From: david on
On Sun, 03 Jan 2010 00:58:32 +0000, usbProblems rearranged some electrons
to say:

> USBproblems(a)hotmail.com wrote:
>>> When I transfer files to flash drive, I am only seeing speeds of about
>>> 8-10 MB/second.
>>>
>>> If USB 2 is capable of 60MB/second why the am I only getting 8-10.
>>> When the copy first starts, I have seen 60 MB/second for 5-10
>>> seconds, but then it gradually slows down to 8-10. Seems the larger
>>> the file the slower the transfer.
>>>
>>> WIN 7 Pro x64... fwiw.
>>>
>>> Thanks in advance.
>>>
>>>
>>The interface rate on the cable is 480 megabits/sec. Dividing by 8, that
>>is 60MB/sec.
>
>>USB protocol uses packets. The packets have headers. That wastes some of
>>the bandwidth. Now your usable rate is 57MB/sec.
>
>>Next, comes the packet protocol. There isn't a 100% stream of data
>>packets in one direction. The protocol is more complicated than that.
>>USB uses a polled protocol.
>
>>The end result, is you see things like hard drives in external USB
>>enclosures, do stuff at about 30MB/sec. Performance tops out, somewhere
>>in that vicinity.
>
>>This is an example of a dual channel USB stick. There are two flash
>>chips and a controller chip. It is a way of combining the bandwidth of
>>two chips. The numbers stated here are peak values. After a while, you
>>may see the average rate being somewhat lower than the manufacturer's
>>stated values. The peak rates are most likely to be evident when
>>transferring a single big file. Small file transfers will be
>>considerably slower.
>
>>http://www.ocztechnology.com/products/flash_drives/
ocz_rally2_turbo_usb_2_0_flash_drive
>
>>Try a dual channel stick, and see if you get double what you're
>>currently seeing. More than one manufacturer should make them. It is
>>pretty hard to fit more channels than that, without making the device
>>bulky.
>
>> Paul
>
> Hi Paul, as luck would have it I had a Super Talent dual channel stick
> and yes it does acheive twice the speed. Accourding to the profram
> USBDeview:
>
> Data traveler write speed is 6.26 while the Super talent is 13.2. When I
> plugged it into the motherboard USB port it actually slowed to 12.6.
>
> I guess I don't get where they can say USB 2.0 has a rate of
> 480mbs/second. I mean.. under what circumstance does it even acheive 90%
> of that?
>
> ANy way thanks for your help.

That is the maximum transfer rate of the physical interface. It does not
take into account the protocols involved. FWIW we designed a custom USB
interface to capture high speed serial data from a large number of
devices. Because it was custom, we were able to optimize the transfer
rate by using mostly high speed isochronous block transfers.
From: Sam Yosemite on
david <none(a)nospam.com> wrote in
news:hhq6cb$211$1(a)news.eternal-september.org:

> On Sun, 03 Jan 2010 00:58:32 +0000, usbProblems rearranged some
> electrons to say:
>
>> USBproblems(a)hotmail.com wrote:
>>>> When I transfer files to flash drive, I am only seeing speeds of
>>>> about 8-10 MB/second.
>>>>
>>>> If USB 2 is capable of 60MB/second why the am I only getting 8-10.
>>>> When the copy first starts, I have seen 60 MB/second for 5-10
>>>> seconds, but then it gradually slows down to 8-10. Seems the larger
>>>> the file the slower the transfer.
>>>>
>>>> WIN 7 Pro x64... fwiw.
>>>>
>>>> Thanks in advance.
>>>>
>>>>
>>>The interface rate on the cable is 480 megabits/sec. Dividing by 8,
>>>that is 60MB/sec.
>>
>>>USB protocol uses packets. The packets have headers. That wastes some
>>>of the bandwidth. Now your usable rate is 57MB/sec.
>>
>>>Next, comes the packet protocol. There isn't a 100% stream of data
>>>packets in one direction. The protocol is more complicated than that.
>>>USB uses a polled protocol.
>>
>>>The end result, is you see things like hard drives in external USB
>>>enclosures, do stuff at about 30MB/sec. Performance tops out,
>>>somewhere in that vicinity.
>>
>>>This is an example of a dual channel USB stick. There are two flash
>>>chips and a controller chip. It is a way of combining the bandwidth
>>>of two chips. The numbers stated here are peak values. After a while,
>>>you may see the average rate being somewhat lower than the
>>>manufacturer's stated values. The peak rates are most likely to be
>>>evident when transferring a single big file. Small file transfers
>>>will be considerably slower.
>>
>>>http://www.ocztechnology.com/products/flash_drives/
> ocz_rally2_turbo_usb_2_0_flash_drive
>>
>>>Try a dual channel stick, and see if you get double what you're
>>>currently seeing. More than one manufacturer should make them. It is
>>>pretty hard to fit more channels than that, without making the device
>>>bulky.
>>
>>> Paul
>>
>> Hi Paul, as luck would have it I had a Super Talent dual channel
>> stick and yes it does acheive twice the speed. Accourding to the
>> profram USBDeview:
>>
>> Data traveler write speed is 6.26 while the Super talent is 13.2.
>> When I plugged it into the motherboard USB port it actually slowed to
>> 12.6.
>>
>> I guess I don't get where they can say USB 2.0 has a rate of
>> 480mbs/second. I mean.. under what circumstance does it even acheive
>> 90% of that?
>>
>> ANy way thanks for your help.
>
> That is the maximum transfer rate of the physical interface. It does
> not take into account the protocols involved. FWIW we designed a
> custom USB interface to capture high speed serial data from a large
> number of devices. Because it was custom, we were able to optimize
> the transfer rate by using mostly high speed isochronous block
> transfers.
>

