First Actual CPU Energy Use Statistics Published

BBCWatcher writes "CNN is reporting that the U.S. Environmental Protection Agency in August asked server manufacturers to develop 'miles per gallon' ratings for their equipment that would provide accurate assessments of energy efficiency. IBM says it is now providing 'typical usage ratings' for its line of z9 mainframe computers, in addition to previously available maximum power ratings. More than 1,000 z9s around the world started reporting (with the owners' permission) on May 11th their actual installed power and cooling demands, so IBM can publish statistics such as how much energy is required to turn on an additional processor to run multiple Linux virtual servers. The answer? About 20 total watts. 'Over time every vendor is going to be asked to provide typical energy use numbers for their equipment. It's what the EPA wants, and this allows us to move beyond simple performance benchmarking to energy benchmarking.'"

Comments

damn lies

[Bob-taro]

Oh, but there are so many ways to play with statistics. Hey, is the the first post?

Re:damn lies

[Anonymous Coward]

Hey, is the the first post?

Statistically speaking, no.

Re:damn lies

[Bob-taro]

Ha! I confess, I cut that post short to try to get the first post (I'd never gotten one before!)

Anyway, the problem with trying to get some "miles per gallon" efficiency rating on computers is defining the "miles". For example, if computer A is 2 times faster and uses 1.5 times the energy compared to computer B at full load, and both computers are run at full load 8 hrs a day (doing some serious number crunching), which computer is more efficient? A is using more power, but is doing twice the amount of "work" of B. So do you measure straight Watts? Watts / MFLOPS? If you use MFLOPS, how do you account for differences in architecture?

Re:damn lies

[Colin Smith]

Oh, but there are so many ways to play with statistics All the more reason to understand how they work.

Re:damn lies

[blindd0t]

Good point. I think we can all attest that the 'mpg' analogy isn't so good in its present state, as it has been highly inaccurate in many (if not most) recent cases. Perhaps a simple energy benchmark similar to audio amplifiers would suffice: show me peak & continuous power. If it's on, what does it consume continuously (or as a minimum - i.e. just 'on' but idle), and what is the most it could possibly consume at any given time? I know this doesn't cover efficiency very well (i.e. watts/FLOP); however, efficiency is relative to the application. Let the consumer decide what benefits him/her the best between continuous and peak consumption.

Re:damn lies

[vtscott]

I agree. While this could be a useful measure, companies will find ways to game the system like they always do. The problem is, if they just publish hard numbers and exact specs they will be difficult to interpret. When they publish these "more useful" stats, they'll just do everything they can to get the biggest (or smallest) numbers while sacrificing everything else.

For example... I was recently shopping for home theater projectors and was doing a lot of comparisons between brands. The two most important things I was looking for were picture quality and brightness. Both have standard metrics such as contrast ratio and lumens. The problem is that the manufacturers know they're being judged on these two numbers, so they just play with the numbers until they get them as high as possible. Sure, you can get theoretically get a picture with the advertised contrast ratio, but the brightness will be terrible. And yes, you can get a picture as bright as advertised, but it will look like crap. The numbers are supposed to give you an indication of the quality of the projector, but instead it's just a dick measuring contest between manufacturers and most would likely sacrifice overall quality just to up their numbers.

I would expect the same will go for CPU manufacturers in this case. Processors will be advertised at XGHz and Ywatts*, but you definitely won't see both X and Y true at the same time.

*Y watts based on our definition of "typical use" while running at a much slower clock speed than X

Re:damn lies

[Runesabre]

I see what you're saying and definitely would be suspectful of a rating that included speed and energy consumption all rolled into a single number.

I see this as another important rating to be used simply to evaluate a chip based purely on energy consumption. It should also be a good motivator for chip makers who haven't concerned themselves with energy consumption when their chip suddenly is viewed unfavorable when compared to more energy efficient yet equal performing chips from rivals.

