feature article
Subscribe Now

Take as Much Time as You Need

Those of you familiar with San Francisco may have traveled a beautiful seaside stretch of road called the Great Highway, which runs from Ocean Beach down to the zoo. The portion south of Golden Gate Park is less traveled and passes through sparsely vegetated dunes with no intersecting roads. However, even though the road is isolated from the Sunset District’s vehicular traffic, the beach is made accessible to local pedestrians via a series of crosswalks protected by stoplights.

Coursing down this way at 2 in the morning, you’ve got a good chance of being free of competing traffic, and there are no pedestrians crossing. So the temptation arises to take advantage of a long, uninterrupted stretch of asphalt and clear some of the carbon from your cylinders. But, for no apparent reason, and with no other cars or people in sight, the stoplight at the next crosswalk will turn red, and your straightaway dreams are dashed, at least for the moment. Presently, the light will, of course, turn green, giving you one more opportunity to feel the rumble of latent horsepower heaving and straining under the bridle of provincial speed limits. Again you go, and again the next light turns red, and you must once more concede an ignominious halt.

You can repeat this as many times as you want, and lo, you will find that you are stopping at every light. After driving this road a couple times, with the same result each time, it occurs to you: the lights are timed. You will get through when the lights let you get through, no faster. An impressive display of acceleration will get you to the end no earlier, and, in fact, taking into account the full impact of your frustrated attempt at glory, you will find that you have burned an inordinate amount of fuel with nothing to show for it, save perhaps a few well-placed black marks on the road and a commensurate reduction in the life of your tires.

The trick lies in slowing down: if you go at just the right speed – something under 30 mph – you will find that, once you get in synch, each red light will obligingly turn green for you as you arrive. You will be tempted to feather your brake as you approach the next red light, but if you keep the faith, the lights will gracefully guide you down the entire stretch of road without your having to touch either the brake or the accelerator. Attempting to get from one light to the next any faster provides no benefit, and only costs more in energy.

The philosophy of going only as fast as needed is taking hold in circuit design. Time was when the one clock frequency on a chip was determined by the neediest portion of the system, and all other parts of the circuit hewed to that clock. So if video needed compressing, an audio packet needed decoding, an Ethernet packet needed exploring, it was done at a rate determined by some other part of the circuit and might complete long before needed.

So what’s wrong with that? Looking at it one way, let’s say that it takes 30,000 clock cycles to complete a task, and that task involves 50 signals making a combined, oh, 600 transitions (just making numbers up). If you do that task in 30 µs or in 30 ms (with a 1000x slower clock), it will still involve 30,000 clock cycles and 600 transitions on 50 signals. Since dynamic power is determined by transitions, on the face of it, you’ve done the same amount of “work,” and so have burned the same amount of energy (assuming the same slew rates). Power will be lower in the slower case, since it takes longer to complete the task, and so the rate of energy burn is reduced, but, in the end, the energy consumed is the same. So slowing down would appear to have no real impact.

But today’s chips aren’t the simple monolithic entities of yore. Modern behemoths have multiple frequency domains and – importantly – multiple power domains. The real benefit of slowing down to smell the sea salt comes not from the reduced frequency itself, but from the fact that, with a less frenetic frequency requirement, the circuit can be fed by a lower voltage. While power is linearly related to frequency, it goes as the square of voltage. So reducing the frequency reduces power somewhat, without reducing the energy used, but reducing the voltage has a beneficial impact on both power and energy.

Reducing the voltage also cuts back on the static leakage in a manner that simply reducing frequency can’t accomplish. You may also be able to raise the VTs on the transistors since they don’t need to operate as quickly. That further reduces gate leakage and helps trim the static power consumption. All of the required work of the chip can be completed on schedule, with reductions in dynamic and static power, and, most critically, with a lower energy bill.

So just as powering your way from stoplight to stoplight uses more fuel to no advantage, so driving your circuits faster than necessary uses more energy, also to no advantage. Granted, creating a new power domain involves some circuitry and some complexity in the design and verification of the circuit, so it’s not entirely without cost. However, such increased sophistication has become standard fare in any self-respecting SoC, and so, in practical fact, it really imposes no significant extra burden. Since it’s not necessary (or advisable, or even possible) to optimize every circuit to the point where each little block is exactly optimally timed, you can combine many such blocks in a single domain and thus better amortize the cost of managing the domain.

So consider it permission to relax. First car to the stoplight does not win. Hurry up and wait seems a bit silly, like testosterone dashed with cold water. Heck, it’s always good form to look busy: if you complete your work too fast, you end up sitting around looking idle, which is never recommended when the boss comes around. So pace yourself; don’t be late but don’t be early. Reduce your stress level and your blood pressure. It will all come together, and you’ll have energy to spare when the day is done.

And because you don’t have to pay so much attention to the stoplights, you can look around and enjoy the scenery. OK, fair enough, you can’t see the ocean because of the dunes, and it’s hard to see much at 2 AM anyway. But at a more relaxed pace, you can open the window and let the gentle breeze waft the savory scents of sand and seaweed through the car, and, if you’re lucky, you can watch the fog muffle the beam of your headlights as it envelops you in soft marine silence.

Leave a Reply

featured blogs
Oct 22, 2021
Voltus TM IC Power Integrity Solution is a power integrity and analysis signoff solution that is integrated with the full suite of design implementation and signoff tools of Cadence to deliver the... [[ Click on the title to access the full blog on the Cadence Community site...
Oct 21, 2021
We share AI chip design insights from AI Hardware Summit 2021, including wafer scale AI accelerator chips, high-bandwidth memory interfaces, and custom SoCs. The post 4 Futuristic Design Takeaways from the AI Hardware Summit 2021 appeared first on From Silicon To Software....
Oct 20, 2021
I've seen a lot of things in my time, but I don't think I was ready to see a robot that can walk, fly, ride a skateboard, and balance on a slackline....
Oct 4, 2021
The latest version of Intel® Quartus® Prime software version 21.3 has been released. It introduces many new intuitive features and improvements that make it easier to design with Intel® FPGAs, including the new Intel® Agilex'„¢ FPGAs. These new features and improvements...

featured video

Fast & Accurate 3D Object Detection for LiDAR with DesignWare ARC EV Processor IP

Sponsored by Synopsys

This demo, developed in partnership with Sensor Cortek, executes the FA3D algorithm on ARC EV7x processor with DNN engine. It shows 3D boxes rendered onto objects detected in the video frames, enabling the development of driver assistance systems.

Click here for more information

featured paper

3 ways to speed cycle time when designing with brushless-DC motors

Sponsored by Texas Instruments

Designing systems with BLDC motors can be challenging, because it usually requires complex hardware and optimized software designs to deliver reliable real-time control. This article discusses three methods for simplifying the process.

Click to read more

featured chalk talk

Automotive Infotainment

Sponsored by Mouser Electronics and KEMET

In today’s fast-moving automotive electronics design environment, passive components are often one of the last things engineers consider. But, choosing the right passives is now more important than ever, and there is an exciting and sometimes bewildering range of options to choose from. In this episode of Chalk Talk, Amelia Dalton chats with Peter Blais from KEMET about choosing the right passives and the right power distribution for your next automotive design.

Click here for more information about KEMET Electronics Low Voltage DC Auto Infotainment Solutions