Google Calls For Power Supply Design Changes

Raindance writes "The New York Times reports that Google is calling 'for a shift from multivoltage power supplies to a single 12-volt standard. Although voltage conversion would still take place on the PC motherboard, the simpler design of the new power supply would make it easier to achieve higher overall efficiencies ... The Google white paper argues that the opportunity for power savings is immense — by deploying the new power supplies in 100 million desktop PC's running eight hours a day, it will be possible to save 40 billion kilowatt-hours over three years, or more than $5 billion at California's energy rates.' This may have something to do with the electricity bill for Google's estimated 450,000 servers."

What in a modern computer actually uses 12V?

[Marxist Hacker 42]

In the old days, disk drive motors and fans. But many of these now run on 5V, hence the cheap USB-powered drive cases out there. Chips at CMOS power levels run at 3.3v, TTL is 5v, but hardly anything runs at 12v anymore. It seems to me that if they'd just pick their hardware carfully, they could run their entire server rack off of 5v+- rails.

Re:What in a modern computer actually uses 12V?

[Anonymous Coward]

Video cards use a ton of 12v power, enough that high-end cards get a dedicated connector featuring two wires of it.

Re:What in a modern computer actually uses 12V?

[ArcherB]

This is true, but Google is not throwing 7950's in their servers. These systems run with on-board video at best. Google has no need for a video card that can do anything more than text, as with all non-windows based servers. For that matter, after the first boot, there is no need for a video card at all.

Re:What in a modern computer actually uses 12V?

[Jahz]

Actually I would bet that Google servers DON'T have a video card, and that all of them have RJ-45 SOL support (or something like it). The reason being that Google has admitted that they fully embrace the commodity distributed server system. Google will periodically host talks at my university where they explain all this in [too much] detail.

Basically, when a machine fails, it is pulled from the rack and replaced with an identical machine with a cookie cutter image. Kinda like the Borg :)

When a box fails it is probably instantly detected by some machine monitor and taken offline (think: the 'crop' tenders in the Matrix I). The sysadmins arent going to waste time plugging a video cable into the rack... just pull it. Toss the box into a repair queue and let the tech's put a video card into it if needed. Remeber: 100's of machines fail for them every day. That's a fact from the Google talk in 05.

Re:

[drsmithy]

Actually I would bet that Google servers DON'T have a video card, and that all of them have RJ-45 SOL support (or something like it). The reason being that Google has admitted that they fully embrace the commodity distributed server system. Google will periodically host talks at my university where they explain all this in [too much] detail.

Who sells commodity servers without motherboard-integrated video cards ?

Re:

[dcam]

Someone who wants to sell to google? They must get through a *lot* of machines, this would give them some buying power.

Re:What in a modern computer actually uses 12V?

[RedWizzard]

With 450,000 computers, couldn't Google afford to get their own custom designed power supplies built?

They can and do - read their white paper [google.com]. What they are saying is that if everyone else did what they do the power savings would be huge.

Re:What in a modern computer actually uses 12V?

[poot_rootbeer]

Google is not throwing 7950's in their servers. These systems run with on-board video at best. Google has no need for a video card that can do anything more than text, as with all non-windows based servers. For that matter, after the first boot, there is no need for a video card at all.

Seems to me Google doesn't want to fracture the commodity hardware market into server-class hardware using 5VDC power and desktop-class hardware using 12VDC. One standard, applied equally across the entire range of products.

Re:

[Captain Splendid]

Mod Parent up please. That would definitely be the non-dick move, which is what we'd all like to expect from Google.

Re:

[kfg]

One standard, applied equally across the entire range of products.

The one used by the majority of DC electric devices, not just computers. The one compatable with existing external power supplies such as solar, home gas powered generators, your car battery, etc.

If motherboards were designed to run on 12v DC you could put a socket on the back of the case and jack into anything that gave you 12v DC. You could take your home desktop straight to the RV, boat, or cabin in the woods running off a turbine in the l

Re:

[mrchaotica]

You'd think that with all the parallel operations modern GPUs are capable of, Google would find a way to use them in their database (or at least use them for something)...

Re:

[Marxist Hacker 42]

Thank you- you're the first person to teach me something NEW. I didn't know GPUs still used 12V technology- especially since they all seem to output TTL level voltages.

Re:

[Bassman59]

Video cards use a ton of 12v power, enough that high-end cards get a dedicated connector featuring two wires of it.

Video cards with a disk-drive-type power connector always use point-of-load switch-mode supplies to convert the +12V to whatever voltages are needed by the chips on the board. Nothing on the board uses the 12V directly, except maybe the fan (if that).

They use the disk-drive connector because:

• The GPUs and other devices on the board use more current than you can push through the PCI/AGP or P

Re:What in a modern computer actually uses 12V?

[Anonymous Coward]

Almost nothing, but that is irrelevant. A modern PC uses so much power that it would be plain stupid to try and deliver it at lower voltages. The power is stepped down close to the actual load, because otherwise you'd need much heavier wires or lose much power to heated cables. That kinda is the point of the proposal: Every PC already has the necessary regulators because there's simply no other sane way to deliver the kind of power that graphics cards and CPUs consume. So what's the point in keeping the power supply complicated when the main consumers in a PC use the 12V line anyway?

Re:What in a modern computer actually uses 12V?