Most IT purchasers are going to understand speed and energy are separate, often opposing numbers just like most understand a semi will have a lower gas mileage than a car. Even with far better gas mileage you still wouldn't buy a car for freight hauling nor would you buy a semi for picking kids up from school, taking kids to soccer, grocery shopping and driving to Wal-mart to stand in line at midnight for the latest great Halo game for your kids. :)

Re:damn lies

[Anonymous Coward]

A is using more power, but is doing twice the amount of "work" of B. So do you measure straight Watts? Watts / MFLOPS? If you use MFLOPS, how do you account for differences in architecture?

Why not extend this concept to include the computer OS then? For example, an app running on the bloated Vista doesn't get the same resources to complete a task timely compared to running it on XP or on linux on the same CPU. It could get messy though, since it needs to account many more variables. It would also be interesting to have statistics of CPU energy used by malwares.

Re:damn lies

[Reziac]

I suppose you just run the numbers, same as for anything else.

Example which by unfortunate chance is handy: I have to replace my well pump. I can either go with a 10 horsepower pump that does 60 gallons per minute, or a 3 horsepower pump that does 18 gallons per minute and costs about $1000 less up front. Now, 10HP uses one helluva lot more electricity than 3HP... but it only needs to run a fraction as much time to pump the same water. Turns out when we ran the numbers, the 10HP was actually more cost-effective since its net use was about 5% less than the 3HP -- which makes up the difference in price in about two years, AND the 10HP is liable to have better longevity since it's made better and won't have to work as many hours; and if I really need 60GPM of water, I can have it, which the other pump cannot deliver at all. So that upfront extra $1000 works out over its 30 year lifespan to several thousand in direct savings (varying with the price of electricity), and some unknown amount of savings wrt other factors (longevity, volume).

This is actually quite similar to figuring needs for a data center -- if you have NN-big servers using XX-amps power and putting out YY-BTUs worth of heat, where is the balance between electrical use, cooling costs, and smaller units that won't do as much work? So you run the numbers, because they may not be what you'd expect.

BTW per figures I saw a couple years ago, about 40% of the electricity used in California goes to feed the starving data centers.

Re:damn lies

[edmudama]

In many industries, this eventually gets resolved as the manufacturers realize that everyone is gaming the system, so they define new metrics that are perhaps harder to game, to help distinguish the quality of their products.

One example would be the move from "response time" in LCD panels, to "grey to grey response time" which prevents advertising of just one of the hyperfast transitions. (Can't remember whether black-to-white or white-to-black is the faster one)

Re:damn lies

[Marxist Hacker 42]

Like it costs The Dalles Google Center about $10/year at current hydropower prices to turn on an extra processor?

Re:damn lies

[pz]

Anyway, the problem with trying to get some "miles per gallon" efficiency rating on computers is defining the "miles". For example, if computer A is 2 times faster and uses 1.5 times the energy compared to computer B at full load, and both computers are run at full load 8 hrs a day (doing some serious number crunching), which computer is more efficient? A is using more power, but is doing twice the amount of "work" of B. So do you measure straight Watts? Watts / MFLOPS? If you use MFLOPS, how do you account for differences in architecture?

The standard performance in business applications is database transactions per second. It should be pretty easy to publish transactions per Watt instead, for a given power use. Beyond that, for general purpose use, there are standard benchmarks already for an absolute number of operations per second on various tasks, like SPECint, SPECfloat, etc. There's no reason these too can't be normalized by power consumed, eg, SPECint/W.

Re:damn lies

[dgatwood]

While you're at it, account for a more typical load of 10%, account for periods overnight when the machine may be completely idle (for a US-centric web site, for example), etc.

First, everyone has to agree on a series of load metrics which when combined are sufficient to get a good approximation of performance under various types of load (disk-heavy load, CPU-heavy load, etc.). Only then can we really answer the question of how much power a server is going to waste.... :-)

Re:damn lies

[aichpvee]

Oh man, this isn't going to end well. If anyone starts weighing the value of the content being hosted 99% of the Internet will be rated as wasteful.

about time.

[cliffski]

about time. energy prices aren't going down any time soon, and if this means a spread to accurate energy consumption cost reporting for all computer equipment, that can only be good news.

Re:about time.

