CPU Interview with Doug Dodson
Home > About > CPU Interview with Doug Dodson
Computer Power User
September 2005 • Vol.5 Issue 9
Page(s) 103-104 in print issue
Of all the components that are in a PC, power supplies have probably been the most ignored. But from Extreme Edition CPUs to SLI graphics, the power requirements of enthusiast systems in particular have recently skyrocketed. We cornered Doug Dodson, the president and founder of PC Power & Cooling, the only American power supply manufacturer and arguably the finest maker in the world, to explain some of the inside tricks of his trade and what PSU buyers should really know before buying.
by William Van Winkle
CPU: Of all things, how did you get into PC power supplies?
Dodson: I was an electronic hobbyist since I was a kid, and my background was in economics and commodity trading. I had one of the first XT computers. The computer was extremely noisy, so I created a special large-diameter cooling fan to take the place of the small fan that was inside of it and just did it for myself to reduce noise. Then people came and saw it. They wanted one, and one thing led to another. So I started making these things in my garage in 1985. Then I realized that the entire power supply needed to be redesigned because it was just poorly designed by IBM in the early days. The fan would cavitate. There was no real airflow coming out, so you had a ton of noise and no cooling. We started developing power supplies that were both quieter and cooler than the stock XT power supplies and we started getting customers for both attributes.
CPU: How could people perceive a difference in power quality back then?
Dodson: We were the first ones with independent regulation, meaning that we gave people rails that were independent of each other so that they could be better regulated, rather than the cheap design, which is combining two rails together and then neither one of them regulates very well, but you save on components. We've always opted for really tight regulation and much cleaner power by using bigger capacitors to filter the DC output onto a much cleaner output than what the other guys had. Of course, these capacitors cost more. We used triple-stage capacitors. On the input side, where you're converting the AC to DC, we also used bigger capacitors, and the effect of that was we were able to better handle sags and surges and basically had a better holdup time. Rather than the typical 10 milliseconds, which is what most power supplies were doing, ours were producing power for as much as 300 milliseconds. So people noticed when they had our products that when the lights would dim and the blackouts would occur and they'd look around the room--like a student's room or a university or school where they had 10 or 20 computers--and notice every single one of them was down except for ours. Most outages are anywhere from 10 to 300 milliseconds. So if you can give people enough capacitance to ride through that, you save a lot.
CPU: One of your apparent pet peeves has to do with PSU temperature ratings. Why exactly?
Dodson: People don't understand that the actual temperature that the power supply and its components see is far higher than room temperature. When a power supply is rated for 25 degrees C, like most of our competitors, especially the ones that are designed for home use, that's a totally fabricated number put together in order to exaggerate the wattage claim.
CPU: Why fabricated?
Dodson: There is no way in the world that the air going to the power supply is going to be 25 degrees Celsius in the real world because with a 15-degree buildup inside the machine first, before the air reaches the power supply, that means that the room temperature would have to be 10 degrees C. Now you're talking about it being in the 50s Fahrenheit. What these guys selling power supplies rated for 25 C are saying is that as long as you use the power supply in a room that's 50 degrees Fahrenheit, you're OK. But if you were to actually go through the de-rating that they keep very secret and re-rate those power supplies that are rated at 25 C at a more realistic 40, you'd have to cut the wattage rating by at least 33%, compared to what they're advertising. You have a 400W power supply, but the fine print is, yeah, at 25 degrees C. At 40 degrees C, it's a 300W power supply, and the public doesn't understand that it's not getting an apples to apples sort of rating when it is out there, buying it in the store, because these manufacturers don't even specify what the ambient temperature is for full load. That's one of the most common deceptive practices in the power supply business.
CPU: But if this were such a big deal, shouldn't it have been more publicized by now? Does it really matter in the real world?
