31 March 2008

Lifetime Electricity Costs for High-end Video Cards

So a few weeks ago I discussed how we're going to need a greater emphasis on low power consumption electronic design in the future. I thought it would be helpful to actually put down some numbers and see how big a cost power consumption is for high-end computer equipment. On some equipment, especially that used in server farms such as hard drives, power draw is already an important metric. For other components, we often are not even given consumption numbers by the manufacturers.

High-end graphics cards are becoming particularly power hungry, as this chart by anandtech.com shows. The two current best-performance at a reasonable price-point video cards are the 8800GT by NVidia and Radeon 3870 by ATI. Even idling, these systems chew through an impressive amount of power — 165 W in the case of the NVidia product and 125 W for the ATI one. I don't know if anandtech.com's methodology is normalized for the efficiency of the power supply or not, but regardless they are burning a lot of power for doing next to nothing. Average residential electricity prices are up to 10.4 ¢/kW·h now in the US for 2006.

Let's assume that the lifetime of a card is always on at idle for two years. "Idle" in this case would basically extend to using any 2-D application, such as browsing the internet or using a word processor. Only 3-D accelerated games are going to stress these systems to any significant degree.

Video Card

NVidia 8800GT

ATI Radeon 3870

Initial Purchase Price

$260

$255

Idle Power Consumption

165 W

125 W

Expected Lifetime

17500 hours

17500 hours

Lifetime Est. Power Consumption

2187.5 kW·h

2887.5 kW·h

Lifetime Electricity Cost

$300.30

$227.50

Total Cost

$560.30

$482.50


As we can see, the operational cost of these two cards is roughly comparable to the purchase price. Even with the modern price of gasoline, automobiles don't have such a high proportion of their lifetime cost associated with fuel.

Both of these cards come with 512 MB of fast memory spread out over eight chips. However, a single buffer at 1280x1024 pixels with 32-bit resolution requires less than 6 MB of RAM, so there's no need to maintain all that memory powered on for the vast majority of computer applications. One chip should suffice for a triple buffered display.

Similarly, it should be technically possible to clock down the processor and bus speeds dynamically to reduce the power consumption of the GPU. Alternatively, one could embed a slow GPU for 2D applications. Most motherboards are available with on-board video on the Northbridge chipset which is just fine for web browsing (useful if you ever want to flash the BIOS on your video card BTW). The marginal cost of on-board graphics is probably around $5 to the manufacturer.

I find it somewhat surprising that neither of the major graphics manufacturers have tried to radically improve the power performance of their cards. There is, potentially, a major competitive advantage to be had. For example, if ATI was to spend $5 per card and drop the idle power requirement to 1/8th that of the Nvidia model, and advertise that fact and the estimated savings aggressively, they could recapture a lot of the market share they've ceded since the heyday of the Radeon 9700 Pro.

16 March 2008

Carbon Trading, Bubble Hysteria

In the past, I thought that carbon trading of the style proposed by the Kyoto treaty could be a positive way to affect change, both from the point of view of climate change and peak oil. I have gradually come to change my mind, and I now favour a vanilla carbon tax with no loop holes. My decision was largely made watching the fallout from the dot-com bust, and now the US mortgage security shenanigan's.

Anything that Wall Street can game to enrich themselves, they will game. These crony capitalists with their derivatives and good-old-boys compensation schemes are really the enemy of free market entrepreneurship. If you bought $50 puts on Bear Sterns on Monday (10Mar2008), you gained a lot of money, but no wealth was crated.

I fail to see any advantage in giving Wall Street access to the carbon market.

A lot of people suspect that the recent run-up in commodities is largely due to money flowing out of mortgage securities and into commodities. I am not convinced of this, due to a number of factors.

Past pump and dumps in commodities — such as nickel — can work because you can store an entire years worth of the world's nickel production in a single large warehouse. On the other hand, a day's worth of oil production is roughly a cube 300 m on each side. It's very difficult to take oil out of the system unless you are a national oil company.

Furthermore, demand remains remarkably inelastic. Predictions of any tipping point where demand suddenly falls off at some price-point haven't panned out. When oil is consumed, it's really gone.

