The following figure shows the design stages in which you use power analysis tools, and compares the accuracy of the estimations for different input types:

The following table lists the differences between the EPE and the Intel^® Quartus^® Prime Power Analyzer.

The following revision history applies to this chapter:

The following figure shows the power dissipation of Stratix^® and Cyclone^® devices in different designs. The analysis considers a fixed clock rate of 100  MHz and exhibits varied logic resource utilization across available resources.

In Stratix^® and Cyclone^® device families, a series of column and row interconnect wires of varying lengths provide signal interconnections between logic array blocks (LABs), memory block structures, and digital signal processing (DSP) blocks or multiplier blocks. These interconnects dissipate the largest component of device power.

FPGA combinational logic is another source of power consumption. For more information about ALMs and LEs in Cyclone^® or Stratix^® devices, refer to the respective device handbook.

Memory and clock resources are other major consumers of power in FPGAs. Stratix^® devices feature the TriMatrix memory architecture. TriMatrix memory includes 512-bit M512 blocks, 4-Kbit M4K blocks, and 512-Kbit M-RAM blocks, which are configurable to support many features. Stratix^® IV TriMatrix on-chip memory is an enhancement based upon the Stratix^® II FPGA TriMatrix memory and includes three sizes of memory blocks: MLAB blocks, M9K blocks, and M144K blocks. Stratix^® IV and Stratix^® V devices feature Programmable Power Technology, an advanced architecture that enables a smooth trade-off between speed and power. The core of each Stratix^® IV and Stratix^® V device is divided into tiles, each of which may be put into a high-speed or low-power mode. The primary benefit of Programmable Power Technology is to reduce static power, with a secondary benefit being a small reduction in dynamic power. Cyclone^® IV  GX devices have 9-Kbit M9K memory blocks.

The DSE II offers two options in Exploration mode that target power optimization: Power (High Effort) and Power (Aggressive). In both cases, the target is an overall improvement in the design's power; specifically, reducing the total thermal power in the design.

When the optimization targets power, the DSE II runs the Intel^® Quartus^® Prime Power Analyzer for every group of settings. The resultant reports help you debug the design and determine trade-offs between power requirements and performance optimization.

Intel^® Quartus^® Prime software settings that control power-driven compilation are located in the Power optimization during synthesis list in the Advanced Settings (Synthesis) dialog box, and the Power optimization during fitting list on the Advanced Fitter Settings dialog box. The following sections describes these power optimization options at the Analysis and Synthesis and Fitter levels.

The Power Optimization During Synthesis logic option determines how aggressively Analysis & Synthesis optimizes the design for power. To access this option at a project level, click Assignments > Settings > Compiler Settings > Advanced Settings (Synthesis).

You can also control memory optimization options from the Intel^® Quartus^® Prime Settings dialog box. The Default Parameters page allows you to edit the Low_Power_Mode parameter. The settings for this parameter are equivalent to the values of the Power Optimization During Synthesis logic options. The Low_Power_Mode parameter always takes precedence over the Optimize Power for Synthesis option for power optimization on memory.

Memory blocks can represent a large fraction of total design dynamic power. Minimizing the number of memory blocks accessed during each clock cycle can significantly reduce memory power.

In the default implementation of a simple dual-port memory block, write-clock enable signals and read-clock enable signals connect to V_CC, making both read and write memory ports active during each clock cycle.

Memory transformation moves the read-enable and write-enable signals to the respective read-clock enable and write-clock enable signals. This technique reduces the design’s memory power consumption, because memory ports are shut down when they are not accessed.

For Stratix^® IV and Stratix^® V devices, the memory transformation takes place at the Fitter level by selecting the Normal compilation settings for the power optimization option.

In Cyclone^® IV GX and Stratix^® IV devices, the read-during-write behavior impacts the power of single-port and bidirectional dual-port RAMs. As a best practice, you can allow optimization by setting the read-during-write parameter to “Don’t care” at the HDL level, and set the read-enable signal to the inversion of the existing write-enable signal (if one exists). This allows the core of the RAM to shut down, which prevents switching, saving a significant amount of power.

