XI2000A issues

Hello,

Hard to tell what the root cause is, can you share your schematics?
Can you elaborate more about the failure mode?
Does this happens in both windows XP and windows 7?
Have you tried different PCI cards? Does your PCI cards have different drivers for WinXP and Win7 ? The incorrect driver could cause timing issues.

Are you configuring the XIo2000A with external arbiter enabled? Terminal EXT_ARB_EN
Regards

First of all, I appreciate you engaging with us on this issue.

It is now the Easter holidays here so for some more detailed information I will have to get back to you next week.

It is definitely not an issue with one PCI card, one Windows 7 machine or one XI2000A chip. We have tried multiple combinations of these both in our office and at our client's place of business and the failure is reproducible.

We have not tried Windows XP. We reached the point where we had a Windows 7 machine which happened to have two PCI slots and a PCI Express slot. Note this machine would not be usable by our clients, it's totally the wrong size. With the exact same software and drivers, everything ran fine for days when the PCI Cards were in the machine's internal PCI slots, but when the same PCI cards were put in the Startech box containing the XI2000A chip, the failure mode could be reproduced within seconds. As Windows XP is not an option for our client (this is the exact reason we are doing the work) we did not try to reproduce using the Startech box connected to an XP machine.

It is the exact same driver for Windows 7 (32-bit) and XP. However, as described in the previous paragraph, this driver does work fine with Windows 7, just not when we go through the PCI to PCI Express conversion process.

How exactly would using the external arbiter help? Is this something that can be configured from software, assuming that we can get a handle to the chip?

I will get back to you with further details on how we came to the conclusion that the XI2000A chip is not granting control of the bus.

Hello,

Please describe in more detail the failure mode, exactly at what point or which operation does the system hangs?
Please disable ASPM in the system. If there is no BIOS option to do this go to Control Panel>Power Options>Edit Plan settings>Change advanced settings and disable all power savings fr the PCIe.

Does this XIO2000A has an external EEPROM? Can you read those values? (we can provide a tool to do that, please provide an email address).

Regards

Elias,

It's rather difficult because I am trying to coordinate with Startech who provide the equipment which hosts your chip to get answers to your questions. I do not yet know if the XIO2000A is configured with an external arbiter enabled. I believe there is an EEPROM.

We have not yet attempted to read or write any registers in the XIO2000A beyond those which Windows 7 (32-bit) would configure as part of its native PCI Express / PCI support.

The problem occurs when the AMCC S5935 chip on our acquisition card is bus mastering to write data into system memory. If we run at our peak data write speed to the PCI bus (100Mbps) then within a few seconds the system hangs. Scoping the PCI signals after the failure, we see the S5935 asserting REQ# but the XIO2000A not responding with GNT#.

If we slow down the mean data write speed, then the system can run successfully for much longer (hours rather than seconds) without hanging. However we don’t think that 100Mbps is an unrealistic data rate to expect.

Reading the errata for the XIO2000A silicon, there appear to be a number of cases where system hangs can occur when a PCI device is bus mastering and data rates are increased, although the specific failure case that we have observed does not appear to be listed.

Could you provide the tool to read the EEPROM to phil.eccles@blueyonder.co.uk

Hello,
Your email server is rejecting the message with the tool, I am sending a .zip file
Regards