KCU105 vs KCU116: Kintex UltraScale Development Board Comparison

Selecting between the KCU105 and KCU116 evaluation kits isn’t straightforward—they target different generations of Kintex UltraScale architecture with distinct price points and capabilities. Having worked with both boards on production prototypes, I’ve found each excels in specific scenarios. This comparison breaks down the EK-U1-KCU105-G and EK-U1-KCU116-G specifications, highlighting the practical differences that matter when choosing a development platform.

KCU105 vs KCU116: Quick Overview

The fundamental difference between these boards comes down to architecture generation. The KCU105 features the original Kintex UltraScale XCKU040 FPGA built on 20nm technology, while the KCU116 uses the newer Kintex UltraScale+ XCKU5P on 16nm FinFET+. This generational shift translates to meaningful differences in power efficiency, transceiver speeds, and on-chip memory—but the KCU105’s larger logic capacity still makes it the better choice for certain applications.

Specification	KCU105 (EK-U1-KCU105-G)	KCU116 (EK-U1-KCU116-G)
FPGA Device	XCKU040-2FFVA1156E	XCKU5P-2FFVB676E
Architecture	Kintex UltraScale	Kintex UltraScale+
Process Node	20nm	16nm FinFET+
Logic Cells	530,250	475,400
DSP Slices	1,920	1,824
Block RAM	21.1 Mb	34.9 Mb
UltraRAM	None	36 Mb
Transceivers	20x GTH (16.3 Gb/s)	16x GTY (32.75 Gb/s)
DDR4 Interface	64-bit component	32-bit component
Approximate Price	$7,170	$4,995

FPGA Device Comparison: XCKU040 vs XCKU5P

Understanding the underlying FPGA capabilities is essential before evaluating board-level features. The Xilinx KCU105 and Xilinx KCU116 target different device tiers within their respective families.

KCU105 FPGA Specifications (XCKU040)

The XCKU040 sits in the mid-range of the Kintex UltraScale family, offering substantial resources for complex designs:

Resource	XCKU040 Value
System Logic Cells	530,250
CLB Flip-Flops	484,800
CLB LUTs	242,400
Distributed RAM (Mb)	5.1
Block RAM (36Kb each)	600
Block RAM Total (Mb)	21.1
DSP48E2 Slices	1,920
GTH Transceivers	20
Max GTH Speed	16.3 Gb/s
Max I/O Pins	520
PCIe Hard Blocks	3 (Gen3 x8)

KCU116 FPGA Specifications (XCKU5P)

The XCKU5P represents the entry point for Kintex UltraScale+ with next-generation transceivers and on-chip UltraRAM:

Resource	XCKU5P Value
System Logic Cells	475,400
CLB Flip-Flops	443,520
CLB LUTs	221,760
Distributed RAM (Mb)	3.6
Block RAM (36Kb each)	480
Block RAM Total (Mb)	16.9
UltraRAM (Mb)	36
Total On-Chip Memory (Mb)	52.9
DSP48E2 Slices	1,824
GTY Transceivers	16
Max GTY Speed	32.75 Gb/s
Max I/O Pins	304
PCIe Hard Blocks	1 (Gen3 x8, Gen4 capable)

Architecture Differences: UltraScale vs UltraScale+

The move from UltraScale to UltraScale+ brings several architectural improvements:

Feature	UltraScale (KCU105)	UltraScale+ (KCU116)
Process Technology	20nm planar	16nm FinFET+
Power Efficiency	Baseline	Up to 30% lower
UltraRAM	Not available	Up to 36 Mb
Max Transceiver Speed	16.3 Gb/s (GTH)	32.75 Gb/s (GTY)
DDR4 Speed	Up to 2400 MT/s	Up to 2666 MT/s
PCIe Generation	Gen3	Gen3 (Gen4 capable)
100G Ethernet MAC	Soft IP required	Hard IP available

The UltraScale+ architecture’s FinFET transistors provide better power efficiency at equivalent performance levels, while UltraRAM blocks offer high-density, low-latency on-chip memory that reduces external memory bandwidth requirements.

Board Feature Comparison

Beyond FPGA specifications, board-level features determine practical usability for specific applications.

