreciate like you'll never know, whichever connections on each pc, I'll use the fastest methods in accordance with your help responses. God bless all of you, best wishes for cont'd success (55 yr old mom)


(working on a legal deadline for aug 10 2010, 2 towers, 1 laptop, desktop & wand scanners, internet research, countless law books, rules, requirements, legal software, anything to speed up accomplishing successfully, am in federal appeal, pro-se, doing order from Amazon file transfer speeds are my present needs, found you via my google search, in awe all of you, education, experience, thank you! don't have time to edit my typos & grammar, back to work, had done benchmark reviews yesterday, cpu speeds, cpus overheating/coolest...)
 

deepthinker22

Distinguished
Aug 10, 2010
3
0
18,510

lol, which is why I get so upset when people think that CE is the same as CS, and more so when somebody confuses me as an EE for a CS.
 

Patxi

Distinguished
Aug 25, 2010
2
0
18,510
Regardless of what anyone says, eSATA is extremely quicker in transfer rates than USB or Firewire.

I got fed up with trying to back up my 300GB hard drive through a USB port because it took 1 hour and 25 minutes. I bought an eSATA capable Western Digital Backup Drive and when I did the exact same back up through the eSATA connection it took 28 minutes!!! That's nearly 66% faster.
I see why you signed in as "Anonymous"!
You don't need sata connections ... What you really need is ... a "biological CPU" upgrade!
So from 1 1/2 hours to 1/2 hour is ... "nearly" 66% faster? What about 3 times or "nearly" 300% faster?

I used to have this misconception that computer nerds were good at math.
But I have been "cured" for a few years now!

And pardon me for rounding your quoted times to 1/2 hours. I am not Boole either :sol: :)
 

tanco

Distinguished
Aug 30, 2010
1
0
18,510
I think everyone who reads this will know 'exactly' who needs the 'upgrade' and who is 'not' good at math. Do your sums before 1. being sarcastic, and 2. trying to seem superior.

30mins with an increase of 100% = 60mins

so you think that 30mins with an increase of 300% is what exactly ??
 
G

Guest

Guest
When comparing data typically we assign them order based on chronology. In this case the first test took 1.5 hours. That's what we will call 100%. We then consider the second (and subsequent) tests(s). These will be assigned percentages based on their proportional improvement. Thus: Since 1/2 hour to complete the task is 3 times faster than the original speed, the task is completed at a rate of 300%. 45 minutes would be a rate of 200% and so on.

Although you are correct in noting that the second test observes an increase of 200%, we generally look past a slightly poor choice of words in order to avoid "2. trying to seem superior."

All the best.


 
G

Guest

Guest
HD Tune Pro is a program for measuring the performance of hard drives.

I downloaded a free copy (it's free for about 2 wks) and measured some external Western Digital hard drives.

These two WD drives were purchased as "internal drives" then mounted in an enclosure that contains a power supply and has eSATA connectors on the back to connect up the drives.

see: http://www.cooldrives.com/index.php/midudrnasaen.html

Adding a passive card slot bracket device that just has two eSATA connectors on it that goes to two standard internal SATA cables you plug into available SATA ports on your motherboard, I installed the two drives so that they are seen by the computer as just two more internal drives.

So now the BIOS sees three drives:

The original (boot) drive: Samsung HD160JJ/P
External Drive 1: WDC WD1001FALS-00J7B
External Drive 2: WDC3200AAKS-00G3A

Using HD Tune Pro and the "fast test" options) the results are as follows:

The original (boot) drive: Samsung HD160JJ/P (160 GB)
47.7 MB/s

External Drive 1: WDC WD1001FALS-00J7B (1TB Caviar Black)
89.0 MB/s

External Drive 2: WDC3200AAKS-00G3A (320GB Caviar Blue)
97.5 MB/s

I also have a USB Vantec drive adapter but I haven't tried using it for comparison yet.