[Rei]

I agree. For those with home servers like me who leave their computers on all the time, I'd be willing to pay an extra $5 or so per watt saved ($2 for the price extra watt over the course of an assumed usage lifespan of 3 years, $1 for the reduced cooling cost and increase in part lifespan from it being cooler, and $2 extra for the environmental benefit). If I lived in a place with expensive power, like California, that would probably be $7 or so per watt saved.

Re:about time.

[Anonymous Coward]

Convince the industry that power supply units should contain a standardized watt meter. And then there should be a standardized way to relay that data to the motherboard. Then anybody could get an fairly accurate measurement of standby, peak, and average energy consumption right along with their fan speeds, CPU temp, etc. You could use this to find inefficiencies in a system (maybe as the fans inside get dirty, they take more juice to maintain speed) or compare performance to actual usage.

But yeah, it'd also be nice to know what the levels typically are on a new commercial box under a fairly standard configuration.

Power consumption since mid-80's?

[securityfolk]

I would be interested in seeing statistics from the 80's on, to show how much power an "average" home PC or business workstation consumes. Over the years I've heard things like "it takes more energy to power it on than to just let it run", "it uses as much energy as a blowdryer", and "it uses as much energy as a lightbulb". Also, doesn't it take LOTS more energy to continually refresh RAM than it does to enable a processor? Think we could break usage down by component in these stats?

Cheers, Securityfolk

Re:Power consumption since mid-80's?

[Trinn]

No it does not take more energy to refresh the ram. Here's how to tell that. The ram is built on the *same* process & technology (or perhaps even an older process, making it *less* efficient) yet it never gets hot enough to require a heatsink let alone active cooling. Given the same or less efficient process, the same portion of the energy pumped into the system would get turned into heat.

Re:Power consumption since mid-80's?

[redleaf8]

Just get a Kill-A-Watt. My desktops use about the same power as a lamp with a 75 watt bulb in it, my laptop is 20-30 watts.

Re:Power consumption since mid-80's?

[archen]

"it takes more energy to power it on than to just let it run"

I'm not sure where people get that from (I've heard it myself). Most of my tests on a run of the mill PC (onboard VGA) shows a computer drawing around 80W. There is a small spike when booting, but I honestly have a hard time believing you save money letting the thing run over 10 minutes. As someone else said, get a kill-o-watt and test it out. It's actually really interesting to see how much power stuff like TVs draw, as well as the difference between CF and incandescent lighes.

Of course my testing was extremely informal so take that with a grain of salt. For some reason using an SSD hard drive with FreeBSD on the same hardware (instead of Win2k) dropped the power usage to around 58W. I have no idea why that is.

Re:Power consumption since mid-80's?

[KingMotley]

Is that 80W when the computer is in it S3 sleep state, or just idle at the desktop?

Re:Power consumption since mid-80's?

[Anne Thwacks]

I'm not sure where people get that from

It was defintiley true in the days of thermionic valves - ie the 1950's. It probably has not been true since the transistor was in common use for logic - approximately 1968 if my memory is correct (highly improbable).

My profesiosnal testing shows that unless you havbe very expensive professional equipment, your readings are probably +/- 30% accuracy. Kill-a-watt is not expensive professional industrial test equipment.

If you are concerned about server power consumption, you should be looking at OpenBSD on Sparc64, and not Vista on Intel. Since power (including that required to get rid of waste heat) is the biggest cost in a data centre, you should be worrying about it if you run a datacentre.

If you DONT run a datacentre, stuff the CPU, look at your AC costs! Who cares if the processor is taking 50W or 58W when your AC is using 3kW!

Re:Power consumption since mid-80's?

[timeOday]

Also, doesn't it take LOTS more energy to continually refresh RAM than it does to enable a processor?

Your best window into this issue is laptops, where every watt counts. The simple answer to that particular question is "no." I have a D630 Dell laptop with 4GB RAM. When suspended to RAM, it consumes about 1% of a 56 watt-hour battery, per hour. In contrast, with the processor and screen running the whole battery is emptied in 3 hours.

