USB Tester

[jsteele]

Does anyone know of a free USB testing tool? I'm occasionally seeing the error message that "this device would perform better on a USB 2.0 port ..." If I remove and replug the device I do not get the message again. I'd like to test the system to see if it is really running at USB 2 or not

[christopher_wolf]

Did you try this, first?

Otherwise, all the true USB 2.0 testers I know of are hardware based and rather more expensive than their general utility value for a one time operation. PassMark has some excellent kit, $49 for the hardware loopback (which can be found [url=http://www.passmark.com/products/usb2loopback.htm]here) plus additional software. I, however, would only get something like this if I was testing computers in a lab or office area, or was in the habit of building/selling many PC systems myself for a rough QA gauge.

Software-only testers can really only check to see if the port is responding correctly from the POV of the computer as well as check to see if everything is A'OK with the drivers for USB 2.0. Also, there is the, crude, test of plugging in a Flash USB 2.0 stick and then moving over a file of known dimensions and timing it; however, that depends on how many things are using the bus that the USB 2.0 stick is using at the time (which will slow things down) and include various devices such as Bluetooth adapters, fingerprint readers, etc. So it really won't give you accurate info on the true transfer rate of the USB port in question. Of course, you can download the software for free, but it won't quite give back results unless the USB2 loopback hardware peripheral is plugged into a USB 2.0 port to be tested on the system.

I would try the little "driver dance" solution in the aforementioned linked thread first.

[GomJabbar]

Because of the vast difference in transfer rates between the different USB modes, you should be able to use a timed file transfer to determine if you are getting High Speed USB, Full Speed USP, or Low Speed USB.
http://en.wikipedia.org/wiki/USB

USB supports three data rates.

* A Low Speed rate of 1.5 Mbit/s (183 KiB/s) that is mostly used for Human Interface Devices (HID) such as keyboards, mice, and joysticks.
* A Full Speed rate of 12 Mbit/s (1.5 MiB/s). Full Speed was the fastest rate before the USB 2.0 specification and many devices fall back to Full Speed. Full Speed devices divide the USB bandwidth between them in a first-come first-served basis and it is not uncommon to run out of bandwidth with several isochronous devices. All USB Hubs support Full Speed.
* A Hi-Speed rate of 480 Mbit/s (57 MiB/s).

[christopher_wolf]

That is not reliable for the majority of cases, as I *did* note in my post. It works only in a ridiculously simplified setting where other I/O transfer ops, that aren't accounted for in the simple arithmetic, are kept to an absolute minimum, ideally 0. Such is extremely rarely, if ever, the case on a Windows system that has been operating for any period of time.

I, myself, have done such tests and it depends, really, on the mood of the system and how much other data it is juggling, especially through *any* other communications port as Windows. I have been able, based solely on information garned from such an ad hoc test alone, to "conclude" that a USB port is operating at 1.1 instead of 2.0 when, in fact, I know that it is operating at 2.0 and the system is getting bogged down sharing a bus (as the other devices including Bluetooth and the FPR poll the main hub controller) or dealing with a fragmented cache somewhere. As such, it should not be trusted for getting conclusive data on the port. Badly implemented tests will produced data in bad context as demonstrated.

Go ahead and, with a mouse plugged in or not and with normal use of the system with a browser and maybe a word processor in use or WiFi on, try and measure it that way; I assure, a number significantly less than 2.0 (even far less than even the "realistic" speeds of most USB 2.0 ports) specs will be generated, mistakenly leading one to believe that the port is not operating at USB 2.0 Hi-Speed.

[GomJabbar]

It's easy enough to close all your programs to do a test. Depending upon the file transfer rate achieved, you can determine if you have crossed the threshold of the Low Speed or Full Speed USB connection. Of course to do this test you need to have a known High Speed USB device, such as an Optical drive, USB key, Memory Card Reader, Hard Drive, etc.

I am speaking here of testing a USB port on the ThinkPad itself. Using a Cardbus USB adapter may not give accurate results.

Also, don't use USB hubs or more than one USB device when running the test.

[christopher_wolf]

No, it does not matter if you close all your programs and whatnot. You will still get inaccurate results; I did not talk about a CardBus adapter either and I haven't mentioned it in use with these tests, so that, as well as hubs, are irrelevant. There is a hub in the Thinkpad itself, if one bothers to check, and eliminating all transit through it except for one port is *impossible*, except at the kernel level or below; I don't know why you think I am talking about that as there has been no mention of that until now.

