So how Many Phases do I Need?
The answer to almost any engineering question is, “It depends”. The main thing to realize is that a motherboard meeting Intel’s design guidelines should always work, and the actual implementation of the VRM is completely irrelevant. For most users, the number of phases in the buck converter only makes a neat-sounding marketing feature.
However, if you plan to over-clock the board or don’t provide proper cooling, then there can be advantages to selecting a power module that can split the high current across a large number of buck stages and stay within the voltage margins required by the CPU. Remember, to over-clock, the VRM has to be capable of providing the higher maximum current and slew rate, regardless of the number of buck stages used in providing this current. Using 100 phases with very small transistors will result in a low-noise power supply, but it may not deliver the necessary power.
Do I need this Extra Copper?
Another marketing differentiator has been the promotion of double-thickness copper in the motherboard power planes, moving from 1 oz to 2 oz. What this means is that instead of a PCB metal layer thickness that corresponds to 1 oz of copper rolled out into a sheet that is one square foot, you double the thickness. There is almost no downside to this, except for the cost (and a tiny bit more weight). More metal will dissipate more heat and lead to lower electrical noise.
While Intel can be very specific about the type of VRM they want, the motherboard layout has too much variation for a chip vendor to create strict rules. There are also a lot of ways to remove extra heat from components on a motherboard without increasing the thickness of metal layers. Still, unless there is a large cost delta, it’s probably worth having the 2 oz metal layers, just to spread out the heat a little better and perhaps bring the component temperatures down. Of course, this discussion will need to be revisited when some vendors try to claim an advantage with 3 oz copper.
Is this Topic Interesting?
As Loyd’s Improbable Insights continues to grow, there will certainly be a period of time during which the content is dialed in to match the interests of his readers. We’d welcome feedback about whether engineering-oriented articles like this provide useful information and whether you’d like to see even more detail. A topic like multiphase power is very complex, and there are a lot of gross simplifications in this high-level overview. It was tempting to go into more detail on some important issues we left out of the discussion. Let us know what you think.
Loading tweets...
26 comments
2 pings
Mike says:
August 28, 2009 at 5:13 pm (UTC -7 )
Great article, I am involved in computer controlled forklifts. My hobby is reading the technical aspects of computers or anything electronic. Thanks.
Steve says:
August 28, 2009 at 9:11 pm (UTC -7 )
I did like the article but would have loved more detail.
At what point does increasing the copper thickness reach the point of diminishing returns? That is probably a very difficult question to quantitatively answer.
The same with power phases. For a given power envelope, say a CPU that is consuming 140 watts, and a given MOSFET, say one that is commonly found on high end motherboards, how many phases are optimal? What is the point of diminishing returns?
Bert says:
August 29, 2009 at 12:32 am (UTC -7 )
Technicalities aside, I found this article very interesting. I had actually been wondering if the extra thickness copper layers would be wort their weight, so this came at a good time!
Scott Gardner says:
August 29, 2009 at 10:55 am (UTC -7 )
Guys, I’m glad you liked the article. It’s hard to tell whether a topic like this can generate much interest when it requires a bit more tech savvy to read.
As for diminishing returns, the extra converter phases won’t make any difference if the motherboard meets the required power specs without them. There might be a slight thermal improvement. The best way to quantify this (and the copper thickness question) would be to instrument some boards with temperature probes. You can also use an infrared camera to visually see the thermal improvement. I’m sure the high-end mobo vendors (like Asus and Gigabyte) have done these tests.
Scott
Jack says:
August 30, 2009 at 6:57 am (UTC -7 )
i loved the write up im glad to see a bit more in depth, with all the new savvy tech gear coming out it is hard to find somewhere you can read about it straight up no gimmicks and get gown to the bottom line is it better for me (end user) well done all at improbable insights (glad i found where loyd and co went too)
Scott Gardner says:
August 30, 2009 at 6:14 pm (UTC -7 )
Jack, I’m glad you liked the article, and it’s great that you’ve followed Loyd. Hopefully he keeps growing as a publisher. Maybe someday he can buy Ziff Davis — everyone else has…
Scott
Craig says:
August 31, 2009 at 1:43 pm (UTC -7 )
Thanks Loyd and Scott for a great article. For me this article had the right amount of technical detail that I could follow the answers with out feeling lost. I consider myself to be an above average computer user who isn’t afraid to dig around inside a computer but don’t have much in the way of a engineering background. Perhaps it would be possible, with an article like this one, at the appropriate spots to have a link for more detailed descriptions, for those with higher aptitudes in this area, thus satisfying compugeeks (like me) and technophiles alike. Your doing a great job and it’s good to hear you’ve landed on your feet.
Have Fun!