Dodson: For the last five years, companies have been marketing power supplies as being 400W and 500W on this basis. And they've been getting away with it without any sort of testing or validation by anyone. The reason they got away with it is because computers, up until recently, only needed like 300W. Since they only needed 300W and were actually capable of 300W, you don't get caught if you're saying yours is 500W. You only get caught when suddenly there's technology that really needs 500W. For SLI you've got like 200W just for the video and another 120W for the CPU. So when the temperature inside the power supply starts getting way high, and it's not designed for that, before you know it power supplies are just crapping out left and right. They were never designed to produce a realistic, true 500W in the first place.
CPU: So if we buy a 500W PSU with accurate thermal considerations, we're good to go?
Dodson: Actually, it really isn't the wattage of the power supply that counts today, because if it's all on the wrong rails, it doesn't do you any good. The only rail that matters whatsoever in today's systems is the 12V rail. That's because the video cards, CPUs, and drives are all being run from it. If the power supply design isn't recent and doesn't take this fact into account, it doesn't matter if you have a 500W power supply, if it only has 20A of +12V power. You'll see 600W power supplies that are down at 24A, whereas our 510W puts out a continuous 34A. That's 50% more power than a power supply that, by wattage, is rated higher than we are.
CPU: Should people focus on peak or continuous power ratings?
Dodson: You want to zero in on the peak rating of the 12V output and how long that rating is for. If you get a 15-second rating on that and you have plenty of amps, then you know that even if you have eight to 10 hard drives, you're going to have no trouble bringing them up simultaneously, with no need for sequencing or a second power supply. But a one-second peak isn't going to get you anywhere. If it puts out the power for one second, that's not enough time for the drives to get going.
CPU: But drives have very little draw.
Dodson: Well, drives on startup have 300% more draw than they do after they've stabilized. So even though they may not require much once they're running, like say they require an amp of 12V, which would be 12 watts. Well, at startup, they could require 3A, which is 36 watts. Now if you have, say, five of them, that's a 180W peak load at startup, which would be enough to bring a lot of power supplies to their knees.
CPU: Modular power supplies seem to be gaining a lot of traction in the market, but PCP&C doesn't make any modular designs. Why?
Dodson: Yeah, it's real popular, but the industry hasn't done anything to educate people about modular plugs. The pins that are used in the modular plugs have a very low capacity to pass current. You're losing power through those pins. It's electrical resistance between the male and the female part of the pin, to the extent that the voltage drop in just the pins is equivalent to about two feet of wire. The effect is that modular power supplies, everything else being equal, are capable of about 10% less power than power supplies without modular plugs. And that's under ideal conditions. In real life it gets worse because the pins loosen, corrode, and burn. Over time the resistance builds up. A year down the road, a guy could be running his system and all of a sudden it stops working reliably, and he has no way of knowing that the reason is because the pin inside that modular plug has become corroded or burned. This technique has been out for about six months, but people are going to find that the reliability is a massive failure point. Also, especially given the way people pull their harnesses around and tie them down, it creates a lot of stress on those pins. So instead of the pins having full 100% contact surface to surface, in most cases those pins are only touching maybe 10% or 30% of their surface area. What's rated for 5A going through those pins, in reality if the pins aren't making good contact, your rating's down to maybe 3A, and with that much constriction, those pins just heat up, and it gets worse. For real pros in the industry, no way would they specify that kind of arrangement. This is a consumer-oriented gimmick.
CPU: We've heard speaker sets that are not THX certified that sound better than some that are. So how important is SLI certification or any other certification in the power supply space?
Dodson: I think SLI certification is essential. First of all, you don't pay for SLI, so there is no THX sort of analogy. For anyone that has a power supply good enough to meet the spec, there's absolutely no reason in the world why they wouldn't want it certified.
CPU: What are your thoughts on fanless power supplies?