[Lonewolf666]

Modern CPUs run on core voltages of 1.5 v or less, depending on model. DDR RAM is 2.5V IIRC.
So you will have to convert most of your power from 5 V to something else. And if you have to re-convert anyway, 5V as intermediate voltage is not optimal. When converting to 5V, the voltage drop in the power diodes and in the wires to the mainboard eats a much higher proportion of the power than with 12V as intermediate voltage.
24V or even 48V would be even better. The auto industry is currenly starting to introduce 48V systems BTW.

Re:What in a modern computer actually uses 12V?

[Hirsto]

12V happens to be a sweet spot in terms of cost of the converter components as well as overall efficiency. Wire guage, voltage drop and capacitor size is significantly smaller than 5V or 3.3V primary supply. Think in terms of millions of units per month and compare the price of an NMOS FET and storage capacitor rated for 35V (safety margin in a 24V design) verses the cost of similar FETs and capacitors rated for 20V. In a synchronous buck design you can easily save $0.75 per converter section by using 12V rather than 24V and significantly increase conversion efficiency for free. Assuming a constant switch frequency the switching losses increase with the square of applied voltage, "I squared R" conduction losses in conductors will decrease with the square of current but the voltage dependant switching losses will dominate once the input voltage gets too high. For a given cost the overall converter efficiency is usually highest if your input voltage is relatively close to the output voltage. 12V to 3.3V conversion is significantly more efficient and less costly than 24V to 3.3V conversion.

Re:What in a modern computer actually uses 12V?

[NoMaster]

Actually, when I learned it, I was taught ~32V was the sweet spot between practicality, I^R losses, component size, and load type. Mind you, back then it was electrical / electromechanical loads (lights, motors, contactors, relays, etc) not electronic. And that was from a telco background (-48V), which made me wonder "well, why tell us that about 32V?" ;-)

But the point is, what the sweet spot is depends mostly upon the characteristics of the load - so it's wrong to come out with blanket statements like "12v happens to be a sweet spot in terms of cost of the converter components as well as overall efficiency". Yes, today, particularly with switchmode supplies and the actual maximum load V being 5v or less, it is. Tomorrow, when everything runs on 3.3v or less, it'll be closer to 5v~6v.

The other half of your argument only holds for certain types of power supplies too - but I'll give you a pass on that, seeing as you did explicitly state "synchronous buck" designs. It doesn't necessarily hold true, however, for other classes like linear, boost, buck-boost, etc. Your final assertation, however - that, for a given cost, the overall converter efficiency is usually highest if your input voltage is relatively close to the output voltage - is spot-on. Too far away from that, and the ol' V=I*R rule starts to bite you...

Many network server racks for already run on 48V

[Ungrounded Lightning]

if you have to re-convert anyway, 5V as intermediate voltage is not optimal. When converting to 5V, the voltage drop in the power diodes and in the wires to the mainboard eats a much higher proportion of the power than with 12V as intermediate voltage. 24V or even 48V would be even better.

Telephony has been running on redundant -48V DC supplies to the racks (typically from rooms full of floating storage batteries) since the early relay days. Much modern networking equipment also conforms to this standard,

5V is too low

[EmbeddedJanitor]

If you're cabling up a rack then you should not be delivering low voltages directly to the boards for two main reasons:

1) Losses are I^2R. This means that you have more power loss if you transfer power at low voltages through the same wires, connectors etc. You need switchmode power supplies anyway, so may as well switch down from a from a higher voltage.5V means more current than 12V, meaning thicker wires, higher current connectors etc and less headroom in the system for voltage loss.

Re:What in a modern computer actually uses 12V?

[zootjeff]

If you look at just routing 12 volts everywhere, you just would have to put the regulators in the hard drives, and CDROMS so they don't need 3.3 and 5 volts. Then what do you do about +5 Stanby that allows you to hibernate? Do you still need a stand by voltage? It isn't and easy answer and will take the whole industry to adopt it. Checkout formfactors.org for ATX and BTX specifications that Intel is pushing. What's also interesting is the 600 and 700, etc Watt power supplies just keep their 3.3 and 5 volts at around 30 amps max, but keep adding +12V1 +12V2 +12V3, etc.. Looks like the industry is already going to mostly 12 volts for distribution anyway. But don't you still need PS_ON, PowerOK, etc.. You're just trying to phase out the +5 and +3.3, and -12 which hardly any motherboards use these days, and maybe the +5 Standby, then it's going to happen eventually anyway. Most of the power is going on the 12 volt lines anyway, so having inefficient +3.3 and +5 isn't really a big deal. I've studied this for a while as my big hobby is computers in cars, I built a power supply called DSX12V that takes a 8-16 volt input and makes a solid 12v output that I got over 97% efficiency on. This is good for people sticking computers in cars or running them off banks of batteries for solar power applications etc.

Re:What in a modern computer actually uses 12V?

[seanadams.com]

In the old days, disk drive motors and fans. But many of these now run on 5V, hence the cheap USB-powered drive cases out there. Chips at CMOS power levels run at 3.3v, TTL is 5v, but hardly anything runs at 12v anymore. It seems to me that if they'd just pick their hardware carfully, they could run their entire server rack off of 5v+- rails.

You are correct that hard drives generally use just 5V, but the rest of your points are not even close. Modern CPUs require lower voltages, higher current, and tighter regulation, which is why DC-DC power supplies are now on motherboards instead of running directly from an ATX supply.

