GX 1650 vs XCVU23P
A comparison of the Altera Stratix 10 GX/SX GX 1650 and the AMD Virtex UltraScale+ XCVU23P using the data from both vendors' selection guides.
Side-by-side specifications
| GX 1650 | XCVU23P | |
|---|---|---|
| Vendor / family | Altera Stratix 10 GX/SX | AMD Virtex UltraScale+ |
| Logic (LUT4-equivalent) | 1,101,080 | 1,648,000 |
| LUTs as published | 550,540 ALMs | 1,030,000 LUT6 |
| Vendor logic figure | 1,624,000 LE | 2,252,000 SLC |
| DSP | 3,145 variable-precision DSP blocks | 1,320 DSP slices |
| Block RAM | 114 Mb | 74.3 Mb |
| Large RAM | — | 99 Mb |
| Max. user I/O | 704 | 644 |
| Transceivers | 64 × XCVR 28.3 Gb/s, 32 × XCVR 17.4 Gb/s | 34 × GTY 32.75 Gb/s, 4 × GTM 58 Gb/s |
| Hard PCI Express | Gen3 x16 | Gen3 x16 |
| Hard processor | None | None |
| Process | 14 nm | 16 nm |
| Configuration memory | SRAM | SRAM |
| Packages | 2 | 0 |
| Design tool | Altera Quartus Prime | AMD Vivado |
Highlighted values are higher. Logic is compared in LUT4 equivalents; see how FPGA vendors count logic.
Key differences
The XCVU23P has about 50% more logic capacity (1648K vs 1101K LUT4 equivalents).
They use different logic architectures: the GX 1650 is built from ALMs and the XCVU23P from LUT6, so the raw LUT counts (551K vs 1030K) are not directly comparable.
For arithmetic, the GX 1650 has 3,145 variable-precision DSP blocks against 1,320 DSP slices on the XCVU23P.
The XCVU23P has more on-chip memory (173.3 Mb vs 114 Mb).
Both have transceivers: 96 at up to 28.3 Gb/s on the GX 1650 and 38 at up to 58 Gb/s on the XCVU23P.
The GX 1650 is designed with Altera Quartus Prime and the XCVU23P with AMD Vivado.
Full specifications
See the complete spec sheets with packages and grades: GX 1650 (Stratix 10 GX/SX) and XCVU23P (Virtex UltraScale+).