Here is the sort of chart you're looking for, although it's somewhat dated. And of course it varies by model. I have a T60p Tkinkpad laptop which, by virtue of its Core Duo processor, presumably has good battery life. But since it also has an ATI FireGL video card, the battery life is crappy and it's uncomfortable to use on your lap.

Re:Power consumption since mid-80's?

[nschubach]

...but my memory does have a heat sink. Well, they are heat spreaders, but it's a similar concept. Move the hot to the cold.

Re:Power consumption since mid-80's?

[Anonymous Coward]

That's 80W idle. Actually using the machine doesn't seem to make much of a difference either - With 99% of the CPU time on the idle process I guess that makes sense. At powered down it takes about 4W. S3 is supposed to take around 10W. I just recently figured out how to get some machines to stay in S3 so I've never done testing on that.

Re:Power consumption since mid-80's?

[Not The Real Me]

...If you DONT run a datacentre, stuff the CPU, look at your AC costs! Who cares if the processor is taking 50W or 58W when your AC is using 3kW!...

Pentium I, II and 3's run in the sub 75W range. I've read acticles which claim that Pentium 4's and the multicores processors chew up anywhere from 125W to 300W simply by idling. I have to guess that someone running a P4 with Vista is sucking up 300 watts, and when they play Halo, the consumption probably hits 500 watts.

Re:Power consumption since mid-80's?

[archen]

Actually I wasn't testing the datacenter (which uses vastly more power even when idle). I was testing desktop PCs. Many people at the company I work for tend to leave their PCs on all the time. Generally that bothered me as being wasteful, so I started to set many PCs to turn themselves off an hour or so after everyone left. I started to get some resistance when people complained that they had to turn their PC back on in the morning (yeah, tough life I know).

So I started probing into how much power we were actually burning for no reason in order to back myself up. After a few complaints everyone just dropped the topic. I was also doing some testing in order to get ballpark figures on how long a computer should last on a ups backup.

As for the data center... I gave up. They just burn a lot of power. For every process I manage to streamline and consolidate some function, they go and drop yet ANOTHER windows server that "MUST have its own machine". I have however dropped some regular backup (redundant process) servers to something like the Soekris net5501 that draws something like around 10W.

Re:Power consumption since mid-80's?

[darthflo]

Someone with a Pentium 4 will probably not even /have/ a PSU capable of delivering 500 watts. Afaik the worst P4 consumed 120 Most current desktops do have power-saving features, so I'd guess an office box would consume about a hundred watts on average, while really bad gaming rigs with two or more GPUs, a PPU and a dual- or quadcore processor should come in at around 300 average and up to 600 peak.

Re:Power consumption since mid-80's?

[Anonymous Coward]

Every self-respecting geek should own an AC ammeter.The Kill-a-watt looks like a good one: built in power calculator and very safe to use (personally I use an inductive loop that I got for $10 at the night market but I'm cheap).

An easy way to very roughly assess power usage is to touch whatever you are curious about (cue electrician jokes). To a good approximation all power consumed by any electronic device is turned to heat. So ask yourself: how hot does it get, how much of the day is it hot and how much energy does it take to keep it hot? RAM does not get very hot (cool enough to touch without explicit cooling) so it obviously draws much less power than your CPU (~45c even with a huge heatsink and fan).

Re:Power consumption since mid-80's?

[billcopc]

Your guess is just about right.

I've recently optimized my home power consumption, after my last PC upgrade caused the breakers to blow repeatedly. Here's the breakdown of my various PCs:

Pentium Pro 200mhz (firewall/router) : 40w (no difference under load)
AMD Athlon X2 3800+, underclocked to 950mhz (file server) : 100w idle, 120w load (with 8 hard drives)
Dell 22" entry-level LCD : 45w in-use, 1w standby
Viewsonic 19" CRT : 40w in-use, 1w standby (surprising!)

And finally, my new baby. An Intel Quad-core Q6600 oc'ed to 4ghz, with water cooling, Geforce 8800GTS and a crapload of hard drives. I call her Scorch. She goes to S3 sleep after 15 minutes of inactivity, otherwise I'd be tacking an extra $40-50 onto my hydro bill every month.

