Performance Index

ID Date Classification
615781 12/05/2024 Public
Document Table of Contents

Agilex FPGA

Claim Products/Technologies Segment Measurement Specification or Testing Information Test Date
1.9X better fabric performance compared to Cyclone® V FPGAs Agilex™ 3 FPGA Industrial, Surveillance, Retail, Consumer, Video processing, Medial Fabric performance comparison measurement is done between Cyclone® V FPGA and Agilex™ 3 FPGA The performance QoR (Quality of Results) Fmax (maximum frequency) represents the geomean derived from Altera's QoR design suite of 45 separate designs. Sep-24
38% lower total power vs. Cyclone® V FPGAs Agilex™ 3 FPGA Industrial, Surveillance, Retail, Consumer, Video processing, Medial The total power comparison is done with Agilex™ 3 core and Cyclone V® core. The total power comparison is done with Agilex™ 3 core power at 0.75V and Cyclone V® core power at 1.1V and both devices at 100% utilization running at 150 MHz. Sep-24
2x better fabric performance per watt compared to competing 7nm FPGAs Intel Agilex® 5 FPGA D-Series Mid range FPGA applications The Intel Agilex D-Series FPGA and Versal devices (part number / speed grade) used in the speed grade comparison are as follows: Intel Agilex D-Series FPGA ALD031 mid speed grade FPGA, Versal VM1102 -3HS Speed running at 600MHz for Power comparison. ​

Design profile used for the power comparison:​ 90% of utilization at 600MHz used across both devices for Power comparison. Details of the workload is as below:

Power estimates of Agilex D-series FPGA designs are obtained from internal Intel® analysis and Xilinx data is obtained through the latest version of Xilinx Versal XPE 2022.1.​

Sep-22
1.6x better Performance per Watt compared to competing 16nm FPGA Intel Agilex® 5 FPGA E-Series Embedded, Industrial, Vision The Intel Agilex FPGA code named Sundance Mesa and Ultrascale+ devices (part number/speed grade) used in the perf/watt comparison are as follows: Sundance SMA028 mid speed grade FPGA, Artix AU25P -2 Speed running at 500MHz for Power comparison.

Design profile used for the power comparison: 90% of utilization at 500MHz used across both devices for Power comparison. Details of the workload is as below:

Power estimates of SundanceMesa FPGA designs are obtained from internal Intel® analysis and Xilinx data is obtained through the latest version of Xilinx Ultrascale+ XPE 2022.1.2

Sep-22
Intel® Quartus® software on average delivers faster mid speed grade fabric performance compared to the highest speed grade competing 7nm FPGAs Intel Agilex® FPGA Intel® Quartus® Prime Pro Edition Software Data Center, Cloud, Communication, and Embedded

The Intel Agilex and Versal devices (part number / speed grade) used in the speed grade comparison are as follows: Agilex AGFB014R24A2E2V FPGA, Versal xcvm1802-vsva2197-3HP-e-S FPGA. Design profile used for the comparison: Base Stratix 10 frequency: 450MHz, Agilex Fmax = 450 * 1.74 = 783Mhz, Versal Fmax = 450 * 1.65 = 742.5Mhz.

Hardware Products compared: Intel FPGA AGFB014R24A2E2V vs AMD/Xilinx FPGA xcvm1802-vsva2197-3HP-e-S

Software Product compared: Intel Quartus Prime Pro Edition software version 22.3 vs. AMD/Xilinx Vivado 2022.1