Furthermore, running a rack of servers on 5V rails would be absolutely absurd. Do you have any idea what the amperage would be? The bus bars would have to be several inches thick, the transmission loss would be enormous, and if you accidentally shorted them.... forget it!

Something like 48VDC might work but then you lose out on all the economies of scale driven by the 110/240VAC standard.

Just match the power supply to the motherboard and be done with it. Standardizing on one voltage is impractical, and besides, how would it improve "efficiency"?

Re:

[Alchemar]

I use to work on a lot of embeded controls. Ones with lots of different plug in boards to do different things, including all kinds of control signal inputs and outputs of various voltages. The best design I saw was a 50Khz 48VAC power supply. At those frequencies, even a good wattage xfmr is small enough to be soldered to the board. Every where they needed power they installed a xfmr, bridge, and votage regulator. Had to be a little careful about seperating the power from the signals, but all the power

MOSFETs use 12V

[wtarreau]

Many recent motherboard use 12V to control voltage regulators' MOSFETs gates because the higher the voltage, the lower the internal resistance, so the higher the efficiency. 5V is generally too low to achieve good efficiency, but 12V is fine.
From 12V, the MB can produce 3.3V and 1.xxx Volt for the CPU. It's easy to also provide 5V on the MB.

Re:What in a modern computer actually uses 12V?

[ottffssent]

Disk drive motors use 12V. Laptop drives (2.5" drives) use 5V exclusively, but standard desktop and server drives use 5V and 12V. SATA drives get 3.3V, 5V, and 12V. The VRMs that power your CPU and video card probably take their power off the 12V rail, as do many other components.

The reason you wouldn't want to power a machine off 5V is because you would need huge busses. Suppose you've got 40 svelte 1U servers in a rack, each drawing 100W. That's 4kW. Assuming that's a purely resistive load (hint: it's not), you'd need 800A at 5V for the whole rack. Are you familiar with the big connectors on car batteries? They're designed to pass less than half the 800A you'd need to run a rack off 5V, and your car battery only has to handle that for a few seconds while the engine is starting up; a rack would need to deal with that continuously. And that's for a pretty low-power rack.

Using 12V instead of 5V lets you get away with busses about 40% the size. Also, and probably more importantly, 12V DC is (IIRC - correct me if you're a PSU designer) easier to get efficiently than 5V DC. Once you split the 12V off into a few dozen servers, you can drop it down with small, fairly efficient CMOS regulators.

Re:

[camperslo]

The only things that natively use 12 Volts at a current high enough to be significant are the drives.
I think that misses the point however. Designing a power supply for higher output voltage, and switching regulators for higher input voltage, raises efficiency. (24 or 48 Voltage would likely be better yet, except for the need to come up with 12 Volts too)

It is unfortunate that the article (and the others that I could find) don't link to the white paper for some specifics.
Instead I'll have to base my comme

Re:

[Marxist Hacker 42]

RS232

Ok, so I missed one- but I haven't seen a server with RS232 ports in ages, and USB has pretty much taken over serial communications at this point if you once again, pick your hardware carefully.

Re:

[Craig Davison]

You've never seen a console port on a disk array, router, switch or UPS? That RJ-45 socket speaks RS-232 and will connect to your serial port with the right cable.

(Yes, some UPS's have USB)

Re:

[Bing Tsher E]

Mabye you're correct, but the dual rail +12/-12 needed for 'RS-232' is easily generated right in the interface chip with a few capacitors. Said chip only needs +5v to operate.

Of course, 'violations' of the voltage on 'Rs-232' ports has historically been really really common. Old PCs often had problems operating with serial mice, because the voltage span on the RS-232 ports on some machines was only a few volts. I remember an old Northgate 386 at work that had that problem.

Re:

[Marxist Hacker 42]

You've never seen a console port on a disk array, router, switch or UPS? That RJ-45 socket speaks RS-232 and will connect to your serial port with the right cable.

I have- but what part of choose your hardware carefully do you people not understand? RS232 is a rather outdated protocol at this point. My two latest computer purchases do not speak RS232 natively- but they DO have multiple USB ports.

Re:

[Da Web Guru]

I have- but what part of choose your hardware carefully do you people not understand? RS232 is a rather outdated protocol at this point. My two latest computer purchases do not speak RS232 natively- but they DO have multiple USB ports.

A lot of network hardware manufacturers choose to support RS-232 natively because of the relative simplicity of the protocol when compared to TCP/IP. Often an alternative non-serial product does not exist, so "choose your hardware carefully" is not always an option. Because of

Re:

[dragonman97]

I have- but what part of choose your hardware carefully do you people not understand? RS232 is a rather outdated protocol at this point. My two latest computer purchases do not speak RS232 natively- but they DO have multiple USB ports.

I believe it's the part where you expect me not to buy Cisco hardware because it uses that 'rather outdated protocol.' Any router that has USB on it is probably a toy! I'd just assume not have to connect via a USB -> DB9 dongle, but at some point, it's going to be harder

Re:

[dragonman97]

Be careful - some network appliances actually use connectors that you'd never expect to do curious things. None of my PIXes (PIXen? PIXies?) have USB connectors...they have RJ-45 connectors, for which an adapter/cable exists that pins out to a DB9 connector. (I have made my own converter to allow for various configurations of this.) Dell switches have an SPF module that uses USB cables for stacking - USB A-A male cables to be precise. Those are most definitely /not/ for use with a computer. But really

Re:

[chris_eineke]

WTF? Every ATCA blade server or CGL-compatible blade comes with RS232. Have you never heard of a serial port server? Those boxes are damn convenient. Oh, and... here's a 2006 ATCA blade server I worked with (and still do, sigh) that has a RS232 port: Sun Microsystems CP3020 [sun.com].