In addition, a huge hunk of the recent run-up in crude oil prices is simply due to the devaluation of the US dollar. The proof is in the US dollar index. So yes, the US is getting hosed on their oil consumption but the majority of the world's consumption is pretty well hedged against this rise.

China even subsidizes the cost of oil to their citizen-consumers. They have to do something with their dollar reserves. So even if we see a lot of demand destruction for petroleum from the USA it's not clear if that will really hammer the price of oil back down to $80 for a sustained period. The twin inflationary and deflationary pressures currently at war between the US Federal Reserve and Wall Street respectively make that an extremely difficult call to make.

I know one thing for sure. I will never hire someone with an MBA on their resume.

This brings up another question, namely is there potential for a bubble in investment in the so-called 'Cleantech' sector?

The world economy is in a slow transition from fossil fuels to alternative sources of energy, true or false? If you answer "true," then your only reasonable explanation for a bubble would be that the alternatives are growing at an unsustainable rate relative to the increase in the price of fossil fuels.

Unlike say, Pets.com or granite counter tops, a wind turbine or photovoltaic power has intrinsic value. They produce electricity, which is a very high-quality form of energy. I can calculate the net present value of a set of photovoltaic panels to a rather high degree of accuracy (~10 %), merely by noting the climate in which they are installed and their age.

The gap between the cost of doing work with oil as your energy source compared to electricity continues to enlarge. Consider, with electricity at $0.09/kWh, natural gas futures at $10.00/MMbtu, and oil at $111/bbl, the value of switching to electrons may pay back quickly. Note: these numbers are changing as fast as I can type this article.

Energy

Currency

Energy Cost

(US$/GJ)

Energy to Work
Efficiency

Exergy Cost

($US/GJ)

Electricity

25.00

1.0

25.00

Natural Gas

9.50

0.4

23.70

Crude Oil

17.35

0.35

49.55


At this point, electrifying train tracks or heating your home with a heat pump looks really good going forward (natural gas isn't nearly as fungible as oil). Look at it this way, there are 153 million employed people in the US, and they consume 19.6 million barrels of oil a day. That's $14.20/day or $5190 a year per (money earning) person at current prices. That's a lot of Starbucks.

There is a potentially enormous sum of money to be made in weaning North America, Japan, and Europe off the oil habit. It's not going to be easy since there is still a massive amount fossil fuels in the Earth's crust. The saving grace of the alternative energy industry is that its costs will go down with time whereas fossil fuel companies will have to extract poorer and poorer quality resources and hence become more expensive.

Of course, not everyone involved in cleantech will be idealists. A number of companies will be formed with the express aim of relieving investors of their capital. These fraudsters will primarily aim at people conceited enough to believe that they understand science, but lack the actual formal education to evaluate what they are seeing in numerical terms. I'm looking at the dot-com millionaires here. Beware the Rube Goldberg machine, or the company with salaries a much higher proportion of their expenses than equipment.

I will say, from personal experience, doing research in a corporate environment where every line of research has to have an immediate application and money for equipment is tight isn't very efficient compared to government funded labs. Now the bureaucracy, well...

11 March 2008

Squestration in the Oil Sands

Soooo.... last year the federal government of Canada introduced a bunch of new environmental programs. This year, they threw a lot of that out the window. Now we have a new environmental program: legislating projects that produce large quantities of carbon dioxide to employ sequestration. These large sources are coal plants and oil sands developments. The obvious loophole for everyone to observe is that it only applies to projects started after 2011, and there's evidently no grandfathering.

I'm not sure I believe whether they Conservative government actually intends to go through with this. Afterall, they are a minority government and while the opposition has no stomach for a new election, they aren't likely to last until 2011. The proof will really be in the activity in the oil patch. If they all rush to start projects before 2011 and have nothing scheduled after that, then maybe the Conservatives are actually serious.

Another question that crosses my mind is the quantity of good sequestration locations in close proximity to the main oil sands patch by Fort McMurray. Alberta is, generally speaking, a big sedimentary basin but the Northeast portion of the province is somewhat different if my memory is correct.

Personally, I foresee the cost of sequestering 'dirty' fuel sources such as bitumen or bituminous coal being onerous. Alberta already has the highest electricity prices in the nation and prices can only accelerate with the introduction of sequestration.