feature article
Subscribe Now

Why Hamsters Cannot Save the Planet

While low power was officially the theme of only a panel session at Globalpress Electronics Summit, it was also frequently the theme, sometimes explicit and frequently implicit, of much of what was discussed.

The most striking performance was by John East of FPGA company, Actel. He spoke mainly from the perspective of the USA, which is probably appropriate, as the US, with around 5% of the world’s population, daily consumes 20.7 million barrels of oil, or 25% of world consumption. “This 20-million-barrel-a-day habit costs $1.4 billion a day,” East said, but that was when oil was at $70 a barrel; today it is twice that, so to feed their habit costs the population of the USA around $9 a day for every man, woman and child. Feeding the habit requires the US to import oil and has adverse effects “on the environment, the economy, and the political landscape.”

Where does this oil go? Well 45-50% of fossil fuel pollution is caused by road vehicles, and around 45% is caused by generating electricity. While half of this is used in electric motors, computing and telecoms are growing their share rapidly, and East stated that server farms alone, running search engines, e-commerce, and design applications, are now responsible for burning 1.2% of the US electricity. (Server farms recurred as a topic – see later.)

Surely power can be saved by using less power-hungry silicon? “With the tools at hand, we don’t have the capability to lick the power problem,” East feels. Each time there is a process breakthrough that reduces the power consumed by devices, the sheer growth in use of silicon defeats it within months. So, although we ought to be looking at saving power, we should also be looking at alternative sources of energy.

One idea that, at least initially, looked promising was the hamster car, originally shown at the Toyota Idea Olympics in 2002. This uses a hamster in a wheel to drive a generator, which in turn drives an electric motor to propel the car. However, always the engineer, East started to do the math. We start with assumptions: one hamster power (hp) is probably about one thousandth of a horse power (HP), and the efficiency of a drive wheel/generator/motor is probably around 64%. To get a 64 HP car would therefore need around 100,000 hamsters rushing round inside their little wheels.

Now, East further argues, there are around 250 million cars in use in the US. (Most of them seem to be on 101 whenever I arrive at the San Francisco airport.) That is a lot of hamsters to feed, so oil imports would probably have to be replaced by grain imports. And hamsters have a low conversion efficiency on grain, so there would be a lot of hamster poop to get rid of. In late nineteenth century cities, the crossing sweeper, who swept the road clear of horse dung, was a familiar figure. Somehow a sweeper of hamster poop doesn’t have the same status. (I went on to do a couple more sums – if the USA (the 50 states and District of Columbia) is, according to the CIA, around 10,000,000 sq km., then that averages around 2.5 hamsters every square metre or 10 square feet, not allowing for those breeding and growing up. And since a lot of the USA is fairly empty, cities would need multi-story hamster housing.)

More seriously, he turned to other approaches. In his view, it was a mistake to stop developing nuclear power stations, as these are a relatively efficient and low-polluting source of energy.  Renewable sources all have some problems: biofuels derived from plants are not efficient, (and recently there has been growing concern about plants grown for biofuels displacing plants for human food). Hydroelectric power is well established, but, in the US, in East’s words, “we are tapped out.” Solar power is still at an early stage, although that hasn’t stopped East installing solar panels to generate energy at home nor Actel investing in 200 kW of panels. Wind power is also at an early stage; East calls both “expensive, unsightly and unreliable.”

(Editor’s note – there is a renewable technology that East missed: tidal stream, where underwater turbines are powered by the flow of the tide. By coincidence, the first UK experiments with tidal stream began a week after East’s talk.)

Since there is no immediate solution to generating more energy, then the answer is to use less – and Actel happens to have just introduced some FPGAs, the Igloo family, that use considerably less power than conventional devices. So all Actel has to do is sell four million chips a quarter that consume 0.1 watts less than the competition, and they are saving twice the 200 kW of the company’s solar panels.