On geomean, designs running in the mid speed grade of Intel Agilex FPGAs achieve > 5% higher in Fmax compared to the same design running in the fast speed grade of Xilinx Versal devices (-3H speed grade) when comparing unique designs. September 1st, 2021
Intel Agilex® 50% Fabric Performance improvement over Stratix® 10 FPGAs Intel Agilex® FPGA Data Center, Cloud, Communication and Embedded On average, designs running in the comparable mid-speed grade of Intel® Agilex™ FPGAs achieve a geomean of 50% improvement in Fmax compared to the same designs running in the most popular speed grade of Intel® Stratix® 10 devices. Derived from testing an example design suite comparing maximum clock speed (Fmax) achieved in Intel® Stratix® 10 devices with the Fmax achieved in Intel® Agilex™ devices, using Intel® Quartus® Prime Software. September 2021
Intel Agilex® averages 50% faster in Data Center, Network & Edge Intel Agilex® FPGA Data Center, Cloud, Communication and Embedded On average, designs running in the comparable mid-speed grade of Intel® Agilex™ FPGAs achieve a geomean of 50% improvement in Fmax compared to the same designs running in the most popular speed grade of Intel® Stratix® 10 devices. Derived from testing an example design suite (including designs used in data center, networking and edge applications)comparing maximum clock speed (Fmax) achieved in Intel® Stratix® 10 devices with the Fmax achieved in Intel® Agilex™ devices, using Intel® Quartus® Prime Software. September 2021
Up to 40% Lower Total Power Compared to Intel ® Stratix ® 10 FPGAs Intel Agilex® FPGA Data Center, Cloud, Communication and Embedded Derived from testing an example design suite comparing total power estimates of each design running in Intel® Stratix® 10 FPGAs compared to the total power consumed by the same design running in Intel® Agilex™ FPGAs. Bitmap PHOTO Power estimates of Intel® Stratix® 10 FPGA designs are obtained from Intel® Stratix® 10 and Agilex FPGA are obtained using Intel® FPGA Power and Thermal Calculator (PTC 20.4), and internal Intel® analysis and architecture simulation and modeling. The six workloads compared is shown below: December 2020
50% Higher Performance Compared to competing 7nm FPGA Intel Agilex® FPGA Embedded Derived from a set of five video IP designs comparing Fmax of each design achieved in Xilinx Versal ACAP devices with the Fmax achieved in Intel® Agilex™ devices, using Intel® Quartus® Prime Software (version 20.4) and Xilinx Vivado Software (version 2020.2). On geomean, designs running in the mid speed grade of Intel® Agilex™ FPGAs achieve a 50% higher in Fmax compared to the same designs running in the mid speed grade of Xilinx Versal devices (-2M speed grade), and 42% higher in Fmax compared to the same designs running in the fast speed grade of Xilinx Versal devices (-2H speed grade) and 24% higher in Fmax compared to the same designs running in the mid speed grade of Xilinx 16nm VUP devices (-2 speed grade). January 2021
~2X better Fabric Performance per Watt Compared to competing 7nm FPGA Intel Agilex® FPGA Data Center, Cloud, Communication and Embedded The Intel Agilex and Versal devices (part number/speed grade) used in the perf/watt comparison are as follows: Agilex: AGF014-2, Versal: Equivalent density to AGF014-2 in 2M speed grade. Design profile used for the comparison: Base Stratix 10 frequency: 450MHz, Agilex Fmax = 450 * 1.59 = 716Mhz, Versal Fmax = 450 * 1.19 = 536Mhz. Bitmap PHOTO Resource usage: 60% of AGF014 resource (logic, M20K memory, DSP), power at the respective Fmax Version of all the tools used for this data: Agilex: Quartus 20.4/PTC 21.1 b149, Versal: Vivado 2020.2/XPE: 2020.2. March 2021
Up to 49% faster fabric performance compared to prior generation FPGA for high-speed 5G fronthaul gateway applications Intel Agilex® FPGA Communication Derived from comparing the Fmax result of Agilex FPGA and Stratix 10 FPGA in a fronthaul gateway example design. using the FHGW example design and running the design targetting both Stratix 10 and Agilex in Quartus Prime 20.4 software. February 2021
Intel Agilex® performs 15% to 20% faster in Fmax when compared to competing 7nm FPGAs while also consuming an average of <5% logic for 5G Radio filters Intel Agilex® FPGA Communication Derived by comparing Fmax result of Intel Agilex 014 FPGA vs. Xilinx Versal VM 1802 FPGA Core fabric among 58 unique FIR filter designs. Designs were produced using Intel® Quartus® Prime Software 21.2, Xilinx's Vivado 2021.1. Intel's DSPBA 21.2, Xilinx's Model Composer 2021.1. MATLAB R2020B 64 bit. On geomean, designs running in the lowest speed grade of Intel Agilex 014 FPGAs achieve a 15-20% faster Fmax than their lowest speed grade Xilinx Versal VM 1802 counterpart and a 5% less logic footprint when comparing 58 unique FIR filter configurations. Filter design's target Fmax is 614.4, design variables are: Number of Channels (1,2,4,8,16), Filter Sample Rate (15.36,7.68,30.72 [MHz]), Filter Tap Count (87,107,35), Co_​efficient (Constant, Programmable). Variables were swept for each unique configuration. Oct 29, 2021
Intel Agilex® FPGAs on average are faster in Fmax than competing 7nm FPGAs by 20% in 5G Radio core functions. Intel Agilex® FPGA Communication Derived when comparing Fmax result of Intel Agilex 014 FPGA vs Xilinx Versal VM 1802 FPGA 5G radio unit functions; FFT & CP -, IFFT & CP+, DDC, DUC & CFR, PRACH. Designs were produced using Intel® Quartus® Prime Software 21.3, Xilinx's Vivado 2021.1. Intel's DSPBA 21.3, Xilinx's Model Composer 2021.1. MATLAB R2020B 64 bit. With 5G radio unit functions; FFT & CP -, IFFT & CP+, DDC, DUC & CFR, PRACH, on average Intel Agilex preforms 20% faster in Fmax. Designs running in the lowest speed grade of Intel Agilex 014 FPGA and Xilinx's Versal VM 1802 FPGA. Designs only in core FPGA Fabric. Nov 9, 2021