Re:

[Marxist Hacker 42]

And exactly what do you use such a server for? Not for internet, that's for damned sure, 115kbaud is far too slow to serve a 1.5TB line....

Consoles, of course.

[Junta]

Serial remains one of the most manageable approaches to console management. Video is, obviously, not loggable, not automatically monitorable, not greppable, and not amenable to low throughput, high latency remote access.. Serial devices and consequently drivers for them are so simple and straightforward, and the behavior so deterministic, that it is far preferable to something more complex (ethernet and usb) for a console. Ethernet certainly in questionable circumstances may suggest a driver unload/reload as a step to problem resolution, which is safer if not using as a console (though many times I have used ssh and chained the commands using semicolons). For example in that case, if your path contains an nfs mount, and you forget about it as you yank the network out, your chained command will hang as the shell tries to stat the nfs mount for the path. Part of the problem with relying solely upon the ethernet for console is the ethernet has more than one job to do, so it takes a fair amount more competent engineering to get to work right. Many newer systems offer to redirect textual serial traffic over IPMI, and that is admittedly decent *if* the vendor architects it robustly, which is difficult to ensure beyond hands-on experience with a brand and trusting in their consistancy. For example, e326 servers from IBM I wouldn't trust the net console, but an IBM x3455 I would be more confident in. USB, again, has similar complexity issues (it's multiplexed for keyboard/mouse/mass/storage/printing/scanning/etc etc). If you theoretically had bi-directional text console over some usb device, it's more difficult for a low level, simple piece of software to set up the usb controller and all requisite activities, then traverse the bus, identify the console devices, and then use it. Just like with an ethernet device where you may have cause to unload and reload a driver, a usb controller out to lunch with respect to a mass storage device would cause a similar issue. Enterprise distribution kernels tend to compile in the serial console and leave the usb controller modular, specifically with serial consoles in mind.

Serial console servers, in answer to your question, provide a scalable way for systems to access via the network serial consoles. By being dedicated, moderately simple systems with 40+ serial cables, they can provide access (via telnet generally) to a rack's worth of 1U servers, automatically log the content, or at the very least provide an administrator with remote console access at will to any given system.

Serial console is not obsolete in the least bit, just because it can't run your '31337' aero interface, or whatever nice and shiny interface that makes poser administrators and PHBs drool, doesn't mean good, serious systems administrators don't consider the technology to be a vital part of a robust management strategy.

Re:

[KermitJunior]

Correct me if I'm wrong, but VNC on a server? You use servers with windowing software on them? You must be new here. Slashdotters (*real* ones) only us CLI on their servers.

Re:

[ErikTheRed]

What happens if/when the system gets borked and you need to have access to a serial console to effect repairs?

Real servers have hardware remote console (such as this [hp.com]). You can remote control the box in just about any state other than "on fire".

Re:What in a modern computer actually uses 12V?

[seanadams.com]

RS-232

Sorry no. Modern rs232 circuits, if it's not already built into the UART, use a chip like max232 that runs off 5V and has a built-in charge pump to generate (close to) RS232 output voltages.

good idea but...

[grapeape]

Its a nice idea and one that is probably a long time coming, but phasing something like that into place will take an incredibly long time. Look at the struggles of PCI express, its still not in 50% of the newer motherboards and systems though its benefits are more than apparent. Its just been in the past couple years that we have seen a shift to full usb and most machines still come with ps2, serial and parallel ports anyway. Dramatic changes to the PC standards are very difficult, there are millions of existing machines that still need support. Perhaps if it was tied to a new socket standard in the future it could slowly be phased in through upgrades, but I see the chances as very very slim.

Re:

[griffjon]

To be fair though, I don't think anyone has a trademark or patent on "12V", unlike getting USB/USB2 certification, PCI Express, etc. etc. etc.

Google vs servers

[perlchild]

I'm surprised Google isn't running off 48V DC power supplies already, which, from what the documentation has shown me, already exhibit some of these savings...
On the other hand, it might have to do with Google's policy to use as much off-the-shelf equipment as possible, which 48V is not(iirc), so unless the "off-the-shelf-standard" changes, google might be in a position that they have to either break their own rules, or pay for following them.

Considering the number of servers google has, I'm surprised they

Re:

[synx]

48 V equipment is COTS - just not for the PC industry. The telco industry uses 48 V all over the place. I remember cisco's 2000 line of routers having a 48 V DC feed/plug.

Of course, how do you know Google isn't doing exactly what you mention? No one knows what really happens inside those giant datacentres. Protect the competitive secrets I guess.

If you've ever read the Xoogler's blog then you'd know that google doesn't exactly build machines the way you and I might. Meaning, your assumption of a case i

DC power supplies

[benow]

At the recent linuxworld in sf, I noticed some dc power supplies being pushed in the rack pc sector. I guess the lack of conversion from AC to DC saves a bit of juice, which makes a difference in large colocation centers. Combined with dc conversion on the motherboard, it would just be a matter of hooking up 12V DC direct to the board, which would be much nicer on the equipment and save a bit of power.