Scott Gardner says:
August 31, 2009 at 2:00 pm (UTC -7 )
Craig, I’m glad you like the article, and you make an excellent point about providing links for more detail. I should include a few links next time. In this case, a pointer to Intel’s design guide would be helpful:
http://www.intel.com/Assets/PDF/designguide/321736.pdf
In the future, I might talk to the mobo vendors a little more and provide some links. Hopefully they would be willing to share some of their engineering data.
Scott
Jeremy says:
August 31, 2009 at 3:25 pm (UTC -7 )
Good job explaining what is going on with these wonderful new engineering terms that have been translated into PR. There is a lot of media out there that tries to explain technology to those with little background in electronics of any sort. I love to see that there are still people unafraid to dig into the more technical aspects and publish what they learn without dumbing it down much. Keep it up!
Tim Handley says:
August 31, 2009 at 9:04 pm (UTC -7 )
Thanks Scott, I’m in the marketing team at Gigabyte and it’s great to hear some positive opinions about our motherboards. From a motherboard vendor’s point of view this is an excellent explanation of where we are in the industry and why we are pushing extreme features like 24 power phases and 2 oz copper PCBs in order to differentiate ourselves. You hit the nail right on the head with your conclusions about better heat dissipation and cleaner signals with 2oz copper PCBs. This has long been a server motherboard spec and we believe that it will be a consumer board industry standard soon – around 80% of Gigabyte models will have 2oz copper PCBs after the P55 chipset launch next week.
We agree with your conclusions about large numbers of power phases being useful for overclocking; part of the dynamic energy saver feature on Gigabyte boards is an LED display that shows how many power phases are operating at any one time. On the GA-P55-UD6 that has 24 power phases, a mere 4 LEDS are lit up most of the time; only during heavy CPU overclocking will you see 20 or 24 LEDS burning.
One may argue that today’s high-end motherboards are way ‘over-spec’, but that is the nature of the current PC component market if you consider that there is no growth in the entry level anymore due to netbooks and low cost OEM desktop solutions. The growth markets for PC components are in the high-end and mainstream markets, e.i. expect to see a lot more excessive specifications on future PC components!
Scott Gardner says:
September 1, 2009 at 6:29 am (UTC -7 )
Tim, it’s good to hear from you. In my email I proposed a follow-up article that walks through some of your thermal engineering data, or Loyd and I can just instrument our own boards. Perhaps some of the readers could chime in to discuss whether such an article has value. As I recall, some of the mobo vendors (Gigabyte perhaps?) already publish Youtube videos that show a thermal imaging comparison. Perhaps an interesting article would explain the thermal plot to analyze the contribution from various components…
BTW, I think it’s great that PC vendors directly join the discussion and contribute to Loyd’s site. Hopefully we’ll see you here more often…
Scott
Scott Gardner says:
September 1, 2009 at 6:36 am (UTC -7 )
Jeremy and Craig, thanks for the feedback. It’s good to know that I can dial up the techno content without making the article too difficult to follow. What I tried to do with ExtremeTech content (and now Loyd’s site) is to make sure I define my terminology first and then go ahead and get more hard-core. While most readers may not have engineering degrees, anybody reading sites like this probably has a passion for technology and will appreciate some exposure to the fundamentals.
Scott says:
September 1, 2009 at 6:49 am (UTC -7 )
Tim, I’m glad you didn’t take offense at my digs toward the marketing guys creating hype and hyperbole. Sometimes I forget that people may not realize I’ve spent a portion of my own career helping with marketing or managing marketing teams. I make fun of myself as well.
Perhaps a lighter article would discuss the role of the product marketing folks who need to turn extremely-technical features into something that can be explained to consumers, while helping them understand why they should care. It’s an interesting job, though marketing guys get only slightly more respect than lawyers.
Scott
Markeyse says:
September 1, 2009 at 9:34 pm (UTC -7 )
Good article. I think Loyd is going to own a big site very soon. lol.
Tim Verry says:
September 4, 2009 at 2:52 pm (UTC -7 )
Good article Scott! The diagrams definitely brought back memories of learning about electricity in school
.
So what you are saying is that at a certain point… adding more power phases doesn’t necessarily help in any (greatly noticeable) way most consumers would operate their computers. However, does having these “over-spec-ed” boards… does the overhead resulting from additional power phases result in easier overclocking?
Also, thanks for the info on the question regarding the denser copper PCBs.
Keep up the good work!
Eric says:
September 5, 2009 at 8:10 am (UTC -7 )
Great piece; more of these would be good. For a guy who went the MA route rather than an MS, it was still digestible and informative.
Took me back to my high school electronics vocational class 20 years ago.