Memory Subsystems

Feature	KCU105	KCU116
DDR4 Width	64-bit	32-bit
DDR4 Capacity	2 GB component	1 GB component
DDR4 Speed	Up to 2400 MT/s	Up to 2666 MT/s
DDR4 Bandwidth	~19.2 GB/s	~10.6 GB/s
Quad SPI Flash	512 Mb	64 Mb
SD Card Slot	Yes (micro-SD)	Yes (micro-SD)

The KCU105’s 64-bit DDR4 interface provides nearly double the memory bandwidth compared to the KCU116’s 32-bit interface. For memory-intensive applications like video processing or large data buffering, this difference is significant. However, the KCU116’s UltraRAM can offset this for designs that can leverage on-chip memory.

High-Speed Connectivity

Feature	KCU105	KCU116
SFP+ Cages	2x (10G each)	4x zSFP+ (28G each)
Max Optical Speed	10 Gb/s	28 Gb/s
PCIe Edge Connector	Gen3 x8	Gen3 x8 (Gen4 capable)
SMA Connectors	4x (GTH TX/RX)	4x (GTY TX/RX)
FMC HPC Connector	Yes (8 GTH lanes)	Yes (4 GTY lanes)
FMC LPC Connector	Yes (1 GTH lane)	No

The KCU116’s zSFP+ cages and 32.75 Gb/s GTY transceivers enable 28G/25G Ethernet and other high-speed protocols impossible on the KCU105. For evaluating next-generation networking or data center applications, the KCU116 is the clear choice.

General Purpose I/O and Expansion

Feature	KCU105	KCU116
User LEDs	8	8
User Switches	5 DIP + 5 push	4 DIP + 5 push
PMOD Headers	2	2
HDMI Output	Yes	Yes
Ethernet PHY	Gigabit (Marvel 88E1111)	Gigabit (TI DP83867)
USB-UART	Dual (CP2105)	Dual (CP2105)

Both boards provide similar general-purpose features, with minor differences in switch counts and PHY vendors.

KCU105 Applications and Use Cases

The EK-U1-KCU105-G excels in scenarios requiring maximum logic density and memory bandwidth:

Data Center and Networking

The XCKU040’s 530K logic cells and 64-bit DDR4 interface handle complex packet processing and network acceleration. The dual SFP+ cages support 10G Ethernet development for data center applications.

DSP-Intensive Processing

With 1,920 DSP slices and substantial block RAM, the KCU105 targets digital signal processing applications including software-defined radio, medical imaging, and 8K video processing pipelines.

Wireless Infrastructure

Base station designs leverage the GTH transceivers for CPRI/OBSAI backhaul while the DSP resources handle digital up/down conversion and beamforming algorithms.

High-Performance Computing

The three PCIe Gen3 hard blocks and extensive transceiver count enable hardware acceleration cards that offload compute-intensive tasks from host processors.

KCU116 Applications and Use Cases

The Xilinx KCU116 targets applications requiring maximum transceiver speed and power efficiency:

25G/28G Ethernet Evaluation

The four zSFP+ cages with 32.75 Gb/s GTY transceivers enable development of 25G Ethernet, 28G Fibre Channel, and other next-generation networking protocols.

Low-Latency Data Acquisition

The combination of high-speed transceivers and UltraRAM provides deterministic, low-latency data paths for test and measurement equipment.

Power-Constrained Designs

The 16nm FinFET+ process delivers up to 30% power reduction versus equivalent UltraScale designs, making the KCU116 suitable for thermally constrained or battery-powered prototypes.

PCIe Gen4 Development

While requiring custom or third-party IP, the XCKU5P’s PCIe block supports Gen4 operation, enabling future-proof accelerator card development.

Kit Contents Comparison

EK-U1-KCU105-G Kit Contents

Item	Description
KCU105 Board	Main evaluation board with XCKU040
SFP+ Modules	2x 10Gbps optical transceivers
Fiber Patch Cable	For SFP+ loopback testing
FMC Loopback Card	For FMC connector validation
Power Supply	12V/5A adapter
USB Cables	For JTAG and UART
Vivado License	Device-locked to XCKU040

EK-U1-KCU116-G Kit Contents

Item	Description
KCU116 Board	Main evaluation board with XCKU5P
SFP28 Loopback Module	28Gbps electrical loopback
Power Supply	12V adapter
USB Cables	For JTAG and UART
Vivado License	Device-locked to XCKU5P

The KCU105 kit includes optical SFP+ modules for immediate 10G testing, while the KCU116 provides an electrical loopback module for 28G transceiver validation.