Proposal spells doom for USB powered devices

[JavaManJim]

You can say goodbye to USB powered devices. An example would be the canned drink cooler.

Thanks,
Jim

Re:

[dgatwood]

Why? I can turn 12VDC into 5VDC (what USB uses) with nothing more than a voltage regulator (or if you want to waste a ton of power, a relatively trivial voltage divider).

white paper

[TheSHAD0W]

Any links to the Google white paper detailing their reasons for this system architecture?

here it is

[TheSHAD0W]

n/m found it.

http://services.google.com/blog_resources/PSU_whit e_paper.pdf [google.com]

I've wanted this for years.

[dgatwood]

The ability to have all my machines powered by a heavy cable carrying 12VDC would be pretty useful for several reasons.

• The UPS could be integrated into the power supply, avoiding lots of energy lost in converting it up to 110VAC and right back down again.
• The power supply would then be external, where it could be a fanless brick instead of being inside the case where it adds heat that must be dissipated.
• A switching power supply is theoretically more efficient than a wall wart. If everything were 12V, all those stupid little outboard devices could draw power off of the same supply source, resulting in better overall efficiency. More importantly, I would never let out the magic smoke when I accidentally plug a wall wart into the wrong device. :-)
• A 12V system can more easily be integrated with solar panels to reduce load on the power grid.

*sigh*

Re:

[sebol]

it make sense...

Next generation computer should have 12v plug and special cable, so that it can take 12v source from outside.
What's important is the cable and socket but be different with 110v or 235v to avoid "accident".

i would love to see conputer running from a car battery

Re:

[mrmeval]

You still miss the Commodore 64 don't you?

Re:

[fm6]

No, he misses the Convergent Technologies NGen. This was a pretty powerful x86 platform that also used external power supplies. The nicest thing about it was that it was quiet: the power supplies (yes, plural; the number you need varied according to your internal hardware) used passive cooling, so only internal heat sources needed to be cooled.

This was 1983, which was when IBM introduced the PC-AT [wikipedia.org], the machine which defines "compatibility" to this very day. And the AT used a big, noisy internal power supp

Re:

[fm6]

You make a good point about wall warts, except you don't go far enough. If all portable devices accepted 12V power, somebody would come out with a single brick with multiple 12V plugs, which would be a godsend to travellers who currently schlep one wall wart for each device.

**big sigh**

Re:

[dgatwood]

Sort of like an iGo?

Unfortunately, they still don't have the one thing I want, which is to combine the iGo with a lithium ion charger for camera batteries with swappable tips. Camera chargers take up more space than other power supplies by far.

The day I have my first battery failure on my Canon Digital Rebel, I'm cutting that sucker open and fitting it with a 9V battery clip and a switching regulator.

Re:

[object88]

If all portable devices accepted 12V power, somebody would come out with a single brick with multiple 12V plugs, which would be a godsend to travellers who currently schlep one wall wart for each device.

Guitar players have just such a device for stompboxes. The vast majority of stompboxes run off a 9V DC, from a battery or wallwart, so there are several bricks you can purchase with just that. Of course, you still have to be careful around those oddballs which call for 5V, 12V, 18V, or inverted power, so i

Re:

[genericacct]

I'm all about the solar angle! Someday I'll wire my house with an off-grid 12-volt solar system, with 12-volt "car lighter" sockets and DC lighting (both LED and mini halogen). Laptop and WiFi router plug in to it.

And everything can plug into the car with the same cord. That's another awesome advantage, being able to put these same computers in cars and RVs.

Re:

[Wesley Felter]

Your points are interesting, but that's not what Google is talking about. They're just proposing that the cable from the power supply to the motherboard inside a PC or server should only carry 12V. The power supply is still internal, and each device still has a separate supply.

I don't get it.

[pavon]

I agree that a standardized 12VDC connector on all electronic devices would be nice, like every other poster here has pointed out, but I don't think that is what google is talking about. You can already get power supplies that take 12VDC in, or even dual 48VDC (telecom standard), and I would be surprised if google isn't using something like that already.

What they are recommending is that the power supply only have 12V out, and all other DC-DC conversions take place on the mother board. Unfortunately, the ar

Offtopic: Growth

[fm6]

The rate at which the Google computing system has grown is as remarkable as its size. In March 2001, when the company was serving about 70 million Web pages daily, it had 8,000 computers, according to a Microsoft researcher granted anonymity to talk about a detailed tour he was given at one of Google's Silicon Valley computing centers. By 2003 the number had grown to 100,000.

I've done my share of Google-bashing (mainly due to their inability to move their newer products beyond "beta"), but here's an accomp

While we're at it let's ditch motherboards too.

[Anonymous Coward]

The motherboard itself is an outdated concept. It's no longer really necessary if you've dealt with small form factor boards you can easily see that the boards are just a substrate to stick the chips on and for that a flat board-like surface doesn't make sense. What you really need is a cubic cartridge like device that gives you access to more surface area for interfaces close to the memory and CPUs and other chips in a smaller area. It would also facilitate cooling reducing power requirements at the system

When you thought it was safe to plug in...

[__aaclcg7560]

Microsoft will specify a Vista-compatible power supply that not only uses a single voltage but can radiate power like Star Trek's newest Enterprise does for nearby devices.