The Compiler includes this optimization technique when the Power Optimization During Synthesis logic option is set to Extra effort.

There is a trade-off between power saved by accessing fewer memories and power consumed by the extra decoder and multiplexor logic. The Intel^® Quartus^® Prime software automatically balances the power savings against the costs to choose the lowest power configuration for each logical RAM. The benchmark data shows that the power-driven synthesis can reduce memory power consumption by as much as 60% in Stratix^® devices.

You can also set the MAXIMUM_DEPTH parameter manually to configure the memory for low power optimization. This technique is the same as the power-aware memory balancer, but it is manual rather than automatic like the Extra effort setting in the Power optimization list. The MAXIMUM_DEPTH parameter always takes precedence over the Optimize Power for Synthesis options for power optimization on memory optimization. You can set the MAXIMUM_DEPTH parameter for memory modules manually in the Intel^® FPGA IP instantiation or in the IP Catalog.

For Stratix^® IV and Stratix^® V devices, the Normal compilation setting enables the Programmable Power Technology to configure tiles as high-speed mode or low-power mode. Programmable Power Technology is always turned ON even when the OFF setting is selected for the Power optimization option. Tiles are the combination of LAB and MLAB pairs (including the adjacent routing associated with LAB and MLAB), which can be configured to operate in high-speed or low-power mode. This level of power optimization does not have any affect on the fitting, timing results, or compile time.

The Extra effort setting the Fitter works to minimize power even after the design meets timing requirements by moving the logic closer during placement to localize high-toggling nets and choosing routes with low capacitance.

The Extra effort setting uses a Value Change Dump (.vcd) file that guides the Fitter to fully optimize the design for power, based on the signal activity of the design. The best power optimization during fitting results from using the most accurate signal activity information. If there is no .vcd file, the Intel^® Quartus^® Prime software estimates the signal activities from the settings in the Power Analyzer Settings page in the Settings dialog box, such as assignments, clock assignments, and vectorless estimation values. The benchmark data shows that the power-driven Fitter technique can reduce power consumption by as much as 19% in Stratix^® devices. On average, you can reduce core dynamic power by 16% with the Extra effort synthesis and Extra effort fitting settings, as compared to the Off settings in both synthesis and Fitter options for power-driven compilation.

The Intel^® Quartus^® Prime software provides Speed, Balanced, or Area for the Optimization Technique option. You can also specify this logic option for specific modules in your design with the Assignment Editor in cases where you want to reduce area using the Area setting (potentially at the expense of register-to-register timing performance) while leaving the default Optimization Technique setting at Balanced (for the best trade-off between area and speed for certain device families). The Speed Optimization Technique can increase the resource usage of your design if the constraints are too aggressive and can also result in increased power consumption.

The benchmark data shows that the area-driven technique can reduce power consumption by as much as 31% in Stratix^® devices and as much as 15% in Cyclone^® devices.

The Perform gate-level register retiming option in the Intel^® Quartus^® Prime software enables the movement of registers across combinational logic to balance timing, allowing the software to trade off the delay between critical and noncritical paths.

Retiming uses fewer registers than pipelining. In this example of gate-level register retiming, the 10 ns critical delay is reduced by moving the register relative to the combinational logic, resulting in the reduction of data depth and switching activity.

Gate-level register retiming makes changes at the gate level. If you are using an atom netlist from a third-party synthesis tool, you must also select the Perform WYSIWYG primitive resynthesis option to undo the atom primitives to gates mapping (so that register retiming can be performed), and then to remap gates to Intel primitives.

When using Intel^® Quartus^® Prime integrated synthesis, retiming occurs during synthesis before the design is mapped to Intel primitives. The benchmark data shows that the combination of WYSIWYG remapping and gate-level register retiming techniques can reduce power consumption by as much as 6% in Stratix^® devices and as much as 21% in Cyclone^® devices.

To set the optimization mode on the Intel^® Quartus^® Prime software, click Assignments > Settings > Compiler Settings.

The two power optimization modes direct the Compiler to prioritize one optimization metric.