Sitting through East’s presentation was rather like being at a revivalist church, only with the true religion being energy saving. That same evangelical feeling came from Chris Rowen of processor-core IP company Tensilica. Rowen also reviewed the issues facing the US and found that electronics, including PCs, all forms of telecoms, consumer electronics, and corporate use, accounted for 6% of electrical consumption, equivalent to the output of thirty 800 MW power stations. Again, he can see no alternative to reducing power use, and, as Moore’s law no longer helped processors, parallelism was the way ahead. But in Rowen’s view, parallelism is more complicated than just shoving an application onto a multicore version of an existing architecture. He wants computing split clearly between operations at the data plane and the control plane, with the data plane being more amenable to significant parallelization. By configuring processors and optimizing the instruction set, Tensilica claims to be able to achieve energy savings of up to 40X, with throughput up by as much as 10X, both compared to an ARM 11, for data intensive applications.

Server farms, which, you will remember, John East said were burning 1.2% of the US electricity output, were discussed in two other presentations: those by 3Leaf Systems and Qimonda. There is a two level problem with farms: individual servers within the farm are inefficient and burn significant amounts of energy, and, consequently, the farm requires even more energy to displace the heat the servers are creating.

3Leaf’s founder and CTO, Bob Quinn, who was talking as part of a Hypertransport Consortium pitch by describing how they use the hypertransport standard in building their virtual servers, sees virtualization as one way to reduce the power consumption of server farms. As virtualization uses one box to replace several, appearing to each user application as a dedicated I/O box or application processor, potentially a farm needs many fewer boxes or grows less quickly.

Qimonda’s Tom Trill was positioning his memory company’s recently announced buried wordline DRAM as a palliative to the energy consumption of the boxes. This technology separates the bitline and wordline within the memory cells, simplifying the cell, improving performance, and reducing power consumption. For servers, lower power consumption by the memory will reduce that needed per box and also will reduce the cooling energy, Trill claimed.

Bringing things to a higher level of abstraction than just the components in a system, Rick Zarr of National Semiconductor felt that, while chip designers and silicon companies have a significant role to play in producing devices that were intrinsically lower powered, there is a responsibility on the system architect to design products that are energy efficient. Efficient design at both subsystem level and at system level in all electronic products can produce significant savings, and, as Zarr pointed out, even small energy savings in products selling in millions of units can help make a difference.

While all the speakers were there to pitch companies and products, it was clear that for many speakers, particularly East and Rowen, there is a personal commitment to energy saving. Both engineers, they are not using global warming and a green agenda as a marketing gimmick, but both are concerned that they, their customers, and the end users should play a significant role in helping to cut back on the way in which humanity is wasting the resources of the planet.

Leave a Reply

featured blogs
Aug 16, 2018
Learn about the challenges and solutions for integrating and verification PCIe(r) Gen4 into an Arm-Based Server SoC. Listen to this relatively short webinar by Arm and Cadence, as they describe the collaboration and results, including methodology and technology for speeding i...
Aug 16, 2018
All of the little details were squared up when the check-plots came out for "final" review. Those same preliminary files were shared with the fab and assembly units and, of course, the vendors have c...
Aug 15, 2018
VITA 57.4 FMC+ Standard As an ANSI/VITA member, Samtec supports the release of the new ANSI/VITA 57.4-2018 FPGA Mezzanine Card Plus Standard. VITA 57.4, also referred to as FMC+, expands upon the I/O capabilities defined in ANSI/VITA 57.1 FMC by adding two new connectors that...
Aug 14, 2018
I worked at HP in Ft. Collins, Colorado back in the 1970s. It was a heady experience. We were designing and building early, pre-PC desktop computers and we owned the market back then. The division I worked for eventually migrated to 32-bit workstations, chased from the deskto...
Jul 30, 2018
As discussed in part 1 of this blog post, each instance of an Achronix Speedcore eFPGA in your ASIC or SoC design must be configured after the system powers up because Speedcore eFPGAs employ nonvolatile SRAM technology to store its configuration bits. The time required to pr...