Why use individual power supplies?

[sharkb8]

Seems like there was some talk a couple of years ago about doing the AC-DC conversion on a massive level, then running individual servers off a server-room wide source. If you create +-12v output on a block of plugs, and +-5v outputs on another set of plugs, you could achieve much better efficiencies. You'd also probably cut your costs significantly. If you put a massive AC-DC transfomer in another area, you could isolate the cooling systems, etc. One large cooling system for a single, large power supply wo

Why not -48?

[AaronW]

A lot of telco equipment is designed to run on -48 volts DC and PC and server power supplies are readily available at this voltage.

The advantage of -48 over 12 volts is that there will be less loss through resistance and smaller conductors can be used. Of course, there is a greater risk of electric shock, but I would think -48 would be pretty safe.

48 volts is also the standard for Power over Ethernet (IEEE 802.3af) [wikipedia.org]. This may not be compatible, though, since telcos run -48, not +48, though some equipment can operate with either (though some cannot).

Re:Why not -48?

[springbox]

48 volts is also the standard for Power over Ethernet (IEEE 802.3af) [wikipedia.org]. This may not be compatible, though, since telcos run -48, not +48

Sure they can! Just reverse the polarity and reroute all power to the main deflector dish!

This ONLY makes sense in a rack, NOT a desktop!

[Bruce Perens]

Low-voltage power supplies in racks might make sense. Not in desktops, because low-voltage power takes requires more copper to distribute it, because there's more current. Copper is very expensive of late.

Bruce

What lower voltage power supply?

[argent]

They're not talking about reducing the voltage the PS uses, they're talking about not having the PS produce things like +5 and -5 as well as +12, INSIDE the computer.

Re:

[SEE]

For efficiency, CPUs are heading below 3.3v, and RAM too.

That's actually why single-voltage PSUs make sense. Your CPU, GPU, and RAM don't care if the PSU is providing 12V, -12V, 5V, or 3.3V, or any combination of them, as long as its VRM steps it down to the 1.7V or whatever it needs. So why have the power supply provide so many different types of power, instead of just one of them? It's all going to be converted by a local VRM anyway.

And a single-voltage power supply is about 85% energy-efficient at con

Bad idea

[ErMaC]

Google's whitepaper is interesting but the fact is that DC in the Datacenter is already happening, and it's not gaining much momentum for multiple reasons.
Google's perspective is rather unique, they use super-cheap desktop systems that individually do not use a lot of power and thus running them off 12v DC might make sense. But in any other, more conventional datacenter, servers have multiple power supplies that can EACH pull 800w of power. Now when you're running 110v AC that means you're pulling ~7 amps through a single cable. You need datacenter grade power cables for this, but it's still sane. Now you can get datacenter equipment that runs 48v DC, but those cables end up running ~15 amps through them, so now you need substantially stronger cable - cable so thick that running it becomes a seriously difficult task due to the guage of the wire!
More likely the direction people are going (and have been for some time) is to 208v AC or 3 phase 220v AC. Now you've just halved the current draw, meaning that your PDUs don't need to be as hefty, your wire doesn't have to be as thick, your coils don't get as hot, etc.
Running 12v DC in any real data center would be ludicrous - the amount of current you'd have to draw through your cables would be way beyond a safe level.
Also AC/DC conversions are cheap these days. And remember, DC can kill you just as easily as AC when your DC Voltage is that low.

RTFA

[Wesley Felter]

Google is proposing 12V between the server's internal power supply and the motherboard. Everything outside the server would still be 208V.

This is about voltage to the boards, not the box

[Animats]

Most of the postings so far have it all wrong. Google is not proposing 12VDC into a desktop PC or 12VDC distribution within the data center. What they're proposing is that the only DC voltage distributed around a computer case should be 12VDC. Any other voltages needed would be converted on the board that needed it.

This is called "point of load conversion", and involves small switching regulators near each load. Here's a tutorial on point of load power conversion. [elecdesign.com]

It's been a long time since CPUs ran directly from the +5 supply. There's already point of load conversion on the motherboard near the CPU. Google just wants to make that work off the +12 supply, and get rid of the current +5/-5/+12/-12 output set.

Mod parent +100 informative...

[argent]

Though getting rid of all the voltage levels will take more than the motherboard work... you'll also need to do something about all the disks and other components that are currently getting a mixed feed.

Re:This is about voltage to the boards, not the bo

[Bruce Perens]

Oh. So, we have lots of switching power supplies and tantalum capacitors (because we have to supply lots of current at low-voltage) on the MB. Thus moving work from a cheap part of the computer to an expensive part. Not sure I want more power-supply electronics on the MB than is already there.

Re:

[inKubus]

There are over 6500 Walmart stores with an average size of 120000 square feet. Every 500 sqaure feet they have a 4-tube fluorescent light fixture, drawing 4x40 or 160 watts. Multiplying out, the total square footage is ~6500*120000=780,000,000 square feet. Divide by 500 square feet to get the total number of fixtures, 1,560,000, and multiply that by 160 watts to get the total watts, 249,600,000. Probably 75% of those Walmart stores are 24 hour, while the rest are 12 hour: (.75*24)+(.25*12) = 21 average

Hey here's one

[eclectro]

S
ave the conversion process all together and tell motherboard makers to rectify 120 V directly on the motherboards. It worked for old TVs, it could work again. Maybe.

