Did Intel's new processors change the way USB data is delivered to the drivers?

[mjplan]

Hi,

I have a product that consists of 2 USB devices, one USB1.1 camera and one USB2.0 streaming device. These two are connected on a hub. This hub is then connected to the computer. The chip on the streaming device is a cypress chip. Each data streamed contains data markers that are parsed on the end point. The data rate is approx 20MB/s. The data from the camera is streamed using directshow. The other USB device is streamed using QuickUSB asynchronously .

With older Intel processors (Core 2 duo) this set-up worked. On the desktop we could see data from both devices at the same time smoothly. The markers were intact. Then we tested on the newer Intel processers (i3, i5, i7). On these new computers, the data from the streaming device would be corrupted. The rate of data requests that would come back 'complete' is 1/20.

I'd suspected power issues, but both devices are powered properly, and just to rule it out we modified the hub so that it would be externally powered. Both devices would not use the power supplied by the usb hub but the external power. The results showed that external power did not improve condition.

If anyone has an idea on this, please let me know?

[Jan Axelson]

A hardware protocol analyzer will show what is happening on the bus. A software-only analyzer might also offer clues.

You could also try connecting the devices without using external hubs, using a different hub for each device, and connecting each device to a different host controller on the PC, to see if any of those make a difference.

Jan

[mjplan]

Thanks for the info. I didn't try protocol analyzer yet. I did download a trial of USBLyzer and will look at the data packets. Hardware analyzers are quite expensive (starting at $1400?).

I did try different hubs or without hubs. Without a hub, data loss was less, but it was still there. Different hubs didn't make any difference. I tried powered and non powered hubs. I was hoping that someone had similar problems where their USB devices stopped working properly on the chipsets/processors.

Cheers.

[Jan Axelson]

If the problem is missing data packets, check the interfaces the host selects during enumeration to see if the host is assigning sufficient bandwidth to each device.

20 MBytes/sec on each of two devices is 320Mbits/sec of data + overhead, which doesn't leave a whole lot of bandwidth for other devices.

Jan

[Snooker]

I've been having connect/disconnect problems with PC's at work.  During my research I found some troubling info in an Intel chipset errata http://www.intel.com/content/www/us/en/chipsets/6-and-c200-chipset-specification-update.html - go to page 16 and review the errata items.  The same errata apply to numerous chipsets not just the 6 series in the link above.  You'll notice that Intel has no plans to fix most or any of these bugs so if you discover a workaround post it here.
-Jim