 

sobank

Distinguished
Nov 19, 2004
63
0
18,630
Well just to add in my real time experience in last two days.

asus p5kc mother board with a sata connectors 7200.11 seagate 500gb drive and a wd20ears 2TB drive (external western digital green drive with esata/usb connection)

http://www.wdc.com/en/products/products.asp?driveid=773
http://www.seagate.com/ww/v/index.jsp?vgnextoid=c89ef141e7f43110VgnVCM100000f5ee0a0aRCRD

So from using usb transfer, I picked up 20mb/s+ while using little bit of pc. (156 gb transfer)
and as i surf tom's forum, I am getting 76mb/s transfer from esata using same drives and settings. (120 gb transfer)

As you will notice, result is not from small transfers.

Now usb vs esata is a major difference if you ask me. Are you sure he was talking about the transfer speed and not disk rev?
 

Patxi

Distinguished
Aug 25, 2010
2
0
18,510



Aaah ... You are funny!
I (really) wasn't trying to "seem" superior when I wrote the original message.
Nor did I dream that I would keep laughing about it so many weeks later.

Without trying to be mean: before we only knew that "tanco" was not very good at math (which in any case is not a sin).
Now we all know that he will most probably never learn!
One will think that after being called to order one will check twice before exposing himself to the "public" again...
Instead he had a belligerent and thoroughly (again) erroneous rebuttal so it is clear that not everything is bad: at least we know tanco is a very daring lad.

I'll explain it so no one needs to upgrade their pretty little CPUs:

1.- when a NFL runner runs the same distance as his colleague in 1/3 of the time in English ones says that the quarterback run three times faster.

2.- if you grandma drives 50 miles in 1 + 1/2 hours (thus at 33.33 miles per hour) and then you (daring devil) drive the same distance in 1/2 hour (thus at 100 miles per hour) then you drove 3 times faster than grandma (but is no biggie ... home we drive much faster but in kilometers).

Please do not read this message as an attack to tanco. It is not.
It is instead meant to have a laugh (a bit at his expense... Sorry!) and to defend the quality of all languages: linguistic or mathematical.

We should all be a bit more careful when expressing ourselves... If users like "tanco" keep assaulting the languages we have and further insisting on their assault in no time we will simply not understand each other in English, math or any other language. When I read his original comment I had no idea of how fast he thought the drive was!

The greatness of the Internet is that anyone can write in it.
The trouble with the Internet is that anyone can write in it: It is indeed very democratic but it is reducing everything to the smallest common denominator...

Have a great one. Everyone! Boolean or Not!
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
Actually 3Gb per second... sound so fast... do the maths
1024*1024*1024*3 bits per sec, divide by 8 to get bytes per second divide that by 1024*1024 to get Megabytes per sec so you
get 1024*1024*1024*3
--------------------------
8*1024*1024

some of the 1024's cancel to give 1024*3/8 = 384 MegaBytes per second...

its supposedly faster than IDE at 100 MegaBytes per second, though I am not sure how less pins can
actually be any faster as surely there is some sort of "packet" shuffling lag, maybe its 384 Megabytes from there to the "packet-shuffling" chip that will be integrated into it...

I suppose the chip controllers for the new SATA drives may respond quicker or run at a higher frequency than the older IDE hard drives, but that would suggest it was a case of upgrading the chips on newer IDE
drives to far outstrip this latest rival...

in short... I don't know either, I do know I have only ever had one SATA drive and it is now uselesss having gone the way of Casper of some mysterious ailment or other, so Im not in the best of moods with it, and when installed it seemed to run slower than the currently installed 40Gb IDE drive (just a temporary until I get a larger one)... and I mean slower, like a reboot mighht hitch the machine for 20 odd minutes, whereas now its less than 1 minute... obviously software must of played some role in that, but
even afetr removing "heavy" anti-viruses and anything else starting at boot time with the SATA it would take 10 minutes to boot, whereas like I said on the IDE drive it takes less than a minute... to add insult to injury for SATA, taht SATA was 160Gb, some 4 and a half times the capacity of this IDE... go figure that then?