PicoPSU

[Michael Woodhams]

There's already a small-scale example of this: the PicoPSU [silentpcreview.com]. You use an external 12V power brick, and then internally replace your entire computer PSU with something about the size of a matchbox. However, it is only 120W, and a bit short on connectors.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Another way to get more efficiency

[Skapare]

Another way to get more efficiency is to operate the Switched-mode power supply [wikipedia.org] at the higher voltage it supports, usually 220 to 250 volts. In most of the world this is already done. In North America computers are typically run on 120 volts (in Japan this is 100 volts). In general, these power supplies are more efficient by about 3% or so, on the higher voltage. Of course, be sure to flip the voltage switch if it has one, or otherwise verify that it does support the higher voltage.

For a single computer, it would not be worth adding the extra circuit to get 240 volts. But if you run several, it could be worth doing so, especially if you have so many that it exceeds the capacity of one 120 volt 15 or 20 amp circuit (you could have twice as many on the same amperage if operating at 240 volts). If you already have a circuit dedicated to the computers, that circuit could be converted from 120 volts to 240 volts by changing out the circuit breaker from a one pole to a two pole type, marking the white neutral wire with red or black tape to comply with electrical code identification requirements, attaching these wires to that new breaker (not to the neutral bus), and installing a 240 volt style outlet (NEMA 6-15R or 6-20R). These are the steps that would be used to install an outlet for a big window air conditioner (which you might need anyway with so many computers). Then you can use this [interpower.com] power cord.

No...

[rsilvergun]

google hires experts on Electrical Engineering to figure out how to reduce the power bill on those 450,000 servers. Hell, I'm all for it. Less power means less heat means quieter fans (w/o spending an arm/leg on an Antec Sonata or whathaveyou).

Re:No...

[TheSHAD0W]

Considering how many machines Google has to maintain, I'm surprised they just don't order motherboards and power supplies to their own spec, and then allow the mfrs to distribute the design to others who request it. They're big enough and have enough whuffie that they can start a trend all by their lonesomes.

Re:No...

[TheSHAD0W]

After reading Google's blog entry [blogspot.com] on the subject, I'm left puzzled by their call for a new standard with no further details, especially since it seems they're already using the technology. A power supply is simple enough, but I'd like to see what sort of strategy they're using for voltage conversion on their motherboards. What connectors are they using for power?

The funny thing is, this idea is relatively old, though AC was used instead of DC. Remember the Imsai 8080? The S-100 bus used an 18V AC supply, and each card had its own DC conversion and voltage regulator(s).

Re:No...

[sillivalley]

The early S-100 systems (Altair, Imsai, Poly, Northstar) used linear supplies and ran unregulated DC on the S-100 bus. Most designers aimed for +8 to +9 volts to feed the onboard +5 volt regulators (and the3 volt or so headroom for 7805 regulators). Can't think of any that ran high current AC on the bus. Some systems, such as the Poly, ran a squared-up 60 HZ signal for real-time clocks.

The heat losses in S-100 on-card linear regulators were immense! That and the weight of the (linear) transformers helped make the Apple ][, with its switching power supply, so popular (I still have an old Poly power transformer; makes a great doorstop).

Some mainframe computers used the scheme mentioned by others -- polyphase high-frequency AC distribution. High frequency (think 800 Hz) power transformers are small and efficient; that's why switching supplies run at high frequencies (in the hundreds of KHz range).

Efficiency is not only about wasting less power, it's about generating less heat!

Re:No...

[kasperd]

Efficiency is not only about wasting less power, it's about generating less heat!Which is of course exactly the same. In the end all the energy you put into a computer turns into heat. The energy wasted in the power supply turns into heat in the power supply, and all the heating of the power supply is energy wasted rather than used to supply the computer.

450,000 servers....

[Dr_Barnowl]

$450,000 dollars? What's that, you say that could pay to employ Fulcrum of Evil at Google for five years?

Ach, shoot, we spent it on USB ports. Never mind.

Re:No...

[alienw]

If you can cut power consumption by 10 watts a machine (quite realistic) and you have 100,000 machines (Google has more) you just saved 1 megawatt of power, or about a million dollars per year in electricity (without even taking into account the electricity required for cooling). That's quite a chunk of cash.

Re:

[MoxFulder]

Yeah, Google hires experts on anything pretty much, I'm told.

Apparently they hired expert ergonomic and industrial designers to figure out how many servers and workstations they could cram into a mobile semi-trailer lab, while still making it comfortable to work in. Kind of a neat optimization problem I think.

Re:

[stunt_penguin]

"how many servers and workstations they could cram into a mobile semi-trailer lab"

I'm guessing the answer was lots and lots...... there are quite a few technical challenges as you say, power, cooling, and making sure that the machines survive the journey, too.

It would be a neat side business if Google went into providing server farms and data centers for other businesses; as other people have mentioned they have a lot of smart people working on the associated problems.

Hey, it could save their asses

Re:No...

[JahToasted]

Why do that when they can just rent out space in one of their super massive server farms. Think about it... you get some good bandwidth, your data will be mirrored on geographically and topographically separate systems. You don't have to worry about hardware failure or anything like that and you'll be able to get all the bandwidth you could ever want. You don't have to worry about database replication or syncing up data or anything like that, its all taken care of for you. Depending on your needs, you can have gmail, google maps, google office, adsense all integrated with whatever it is you're setting up... web app, file server, database system, whatever it is you're setting up you'll be able to get it from Google along with some nice cross platform tools to make it as easy as possible.

