FPGA & adaptive SoC databook · online editionRev. 2026.10

AGF 008 vs XCZU21DR

A comparison of the Altera Agilex 7 F-Series AGF 008 and the AMD Zynq UltraScale+ RFSoC XCZU21DR using the data from both vendors' selection guides.

Side-by-side specifications

AGF 008XCZU21DR
Vendor / familyAltera Agilex 7 F-SeriesAMD Zynq UltraScale+ RFSoC
LUTs as published259,200 ALMs425,000 LUT6
LUT typeadaptive logic module: fracturable 8-input LUT (one 6-input or two independent 4-input functions)6-input look-up table, fracturable into two LUT5 with shared inputs
LUT4 equivalent (approx.)≈518,400≈680,000
CoreScore (SERV cores)——
Vendor logic figure764,640 LE930,000 SLC
DSP2,296 variable-precision DSP blocks4,272 DSP slices
Block RAM74 Mb38 Mb
Large RAM—22.5 Mb
Max. user I/O576280
Transceivers32 × XCVR 58 Gb/s16 × GTY 28.21 Gb/s
Hard PCI ExpressNoneGen3 x16
Hard processor4× Arm Cortex-A53 (optional)4× Arm Cortex-A53
ProcessIntel 716 nm
Configuration memorySRAMSRAM
Packages31
Design toolAltera Quartus PrimeAMD Vivado

Highlighted values are higher. LUT counts are shown as published; the LUT4 equivalent is an approximation to compare different LUT types, see how FPGA vendors count logic. Add up to two more devices to this comparison.

Key differences

The XCZU21DR has about 31% more logic capacity (680K vs 518K LUT4 equivalents).

They use different logic architectures: the AGF 008 is built from ALMs and the XCZU21DR from LUT6, so the raw LUT counts (259K vs 425K) are not directly comparable.

For arithmetic, the XCZU21DR has 4,272 DSP slices against 2,296 variable-precision DSP blocks on the AGF 008.

The AGF 008 has more on-chip memory (74 Mb vs 60.5 Mb).

Both have transceivers: 32 at up to 58 Gb/s on the AGF 008 and 16 at up to 28.21 Gb/s on the XCZU21DR.

The AGF 008 is designed with Altera Quartus Prime and the XCZU21DR with AMD Vivado.

Full specifications

See the complete spec sheets with packages and grades: AGF 008 (Agilex 7 F-Series) and XCZU21DR (Zynq UltraScale+ RFSoC).