Power (High effort—increases runtime)

High effort modes enable additional optimizations that increase compilation time and do not affect design performance. High Power Effort mode guides the Compiler to spend additional compilation time reducing routing utilization, which saves dynamic power.

Power (Aggressive—increases runtime, reduces performance)

Aggressive modes increase compilation time and make trade-offs that may harm other optimization metrics (performance, area, etc.). In Aggressive Power mode, the Compiler attempts to reduce the routing usage of signals with the highest specified (via Signal Activity File) or estimated toggle rates, saving additional dynamic power but potentially affecting performance.

The Optimization Technique logic option specifies the overall optimization goal for Analysis & Synthesis: attempt to maximize performance or minimize logic usage.

Figure 12. Optimization Technique Options

The Power Optimization During Synthesis logic option determines how aggressively Analysis & Synthesis optimizes the design for power.

Figure 13. Power Optimization During Synthesis Options

The Intel^® Quartus^® Prime software optimizes clock routing power automatically, enabling only those portions of the clock network that are necessary to feed downstream registers.

When a memory block is clocked, a sequence of timed events occur within the block to execute a read or write. The circuitry that the clock controls consumes the same amount of power, independent of changes in address or data from one cycle to the next. Thus, the toggle rate of input data and the address bus have no impact on memory power consumption.

The key to reducing memory power consumption is to reduce the number of memory clocking events. You can achieve this reduction through network-wide clock gating, or on a per-memory basis through use of the clock enable signals on the memory ports.

The clock enable signal enables the memory only when necessary, and shuts down for the rest of the time, reducing the overall memory power consumption. You include these enable signals when generating the memory block function.

The Intel^® Quartus^® Prime software automatically chooses the best design memory configuration for optimal power. However, you can set the MAXIMUM_DEPTH parameter for memory modules during the IP core instantiation.

Using the MAXIMUM_DEPTH parameter can save power. For all implementations, a user-provided read enable signal is present to indicate when read data is required. Using this power-saving technique can reduce power consumption by as much as 60%.

As the memory depth becomes more shallow, memory dynamic power decreases because unaddressed M4K blocks can be shut off using a decoded combination of address bits and the read enable signal. For a 128-deep memory block, power used by the extra LEs starts to outweigh the power gain achieved by using a more shallow memory block depth. The power consumption of the memory blocks and associated LEs depends on the memory configuration.

To reduce LAB-wide clock power consumption without disabling the entire clock tree, use the LAB-wide clock enable to gate the LAB-wide clock. The Intel^® Quartus^® Prime software automatically promotes register-level clock enable signals to the LAB-level. A shared gated clock controls all registers within an LAB that share a common clock and clock enable. To take advantage of these clock enables, use a clock enable construct in the relevant HDL code for the registered logic.

IF clk'event AND clock = '1' THEN
    IF logic_is_enabled = '1' THEN
        reg <= value;
    ELSE
        reg <= reg;
    END IF;
END IF;

Use clock enables instead of gated clocks:

assign clk_gate = clk1 & gateA & gateB;
always @ (posedge clk_gate) begin
   sr[N-1:1] <= sr[N-2:0];
   sr[0]<=din1;
end

assign enable = gateA & gateB;
always @(posedge clk2) begin
   if (enable) begin
      sr[N-1:1] <= sr[N-2:0];
      sr[0]<=din2;
   end
end

Reduce LAB-wide clock power consumption without disabling the entire clock tree, use the LAB-wide clock enable to gate the LAB-wide clock.

always @(posedge clk)
begin
   if (ena)
      temp <= dataa;
   else
      temp <= temp;
   end
end

Intel^® FPGAs have different kinds of global signal resources available. Global signals can span the entire chip or smaller regions. Choose the clock networks that can cover all the fanout on a specific domain. For example, you can reduce clock power by switching from a clock network that spans the entire chip to one quarter of the chip, provided all the fanout for that clock is within that region of the chip.

Evaluate the possibility of merging clocks and PLLs in the design.