And because of economies of scale the price will be very reasonable, ie. cheaper than rolling your own solution. Hell, I'd consider it, wouldn't you?

Re:No...

[stunt_penguin]

Actually, you're spot on- why sell the physical machines (a fixed profit per machine sold) when you can sell a service (a steady income) .... that said a lot of companies would rather have 'their' people with their hands on their machines, but I'm sure a lot of organisations would benefit. They could even call the service Google Farm ;)

Come to think of it, given the recent story on slack govornmental data security, maybe they could have Google serve their data for them. They already know how hard it is to get data out of Google, right? :p

Re:No...

[x2A]

Better than that guy they spent $50,000 who said moving the plant from the window and installing a water feature would allow the energy would flow much better...

If google come out with a "can save energy this way...", and gets the world to follow, the marketing value speaks for itself. That kind of reputation doesn't come easily.

Cargo container server room

[yppiz]

The Internet Archive addressed a similar problem: can you build an Internet Archive in a storage container and ship it? They came up with a design for this around standard racks of low-power, low-heat, high-storage nodes.

Their answer is the Petabox [archive.org]. It's a server setup designed to be "shipping-contained friendly", meaning they can build out a container stuffed with these racks, and have it operational on site with connections for power, cooling, and bandwidth. With this design, they can deploy a mirror of t

Re:

[MarcoAtWork]

given that the article says

Although Google does not plan to enter the personal computer market, the company is a large purchaser of microprocessors and has evolved a highly energy-efficient power supply system for its data centers.

I assume Google is employing some smart electrical engineers, which are more than qualified to make this kind of recommendations I would think...

Re:

[Ucklak]

I really don't think that Google is a bunch of shoeless code jockeys playing with database results and fiddling with power connectors to fry capacitors.
I'm pretty sure that there are intelligent, educated people that get a Google payroll paycheck.
I'm also willing to bet that if some of the intelligent, educated people that are experts in this area that aren't on Google's payroll did receive a Google paycheck for their opinion and dissertation on the matter.

Once could conclude that with a server farm of 450,

Re:

[ClosedSource]

I wouldn't look to MS or even Apple for expert opinions on power distribution.

Re:Big ego department

[purpledinoz]

I would bet a lot of the employees at Google have Electrical Engineering degrees. Don't underestimate the brain power Google has in its employee base. But the power supply issue they're trying to address isn't a technical challenge, but a political challenge.

Re:

[chroot_james]

There is no reason to be annoyed by people trying to do good things!

Re:Big ego department

[ve3id]

I have been saying this for years. We lose 10-20 % of energy charging a battery in a UPS with 117V, we lose another 20-30% in the inverter to get it back to 117V, and then we lose another 10% getting the 117V back to usable voltages for the PC.

It does not take an expert in electrical engineering, just common sense.

Can I sue google for stealing my idea?

Re:Big ego department

[Skapare]

There are different kinds of UPSes that do this in different ways. The two major types used for PCs are called "line interactive offline" and "dual conversion online". The first just passes the AC power straight through to the output. If the AC power coming in goes out of range, then it flips a switch internally (relay, contactor, thyristor, etc) to supply the power from an inverter driven by the battery. The second converts the AC coming in to DC all the time, and converts that DC back to AC for output. It then does the switching in DC, or parallels the DC with the battery directly. These variations are classified as "topology" by many manufacturers.

Both of these kinds can have inverters that produce square waves, pseudo-sine waves, or very nice clean sine waves. The dual conversion type can also isolate a poor power factor (the deviation of the current wavefrom from being a sine wave in sync with the voltage) of the PC power supply from the power source. A poor power factor means the product of the average current times the average voltage (apparent power) exceeds the actual real power (average of all the products of the voltage and current and each point in time) being used, which results in reduced efficiency and other problems.

• Uninterruptible power supply [wikipedia.org]
• Power factor [wikipedia.org]
• Switched-mode power supply [wikipedia.org]
• AC power [wikipedia.org]

Re:

[fm6]

Google manufacturered those 450,000 servers themselves. That gives them some expertise. If they had given that many people medical care, I'd listen to their medical advice too.

Re:

[jandrese]

I wouldn't count on it. When you're buying in quantities as large as Google is there is no small chance that they designed thier own motherboard and case specifically for their purposes. There's no point getting the floppy controller, USB, and any of the other stuff that you normally get (even on server machines) that's totally useless to Google. They probably paid a lot of attention to the power consumption, not only to install the smallest reliable power supply possible, but also to figure out how many

Re:

[zootjeff]

For your 12 volt battery that can vary more than 10% as it discharges, you'll need something like what I have designed for use in cars. Some motherboard makers already make computers that run on 12 volts. The Commell LV-673NS Pentium M Mini-ITX Mainboard already runs on 12 volts (+- 5%), and then if you use the Mpegbox DSX12V which is 95+ efficient, then it can run off a battery or in a car.

Re:

[chill]

Apples and Oranges.

Most anyone running a data center or who has a lot of servers cares about power and heat. The electricity bill -- including all that extra air conditioning -- is a whopper in some places.

Those people, however, don't give a rat's ass about the latest video card. People with rooms full of server are probably using an IP KVM or running them totally headless.