Personally I'd be dubious of the numbers ANYWAY... remember they are trying to out do rivals and sell you a product, probably cheaper in some way to produce than its chief rivals, maybe thats the feature not the
actual performance.
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
I do remember some years ago that it was discussed that RAID was laggy compared to IDE, but can't think where... and IDE was still the far superior system
for internal drives due to its larger and more direct interface to the motherboard...
but that was a while ago like I said...
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
Also I'd be dubious of anyone and I mean ANYONE who says SATA runs cooler, the one I was forced to remove was HOT... at around well I'd guess 40 or 50 degrees Celcius... now I don't know much but I DO know that I've NEVER experience that with IDE... the one now installed is nowhere near the temperature, even in exactly the same place in the machine.

I do know that even in my humble real-world experiences, or perhaps because of them, that when I see what is said about SATA and compare it to what I have experienced their seems to be more than just read/write/ redundancy /latency numbers that don't add up... there seems to be to me at least a whole world of absolute BS around them...
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
Oh as for EM interference... less pins, more data... hmmmhigher frequency then... that means greater rate of change of electric field... correct my schoolboy A level physics if its wrong, after all its been some years... but doesn;t that lead to GREATER electromagnetic fields...

One bonus is the thin power leads have less internal resistance (moderately) so reduce draw on power supplies... however the sh*tty little plastic connectors are easier to damage as you can well imagine and far harder to then rectify...

for those wondering at their own personal findings of 34 or 43Mb/sec, they do know there is a jumper you have to set to achieve the maximum speed of the drive, quite often this is not done by the manufacturers of pre-built machines and I suspect its not just omitted for fun... be dubious about setting the jumper, I ain't gonna say the system won't cope with it, maybe it will, but if it couldn;t then you'd have a serious problem...

As for connectors... I have ranted elsewhere on petty pretty little plastic connectors proliferating every corner and recess of our manly domain, at least if the IDE power pins splayed you could do the many thing and squeeze em in with a screwdriver...

Now 'cos of the design you'd probably have to cut one off somewhere else and solder it on with a lot of patience and equal amount of carefully wrapping the individual wires with insulating tape (crude but effective)



OR you'd have to remove it inside the power supply and solder on a new one there.

OR you'd have to buy a new power supply...

what a jip! the old IDE connectors you could actually yank pins from other connectors and force them into replace other ones and all sorts... its a right bloody con I tell ya...

Same with mobile phones... what was wrong with those single slide-in pin power supplies... ya know the ones where you got variable sizes, looked not unlike speaker sockets... but you could get five different sized chargers on on mains fed charger... often with variable ampage and voltage outputs at the click of a switch and all sorts... why now do we endure the easily damaged pretty crap often unique to the phone brand and make...

The protest march begins: Bring back MANLY connectors for MANLY things... sorry off on one.
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
Actually one last note on comments on SATA over IDE, the thinner power leads do as they say have lower internal resistance, however unless the voltage is supstantially lower... well we all know what voltage does over thinner wires don't we... well we should, phyyiscs would tell you that its gonna result in greater heat output over the surface area of the individual wires... you only have to look at a an old fashioned filament light bulb or understand the principles of how a fuse blows when it draws greater ampage through the fine wire and burns out to get the idea to the full extent... in short any claim that it results in cooler leads is an absolute out and out crock... it doesn't even take a genius to see it, just a little time to actual think about it....

so you have dainty plastic connectors and a higher temperature... I'd imagine that slow expansion of said connectors will NOT be an uncommon fault in years to come...
 

sdhobbs

Distinguished
Nov 22, 2010
15
0
18,510
Sorry should use the proper term, you have pretty thermo-plastic connectors... ie they will be the plastic that gets soft as temperature increases (the common one) as opposed to thermo-setting-plastics which may handle it ok...

failing that he says somewhat sarcastically, cos yes I can be... may I suggest bakelite! chuckle.
 
Status
Not open for further replies.