Development Tool Requirements

Both boards use Vivado Design Suite for development, but licensing differs:

Requirement	KCU105	KCU116
Vivado Edition	Design Edition or higher	Design Edition or higher
Included License	Device-locked to XCKU040	Device-locked to XCKU5P
Free WebPACK Support	No	No
IP Core Access	Full (with license)	Full (with license)

Neither board’s FPGA is supported by the free Vivado WebPACK edition. The included device-locked licenses provide full Vivado functionality for the specific FPGA on each board.

Essential Documentation and Resources

KCU105 Documentation

Document	Number	Description
KCU105 Board User Guide	UG917	Complete board reference
KCU105 Getting Started	UG884	Initial setup guide
Kintex UltraScale Data Sheet	DS892	FPGA specifications
UltraScale Architecture Overview	DS890	Architecture details

KCU116 Documentation

Document	Number	Description
KCU116 Board User Guide	UG1239	Complete board reference
KCU116 Getting Started	UG1237	Initial setup guide
Kintex UltraScale+ Data Sheet	DS922	FPGA specifications
UltraScale+ Architecture Overview	DS890	Architecture details

Download Resources

Resource	URL
KCU105 Product Page	amd.com/kcu105
KCU116 Product Page	amd.com/kcu116
Vivado Design Suite	amd.com/vivado
Board Design Files	Available on product pages

KCU105 vs KCU116: Which Should You Choose?

Choose KCU105 When:

• You need maximum logic capacity (530K vs 475K cells)
• Your design requires 64-bit DDR4 memory bandwidth
• 10G transceiver speeds are sufficient
• You need two FMC connectors (HPC + LPC)
• Budget allows for the higher price point

Choose KCU116 When:

• You need 25G/28G transceiver speeds
• Power efficiency is critical
• Your design benefits from UltraRAM
• You’re developing for PCIe Gen4
• Cost is a primary consideration ($2,000+ savings)

Frequently Asked Questions

What is the main difference between Kintex UltraScale and UltraScale+?

Kintex UltraScale uses 20nm planar technology while UltraScale+ uses 16nm FinFET+ process. UltraScale+ provides up to 30% lower power consumption, faster transceivers (32.75 Gb/s vs 16.3 Gb/s), on-chip UltraRAM blocks, and PCIe Gen4 capability. The UltraScale+ architecture is optimized for next-generation applications requiring maximum bandwidth efficiency.

Can I migrate designs between KCU105 and KCU116?

Designs can migrate between UltraScale and UltraScale+ with some effort. The core FPGA fabric is compatible, but transceiver configurations, clocking resources, and UltraRAM usage require attention. Memory interface widths differ (64-bit vs 32-bit DDR4), requiring MIG regeneration. Pin assignments are completely different due to distinct packages and board layouts.

Which board supports PCIe Gen4?

The KCU116’s XCKU5P FPGA supports PCIe Gen4 in compatibility mode, though custom or third-party IP is required as the standard Xilinx PCIe IP operates up to Gen3. The KCU105’s XCKU040 supports only PCIe Gen3. For production Gen4 designs, verify IP availability before committing to the KCU116 platform.

Are these boards supported by Vivado WebPACK?

No, neither the XCKU040 (KCU105) nor XCKU5P (KCU116) is supported by the free Vivado WebPACK edition. Both kits include device-locked Vivado Design Edition licenses valid for the specific FPGA on each board. For development targeting other devices, a separate Vivado license is required.

What is the power consumption difference between KCU105 and KCU116?

The KCU116’s 16nm FinFET+ process provides approximately 30% lower static power and up to 20% lower dynamic power compared to the KCU105’s 20nm technology at equivalent utilization levels. Actual power depends heavily on design complexity, clock frequencies, and transceiver usage. Use Vivado’s XPE (Xilinx Power Estimator) for accurate projections.

Conclusion

The choice between KCU105 and KCU116 ultimately depends on your specific application requirements. The KCU105 delivers more logic capacity and double the DDR4 bandwidth, making it ideal for complex designs with substantial memory requirements. The KCU116 provides faster transceivers, better power efficiency, and UltraRAM—advantages that matter for high-speed networking and power-constrained applications.

For teams evaluating Kintex UltraScale architecture for the first time, the KCU116’s lower price point and more modern process technology make it an attractive entry point. However, if your design already targets UltraScale devices or requires maximum logic density, the KCU105 remains a capable platform with proven reference designs and extensive documentation.

Both evaluation kits provide comprehensive development environments with included Vivado licenses, reference designs, and board design files to accelerate time-to-market for Kintex UltraScale-based products.