Table 3: Known Issues (cont'd)
Area
Description
Comments/Recommendations
General
Driver may not be loaded properly
after boot up (or reboot).
After boot up (or reboot), run these
two commands to load the device
driver:
$ sudo rmmod xclmgmt
$ sudo modprobe xclmgmt
Power
The Alveo U50 card limits HBM power
to 7.8W and FPGA fabric power to 58W.
Exceeding these power limits can cause
system instability.
To manage power consumption, review
design power usage and ensure that it
is within the power limits. Design
power estimates can be obtained using
the
report_power
Tcl command.
After implementing a design in the
Vivado
®
tools or in the Vitis
™
environment, open the implementation
result, add the
set_operating_conditions -
design_power_budget 63
constraint,
and run report power. See Vivado
Design Suite User Guide: Power Analysis
and Optimization (
) for setting up
power analysis.
Actual application power consumption
can be obtained by monitoring the
12V/3V PEX and 12V/3V PEX current
measurements provided by the
xbutil
query—which reports power
consumption at the input to the power
regulator.
For Vivado designs include the CMC IP
so that the system controller can
communicate with the device.
General
The Alveo card has not trained to the
full expected PCI Express link width or
link speed. The output from
xbutil
validate
will look like the following:
$ INFO: Validating device[0]:
INFO: Checking PCIE link
status: FAILED WARNING: Device
trained to lower spec. Expect:
Gen3 x16, Current: Gen2x16
Ensure that the Alveo card is plugged
into a Gen 3x16 or 4x8 capable
slot.Then cold reboot and see if the
card trains to the correct settings.
General
The card is not present when running
xbutil
or
lspci
. The card may not
have been ready when the server
enumerated PCI Express.
Potential Fix: Warm Reboot the server,
Disable Fast Boot.
General
Card does not show up when running
lspci and the red LED on the card is
illuminated.
When card is first installed in server,
BIOS may not recognize the card
correctly and red LED on card is
illuminated, indicating an error.
Cold boot the server four times until
the blue LED on the card is illuminated,
indicating the card is successfully
running.
If the red LED is still illuminated,
disconnect the power to the sever for 5
minutes and repeat the step above.
Chapter 7: Troubleshooting
UG1370 (v1.7) December 9, 2020
Alveo U50 Data Center Accelerator Card Installation Guide
32