Dodson: The fact that it's fanless means that you have to de-rate right off the bat because you have to dissipate the heat somehow, and it's clear that a fan helps in that heat dissipation. So if you have a fanless power supply, you're basically talking about one that has to be way de-rated compared to one without a fan, meaning it costs a lot more per watt. As for noise, I would turn it back around and say that it's somewhat pointless in the sense that you're only as good as your quietest component. If you have a CPU cooler or a video card with a fan or any number of other sources of noise, it seems to me it makes more sense to have a quiet power supply fan than no fan at all. Because with a power supply fan, at least you're not going to be paying a whole lot of extra money for a more unreliable design that doesn't really decrease the overall noise of the system. The only way that a fanless power supply would make any sense is if the entire system was fanless. Actually, we got some in here for testing, and they crapped out about 100W less than what they were rated for, just at room temperature.
CPU: Your product line is unusually Spartan, considering you cater to enthusiasts. You don't even have LEDs.
Dodson: No, we don't do gimmicks. We don't do modular plugs. We don't do LEDs because it necessitates using a cheaper, unreliable fan. We do believe that a power supply should look good, and we address that by giving ours a black finish and cable sheathing.
CPU: Will an SLI-certified power supply handle whatever ATI's Crossfire is going to require?
Dodson: Yeah. If it can run SLI, it should be able to run Crossfire. The only caveat is that the harness needs to have connectors that are adequate to plug into ATI's cards because they're different than NVIDIA's cards.
CPU: Is there such a thing as overbuying a power supply you plan to keep and migrate through multiple configurations?
Dodson: No, and I'll tell you why. Probably the biggest myth of all is that when a person buys a power supply that is higher wattage, that it uses more juice from the wall and will cost them more in electricity. In fact, our 850 actually costs $65 less a year in electricity to run than a 500W power supply. So in a five-year period, you'd save $330, which would be more than the difference in price between our 850 and just an ordinary 500W power supply.
CPU: Is that if you leave it on 24/7, 365 days a year, in California?
Dodson: That's assuming 10 cents per kilowatt hour, which is a pretty much average rate. And I believe it's figuring just letting it run because that's more efficient. The whole reason behind that is that this new 850 is 85% efficient, instead of the typical 70 or 72%. To run a particular computer, you're pulling less power from the wall because more power is being converted into DC power, instead of being converted into heat. That's your efficiency factor. So the more efficient the power supply, the less money it costs to run a system.
CPU: Any comment on that age-old debate about whether to turn PCs off at night?
Dodson: Our position is let it run. That's for voltage and temperature stabilization. Anytime you can keep the computer components at a stable temperature, it's better than expansion and contraction. That's what wears them out.
CPU: Set-top PCs are coming at us very soon. Are there going to be limits imposed upon the performance of those boxes from a power standpoint?
Dodson: Absolutely. Due to size constraints, you're basically looking at practical limits in the 250W range from those small power supplies. You're never going to see an SLI pizza box.
CPU: How would you like to see PC power supplies evolve in the coming years?
Dodson: One of the things that has bothered us about power supplies is this almost insane lack of standardization on connectors. They've done the modular plugs to try to get around that, but years ago, there was one standard, and I think that the entire industry would be helped if we get back to more standardization on the output harness. There are too many permutations and to no advantage. Once you have a 24-pin connector and an 8-pin connector and this SLI connector, my position is, ‘OK, enough's enough, boys. You can make this work.'
As far as form factor is concerned, right now we have an 850W power supply. That one's going to be up to a one kilowatt soon. At that point it will probably be good for a number of years. We're just making some minor modifications. But what I'd like to see is that this new form factor, which is basically an extended ATX, be used in a whole host of server and workstation applications because it's a form factor that's quite efficient, in the sense that it doesn't require all new cases. You can use your old ATX case; it just needs to be 20 inches deep. And yet you can have twice the power that you have today. With that much power, you should be able to handle just about anything on the horizon from a graphics or server standpoint. But there's no mistake about it: We know from video companies that they're going to need that kind of power.
Copyright © 2005 Sandhills Publishing Company U.S.A. All rights reserved.