As a test, I closed all the programs and process I possibly could and then tested it with a 58MB file transfer on a perfectly non-fragmented sector, I even *individually* checked the contig and fragmentation of the files just to be sure and defragmented them individually yet again. I got a calculated speed of 35MBps at best use, even though I *know* that the ports are USB 2.0. Another set of trials, I got near 20MBps, then again at 25MBps, then, when I loaded more programs, 1.5MBps, 1.0MBps, 5.7MBps, 25MBps, then finally, barely 1.5MBps; the error bars would be *huge* for this. Two orders of magnitude at least, not acceptable to *prove* that the port is operating in USB 2.0 all the time and not, as some systems are wont to do, falling back to 1.1 during large data transfers (no log is really generated by this event). It does not matter what the peaks of data transfer rates are, the USB storage device will silently fallback into a slower transfer mode, depending on what its internal controller thinks is appropriate for both the port abilities, the device abilities, and the amount of data being transfered. Therefore, one test is not enough and one would need many such tests; given also the significant inaccuracy of just such a test, that even compounds the problem, increasing the number of tests you need even *more*.

This is with an IEEE verified and compliant USB 2.0 USB Flash storage device that I have independently verified as using 2.0 and not dropping into fallback. I have done this on several Thinkpads I have available, including my T43, and all have similar results and all have fully operational USB 2.0 ports.

Again, you are coming to incorrect conclusions based off of an unsure foundation and references.

[GomJabbar]

A large enough file will average the results.

[christopher_wolf]

I tire of such trolling to waste time; having to demonstrate this yet again....

To test, I applied the same procedure as above, except I increase the file size to ~365MB; again through the, lengthy, tests, I obtained a distribution of file speeds that was very similar to the results of the first trials. You are confusing it stabilizing with it "accelerating" to the correct transfer rate. At various times, it would stabilize at 35MB, then dip to 10, then up to 25, then down to 5 and so forth. This was not at either close to the beginning or end of the file transfer as data obtained there would be even more erratic. This in no way proves that the device isn't, at some load point, falling back onto one of the slower USB transfer modes. Windows will not notify you of this because it sees everything as *set up* correctly for a USB 2.0 data transfer. This performs the same way as on my other systems and some test Thinkpads.

The port can malfunction and, in the middle of a transfer, fall back again with only the transient spike in data transfer rates as a clue. It is a characteristic of the device to negotiate the speed with the port in real time and, as such, can slip into a lower transfer rate abnormally; this indicates a malfunction with either the hardware on the device or the port, or a bug with the drivers. In addition, I am not going to try an even larger file. I have done this for 5.5GB files each averaging near a terabyte per week in my lab; I do know what I am talking about. As such, the results will be much the same and can't really spend time for proving what functional thought in the first place would cause one to arrive at anyway.

[jsteele]

Thanks for all the useful input.

I just tried a fairly crude copy test of a 307MB file and it took just under 4 minutes --- which would put it at USB 1 speeds - DESPITE the fact that I did NOT get the error message when I plugged in the USB drive.

So I guess I will try the Dance of the Drivers and see if that does anything. In my case there is one Intel 2.0 ENHANCED Host Controller thaat has has TWO driveroptions, the Intel driver and a "Standard Universal PCI to USB Host Controller" both of which are digitally signed. Under it are three Intel USD Universal Host Controllers, each of which has TWO Intel USB Universal Host Controllers and the Standard Universal PCI to USB controller. So I'm going to switch to the Standard driver and see if that helps (after I take a system checkpoint

[christopher_wolf]

Thanks for all the useful input.

I just tried a fairly crude copy test of a 307MB file and it took just under 4 minutes --- which would put it at USB 1 speeds - DESPITE the fact that I did NOT get the error message when I plugged in the USB drive.

So I guess I will try the Dance of the Drivers and see if that does anything. In my case there is one Intel 2.0 ENHANCED Host Controller thaat has has TWO driveroptions, the Intel driver and a "Standard Universal PCI to USB Host Controller" both of which are digitally signed. Under it are three Intel USD Universal Host Controllers, each of which has TWO Intel USB Universal Host Controllers and the Standard Universal PCI to USB controller. So I'm going to switch to the Standard driver and see if that helps (after I take a system checkpoint

Please do, *always* take a system restore point before you fiddle with drivers. Either with Windows or with IBM RnR (which does the job nicely). Excellent addition by the way, I must be getting sloppy; sorry about that. I would have felt extremely bad had you done what I said, which didn't include taking backups, and wound up in a sticky situation.