Idle: 350w. 100% CPU load: 550w. Everything loaded: 750+w.

I have no idea why it needs 350w to sit idle, when the X2 it replaced sips on a third of that. Sure, the water rig eats up about 50w, but that leaves 300w just to power a board, cpu and video card. I've heard the Geforce 8800 eats about 80 to 100w doing nothing, which is really pathetic! I think the manufacturers should put more effort into conserving energy. I'm fine with it chewing over half a kilowatt under load, because it's getting the work done in a quarter of the time it used to take, but after the work is done, it should go back to a low power draw.

I contemplated setting up another PC, just for the common, light daily tasks like surfing and email... maybe a laptop or Mac Mini. Ideally I would want my main rig to scale from low power to high power according to demand. CPU throttling does too little, if I don't need the full 4.0ghz, then dropping to 2.7ghz is probably still overkill. I think the throttling should be able to drop down all the way to 1x multiplier (450mhz in my case), which would be plenty for what I'm doing right now: posting on Slashdot. I know there are bullshit manufacturing excuses as to why this isn't possible, but ideally if a 45w machine from ten years ago can handle the current task, then I would expect the latest technology to scale down to 45w (or less) to do that exact same task. Today's power supplies are more efficient, CPUs are supposedly more efficient (on paper), and this machine has more Ram than the old one has disk space. Screw SuperFetch, I could netboot this thing diskless in a heartbeat, so what's wrong with the industry and where do they think they're going ? Will my next PC need a dedicated 30amp circuit to print a Word document ? wtf!

Re:Power consumption since mid-80's?

[ratboy666]

I use a dual PPRO 200 for routing, web serving, mail. Consumption is under 80 watts. For RAID file server, I pressed an IBM GL300 (266 Mhz Pentium II) into service. Only 4 drives at present, though. Seems fully capable of saturating the cable, so I haven't felt the need to replace it (although I have been looking at some lower power systems).

My biggest problem is the MythTV box. Can't for the life of me get the thing to respond to a wake-on-lan packet (the mainboard claims the capability). As a result, the thing is "always on" -- and sucking power. Its a 1.3Ghz AMD.

I use an old 400Mhz HP workstation with an LCD screen as my terminal. Works great. Computations are relegated to another box, which I turn off as much as possible.

Both the PPRO and the PII file server use a UPS, nothing else. So far, its working.

It would be nice to be able to explicitly control the core availability (assuming a power savings) on modern processors. Have it idle on a single core at 40 watts or less, and be able to bring it up depending on use (or expected use). A nice simple command line interface. I won't be going with a quad-core based on power consumption! I will find a mainboard where wake-on-lan actually works, and build a small cluster of those (I want to start investigating near-realtime or realtime video transcode).

Good luck with your rig!

The first one though...

[pigiron]

I'm sure it's true that adding additional virtual servers is only 20 watts. But that first one is a real whopper!

Re:The first one though...

[ToasterMonkey]

IBM mainframes look like some kind of warp reactor. I thought they produced their own energy!

Seriously though... We had a z800 swapped out with a z9 at work. It looked like IBM added a supercharger and changed warp core elements.

Re:The first one though...

[hey!]

Seriously, if you are going to charge a customer what it takes to buy one of these things, it's not that much additional effort to make it look cool. In this case like the love child of the 2001 monolith and a coke machine.

Nice to get a watt/CPU

[Lookin4Trouble]

How about getting a realistic number for BTUs of cooling per HDD/stick of RAM/Processor? my 31 year old Liebert is dying, and the time has come to go to in-row rack-standing AC, but I don't know whether to stick to 2x10Ton or if I need to go for a 3x10 (underfloor in a small datacenter - 30 racks, 250ish nodes). I realize manufacturers have whitepapers out on how much cooling is recommended, but those numbers lie like dogs. "Typical installation: 1 processor, 1 stick of ram, 1 HDD, 1 Power supply" - typical config for my cluster is 4 processors, 8 sticks of RAM, and 2 HDDs on dual power supplies... anyone know where I could get this type of info besides Gartner or the like subscription $ervice$ (yep, they get you coming AND going)

Re:Nice to get a watt/CPU

[guruevi]

Given that you are a small shop, you probably get a detailed energy bill on a monthly basis (not like you're buying bulk energy and nobody's really counting). A little bit of smart engineering would've made sure that you do have a separate counter for your datacenter or some kind of ampmeter where you can calculate what you're actually using (APC power switches in your racks have fairly accurate readings (always round up though)) since you're probably charging your customers depending on their usage.

