AMD high speed transceivers (GTH, GTY) are supported within a RP. As with other reconfigurable site types, the entire PU must be included. For the UltraScale GT transceivers, the PU includes:
- 4 GT_CHANNEL sites (GT Quad)
- Associated GT_COMMON site
- Associated BUFG_GT_SYNC sites
- Associated BUFG_GT sites
- Associated Interconnect and CLB sites
The required GT PU is the entire height of a clock region. As with previous architectures, it is also possible to leave the GT components in static logic and change the functionality through the DRP. For more information on using UltraScale and UltraScale+ transceivers, see the UltraScale Architecture GTH Transceivers User Guide (UG576) or the UltraScale Architecture GTY Transceivers User Guide (UG578).
Virtex UltraScale+ High Bandwidth Memory (HBM) devices support Dynamic Function eXchange just like any other UltraScale+ device. Users have the choice of where the HBM Controllers are placed: in the static region or in the dynamic region. If the AXI High Bandwidth Memory Controller IP is kept in the static region with an active HBM reference clock, the memory interface will remain active and all memory contents are retained. Self-refresh mode can be used for power savings when the memory is not needed. If this IP is placed in a RM, it will be reconfigured and memory contents will be reinitialized per its Vivado customization. For more information on the HBM Controller IP, consult the documentation at https://www.xilinx.com/products/intellectual-property/hbm.html.