Scott Gardner says:
September 7, 2009 at 4:26 pm (UTC -7 )
Guys, I’m glad you can get so nostalgic over power electronics. As for your question, Tim, yes, the overclockers are really the only ones who “need” a gratuitous number of phases in the voltage converter. Loyd’s the guy who gets all the boards from vendors for review, so I actually haven’t seen these new motherboards. I deal more with semiconductor specs. However, the Gigabyte marketing manager mentioned that they even have a display that dynamically shows the number of phases being used. It may not be necessary, but it sounds cool…
Scott
Yiannis Manousakis says:
December 11, 2009 at 2:54 am (UTC -7 )
Very interesting article, thank you.
Yiannis.
Scott Gardner says:
December 11, 2009 at 8:50 am (UTC -7 )
Yiannis,
Glad you found this an interesting topic. After getting distracted by other projects, perhaps it’s time to write some more (non-controversial) content for Loyd’s site. Feel free to toss out some suggestions for technical topics…
Scott
Jim Sands says:
June 2, 2010 at 6:50 pm (UTC -7 )
I always wondered what was up with all of this suspecting that the CPU manufacturer specs if met was all that was necessary. True to a point. Temp is always important and now i know what to look for based on the application I am building for, thus the engineering constant: “It depends”. Thanks!
Scott Gardner says:
June 7, 2010 at 9:19 am (UTC -7 )
Jim,
Thanks for the comments, and I’m glad people are still reading this article. Yeah, I’m a bit chip-centric in my outlook, since I’ve spent most of my career in the semiconductor industry. As I mentioned in the article, the motherboard vendors have a tougher job trying to differentiate their products as more of the technology gets integrated into fewer and fewer components.
Geoff says:
June 9, 2010 at 1:15 am (UTC -7 )
Great article Scott! I stumbled in here from Google and came across this article which was just what I was looking for as an end user, computer enthusiast and EE student. I’ve bookmarked this for future reference. I hope you publish some similar articles in future
Geoff.
Ruzveh says:
August 20, 2010 at 6:21 pm (UTC -7 )
I am still left with a doubt here guys. I am no voice in mechanical spec. I would like to know how much power phase design mdb do i need? I have selected mbd from gigabyte based on my spec i.e. P55A-UD4P which has all the features and perfect for me but comes with only 12phase as against their 24phase. Would it have been a 24 i wud go for it without any doubt but now i am again open for new mdb which has all the features which that mbd has + better power phase design, r u listening gigabyte marketing team member?
Actually i plan to run this mbd on my i7-870 and i have tons of add on cards and other requirements for which i have replaced my smps atleast twice a year. I run 2 different tv tuner cards, 2 sound card, i plan to use only 1 graphics card anytime which is high end like 460 or 5850 or later. Rest all other usb add ons and features will b used. So i assume that i require a better power phase design and more PCI-e 1.0 ports as against graphics ports. I dont need SLI that is why i cant go for UD-6 or UD7 since they r power packed for games only. Their additional PCIe ports r left unused for me.
I am confused here. Please help.
I have seen as high as 48 phase power design form Asus with some 50,000 hrs run time at 100c with their latest mbd. Isnt it exciting Gigabyte
Scott Gardner says:
August 22, 2010 at 11:01 am (UTC -7 )
Ruzveh,
You should be fine with the 12-phase board if you’re not overclocking. The impact of adding lots of peripherals is more likely to stress your power supply — not the voltage regulator to the CPU. The power supply rails that connect to the VRM are likely not shared with supply rails that connect to add-in boards. As I mentioned in the article, the problem is the instantaneous current requirement for the CPU. The VRM in your board should be spec’d to supply this current. Once you over-clock, all bets are off.
Scott
Scott Gardner says:
August 22, 2010 at 11:09 am (UTC -7 )
BTW, Ruzveh, I read an interesting article on Tom’s Hardware (where Loyd does some freelancing), and Thomans Soderstrom took some power and temperature measurements of an i7-870 board being overclocked. The power jumps dramatically as the voltage is increased to achieve the faster speeds. More importantly, the VRM temperature (above ambient) doubles. Apparently, this is enough to destroy some motherboards, though the author mentions that some VRM’s included a thermal shut-off (similar to the protection circuitry in a CPU).
Here’s the link to the relevant page in the article: http://www.tomshardware.com/reviews/p55-motherboard-overclock,2460-7.html
Scott
TomBrooklyn says:
November 23, 2011 at 4:03 am (UTC -7 )
This is useful information. Most readers, if like myself, don’t need highly technical explanations. But we do want to get an understanding of what manufacturers are trying to sell us.
[Info] says:
October 18, 2009 at 2:23 pm (UTC -7 )
[...] [...]
ASUS P7P55D Premium Motherboard Review - says:
November 29, 2009 at 3:47 am (UTC -7 )
[...] [...]