May I ask who "we" is and what rates you acheived?

Thanks,

Sam
From: david on
On Sun, 03 Jan 2010 16:29:03 +0000, Sam Yosemite rearranged some electrons
to say:

> david <none(a)nospam.com> wrote in
> news:hhq6cb$211$1(a)news.eternal-september.org:
>
>> On Sun, 03 Jan 2010 00:58:32 +0000, usbProblems rearranged some
>> electrons to say:
>>
>>> USBproblems(a)hotmail.com wrote:
>>>>> When I transfer files to flash drive, I am only seeing speeds of
>>>>> about 8-10 MB/second.
>>>>>
>>>>> If USB 2 is capable of 60MB/second why the am I only getting 8-10.
>>>>> When the copy first starts, I have seen 60 MB/second for 5-10
>>>>> seconds, but then it gradually slows down to 8-10. Seems the larger
>>>>> the file the slower the transfer.
>>>>>
>>>>> WIN 7 Pro x64... fwiw.
>>>>>
>>>>> Thanks in advance.
>>>>>
>>>>>
>>>>The interface rate on the cable is 480 megabits/sec. Dividing by 8,
>>>>that is 60MB/sec.
>>>
>>>>USB protocol uses packets. The packets have headers. That wastes some
>>>>of the bandwidth. Now your usable rate is 57MB/sec.
>>>
>>>>Next, comes the packet protocol. There isn't a 100% stream of data
>>>>packets in one direction. The protocol is more complicated than that.
>>>>USB uses a polled protocol.
>>>
>>>>The end result, is you see things like hard drives in external USB
>>>>enclosures, do stuff at about 30MB/sec. Performance tops out,
>>>>somewhere in that vicinity.
>>>
>>>>This is an example of a dual channel USB stick. There are two flash
>>>>chips and a controller chip. It is a way of combining the bandwidth of
>>>>two chips. The numbers stated here are peak values. After a while, you
>>>>may see the average rate being somewhat lower than the manufacturer's
>>>>stated values. The peak rates are most likely to be evident when
>>>>transferring a single big file. Small file transfers will be
>>>>considerably slower.
>>>
>>>>http://www.ocztechnology.com/products/flash_drives/
>> ocz_rally2_turbo_usb_2_0_flash_drive
>>>
>>>>Try a dual channel stick, and see if you get double what you're
>>>>currently seeing. More than one manufacturer should make them. It is
>>>>pretty hard to fit more channels than that, without making the device
>>>>bulky.
>>>
>>>> Paul
>>>
>>> Hi Paul, as luck would have it I had a Super Talent dual channel stick
>>> and yes it does acheive twice the speed. Accourding to the profram
>>> USBDeview:
>>>
>>> Data traveler write speed is 6.26 while the Super talent is 13.2. When
>>> I plugged it into the motherboard USB port it actually slowed to 12.6.
>>>
>>> I guess I don't get where they can say USB 2.0 has a rate of
>>> 480mbs/second. I mean.. under what circumstance does it even acheive
>>> 90% of that?
>>>
>>> ANy way thanks for your help.
>>
>> That is the maximum transfer rate of the physical interface. It does
>> not take into account the protocols involved. FWIW we designed a
>> custom USB interface to capture high speed serial data from a large
>> number of devices. Because it was custom, we were able to optimize the
>> transfer rate by using mostly high speed isochronous block transfers.
>>
>>
> May I ask who "we" is and what rates you acheived?
>
> Thanks,
>
> Sam

Sorry, I can't divulge that info. It was done by my company for a
restricted project.