• 2clks & 2PLLs

  Clk1:350Mhz, Fanout 46788

  Clk2: 365Mhz, Fanout 2450

• 1Clk & 1PLL

  Merge clks

  clk: 365Mhz, Fanout 51277

A glitch can propagate to subsequent logic and create unnecessary switching activity, increasing power consumption. Circuits with many XOR functions, such as arithmetic circuits or cyclic redundancy check (CRC) circuits, tend to have many glitches if there are several levels of combinational logic between registers.

Registers stop glitches from propagating through combinational paths. Pipelining is a technique that breaks combinational paths by inserting registers. By reducing logic-level numbers between registers, pipelining can result in higher clock speed operations. However, pipelining increases the latency of a circuit in terms of the number of clock cycles to a first result.

The following figure shows how pipelining breaks a long combinational path.

This reduction in switching activity lowers power dissipation in combinational logic. However, for designs with few glitches, pipelining can increase power consumption by adding unnecessary registers. Pipelining can also increase resource utilization. Benchmark data shows that pipelining can reduce dynamic power consumption by as much as 30% in Cyclone^® and Stratix^® devices.

Nonterminated I/O Standards

Nonterminated I/O standards such as LVTTL and LVCMOS have a rail-to-rail output swing. The voltage difference between logic-high and logic-low signals at the output pin is equal to the V_CCIO supply voltage. If the capacitive loading at the output pin is known, the following expression determines the dynamic power consumed in the I/O buffer:

Transistor-to-transistor logic (TTL) I/O buffers consume very little static power. As a result, the total power that a LVTTL or LVCMOS output consumes is highly dependent on load and switching frequency.

Resistively Terminated I/O Standards

In resistively terminated I/O standards like SSTL and HSTL, the output load voltage swings by a small amount around a bias point. The dynamic power equation above is valid as well, but V is the actual load voltage swing. This voltage is much smaller than V_CCIO, resulting in lower dynamic power when comparing to nonterminated I/O under similar conditions.

Resistively terminated I/O standards dissipate significant static (frequency-independent) power, because the I/O buffer is constantly driving current into the resistive termination network. However, the lower dynamic power of these I/O standards means they often have lower total power than LVCMOS or LVTTL for high-frequency applications. As a best practice, when using resistively terminated standards choose the lowest drive strength I/O setting that meets the speed and waveform requirements to minimize I/O power.

You can save a small amount of static power by connecting unused I/O banks to the lowest possible V_CCIO voltage.

Power-Down Mode

Enable Auto Power-Down directs the controller to place the memory device in power-down mode after a specific number of idle controller clock cycles. You can configure the idle wait time. All ranks must be idle to enter auto power-down.

Auto Power-Down Cycles Number of cycles the controller must be IDLE before entering power down state. You determine the number based on the traffic pattern. If the number is too small, the control enters power down too frequently, affecting efficiency. The Intel^® Arria^® 10 device family supports from 1 to 65534 cycles.

Self-Refresh

Directs the Controller to self-refresh when not sending traffic for very long period. Self-refresh takes more time compared to power down, but the power saving is higher than power down.

When you maximize the packing of DSP blocks, you reduce Logic Utilization, power consumption, and increase efficiency. The HDL coding style grants you control of the DSP resources available in the FPGA.

Transceivers in the device degrade over time unless you preserve them. The Intel^® Quartus^® Prime software generates a warning message if a design contains unused XCVRs.

You do not need to preserve transceivers under 8Gbps. For transceivers over 8Gbps, the best practice is to preserve if there is a possibility for future usage. Otherwise, you can turn the transceivers off. You enable unused transceivers through dynamic reconfiguration or a new device programming file​.

Small Device

Use the smallest device which can fit your design.

Long chains of cascaded logic blocks can create glitches due to path delay differences between the input signals. Inserting Flip-Flops to cut these long chains terminates the propagation of glitches to consecutive logic cells.

Circuits that heavily use of XIO functions (for example, Cyclic redundancy check) tend to glitch significantly when cascaded. Add pipeline registers or re-architect to reduce signal toggling

The following revision history applies to this chapter:

Refer to the following user guides for comprehensive information on all phases of the Intel^® Quartus^® Prime Standard Edition FPGA design flow.