QED.

[GomJabbar]

At various times, it would stabilize at 35MB, then dip to 10, then up to 25, then down to 5 and so forth. This was not at either close to the beginning or end of the file transfer as data obtained there would be even more erratic. This in no way proves that the device isn't, at some load point, falling back onto one of the slower USB transfer modes. Windows will not notify you of this because it sees everything as *set up* correctly for a USB 2.0 data transfer. This performs the same way as on my other systems and some test Thinkpads.

The port can malfunction and, in the middle of a transfer, fall back again with only the transient spike in data transfer rates as a clue. It is a characteristic of the device to negotiate the speed with the port in real time and, as such, can slip into a lower transfer rate abnormally; this indicates a malfunction with either the hardware on the device or the port, or a bug with the drivers.

I won't argue with the above points. Yes, the transfer rate can vary during the time of the transfer. The point I was trying to make is to get an average for the entire transfer. However, a peak transfer rate should still tell one whether it is working in High Speed USB mode.

Unfortunately, at my present location I only have a Lexar 32MB 1.1 USB JumpDrive. Can't really determine too much with that.

A little seaching brought up this article from Tom's Hardware, where he did a comparison test of several USB Key drives. The link below shows his hardware setup. He used a program called h2benchw 3.6 to run his tests. He was testing transfer rates of USB flash drives. Data Transfer on the run: High-Speed USB Flash Drives

The program h2benchw 3.6 can be downloaded for free from this German site: heise mobil. The software is run from within Windows at a Command Prompt. Warning: Some of the commands will erase all data on the drive being tested. I downloaded and ran the software on my Lexar drive, and the results were consistant. I realize this is not much of a test, but the user can try it for himself. Just make sure the correct drive is chosen for the test. One can go to Computer Management > Storage > Disk Management to verify the drive number (i.e. 0, 1, 2, etc.). There is a switch to run the program in English h2bench -english. When I used the command below, I was unsure if my drive would be erased or not, so I backed it up first. After I ran the test, all the data was still on the drive.

The actual command I used to test the interface was: h2bench -english -c 90 2
The -english switch gives the output in English, rather than in German. The -c switch measures interface speed. The 90 is a percentage of total capacity. I chose 90%. The 2 is the drive number for my Lexar USB drive from Disk Management.

See screenshot of my results here: http://img220.imageshack.us/img220/6782/usbspeedxt0.png

PS: Please realize I am not trying to "talk down" to the Great Wolf. I am confident he understands using the command line very well - no doubt much better than I. I simplify everything here for the masses.

EDIT: Apparently Toms Hardware has the for-pay version of h2benchw 3.6. He has a nice pretty GUI interface to work with. I was stuck with the command line.

[christopher_wolf]

That doesn't matter; a test that was inherently inaccurate and error prone was being propounded. Ridiculous. Did it do the correct thing? No. Did the average time matter? No. That doesn't solve anything as, if it takes 4 minutes to transfer 307MB over a USB 2.0 port, then it *isn't USB 2.0*, end of discussion; average tiem notwithstanding. I can take the "average" again and again and it will always be different as well as have an *even larger error and variance* than the data points. Guess what happens when you sum measurements with variance, in this case large variance? Suprise, you get a a number with even *bigger* variance! The variance increases no matter what the operation is. So an "average" will do no good, calculate it yourself.

The h2bench tester is a benchmarking program only, the reason that the warning is there for the data that could be lost is because it takes near complete control of the USB drive and port, shutting out Windows File Protection. WFP slows things down a bit, but it makes sure files aren't damaged or lost due to unsynchronized access by multiple threaded jobs. Benchmarking results are great to link to, but it makes little difference in the real world. Literally no program that doesn't grab kernel level control of the USB port and device can do better than that benchmark. So if you aren't running at 2.0 speeds in the first place, it makes no difference.

Also, a little searching, can be done by the user with their own preferences. Do not tell me what command line I understand or do not understand. Such users, as shown 2 posts back, can simplify and understand things for themselves well enough (and not be patronized). Last I checked, such thought is called "Learning" and not "Regurgitating Search Info and Errorneous Online Edit-Fests"

Christopher "Great Wolf" Wolf

[jsteele]

Well, I tried the various drivers (system restore point taken and rebooting between each change. Result: no change.

[JaneL]

I tire of such trolling to waste time; having to demonstrate this yet again....

You two, get a room. Elsewhere.

[Kyocera]

Seems like this place is turning into a blogosphere for ceratain individuals.