Since you already HAVE a cooling system, you should know whether that size system was enough, too much or too little and then you can just see what to replace it with.

Re:Nice to get a watt/CPU

[Lookin4Trouble]

thanks, but I am looking to expand (always get more than you need, unless budget constrains, which it does)

Speaking of which, I didn't engineer the room - it has been in place 21 years and I've been here all of 5, and all of the energy bills go to someone else.

The Liebert unit in question is rated 20 tons, but I suspect that age and failing parts have pushed it lower than that, and my average temp rises about .1F/month (72.2F last month)

I suspect that I could get by on 20T (120KBTU/hr), but I'd like to have room to expand, and going with a third unit pushes my price/unit down. Maybe it's time to get with my budget folks to see what project we can put on hold until my room is back up to snuff....

Re:Nice to get a watt/CPU

[gregor-e]

If you isolate the power going to your racks from that going to your cooling, can't you simply clamp an ammeter around the power cables going to your racks and assume that every watt going in turns to heat, then multiply watts by 3.4 to get btu/hr cooling?

Re:Nice to get a watt/CPU

[An dochasac]

While it's a baby step in the right direction, Watts alone as a "benchmark" is meaningless as is Watts/CPU. The VIC-20 likely beat the Z9 back in 1980.

If IBM is serious about server energy consumption, they should publish statistics using the SWaP (Space Watts and Performance) benchmark Sun has been promoting for several years or even "MFLOPS/Watt" or "Page serves/second/Watt" If the Z9 can handle a typical highly threaded webserver load with fewer watts than something like Sun's T2000 Niagara while providing identical performance, IBM shouldn't be afraid to prove it.

Until then, I'll assume it's just another useless benchmark configured specifically to make IBM's products look better than its competitors.

Re:Nice to get a watt/CPU

[Lookin4Trouble]

If you isolate the power going to your racks from that going to your cooling, can't you simply clamp an ammeter around the power cables going to your racks and assume that every watt going in turns to heat, then multiply watts by 3.4 to get btu/hr cooling?

Unfortunately, not quite that easy.

1 * 3-phase powers the Air Conditioning (how much of this is really leaking back into the room as heat?!?)

2 * 3-phase powers the UPS/Generator solution, which continually cycles batteries, changing the load on the line frequently, but is not smart enough to tell me exactly what type of draw is running through. Speaking rough numbers based on "percentage output usage", 65KW on main, ~10KW on secondary == 75KW, which translates to me needing roughly 255KBTU/hr or 21.25 tons of cooling.

Re:Nice to get a watt/CPU

[Anonymous Coward]

I know they're not perfect, but the big three server makers (Dell, HP, IBM) all provide power calculators that give you a better picture of utilization than just base configs. You can add/subtract memory, disks, processors and PCI devices from the servers and still get a fairly accurate picture of utilization.

Dell: http://www.dell.com/content/topics/topic.aspx/global/products/pedge/topics/en/config_calculator?c=us&cs=555&l=en&s=biz
HP: http://h30099.www3.hp.com/configurator/powercalcs.asp
IBM: http://www-03.ibm.com/systems/bladecenter/powerconfig/

Re:Nice to get a watt/CPU

[autocracy]

Well, since the compressor motor is on the outside of the server room, and is the primary consumer of energy, I'm going to say the answer is: "A small portion."

Re:Nice to get a watt/CPU

[Chalex]

Just get a node like that and run your software and measure the power utilization, no? E.g. our IBM 1U nodes: x3550 (4 cores, all RAM full, 2 HDs, 1 PSU) use up ~460W each at full CPU + I/O load.

Re:Nice to get a watt/CPU

[Anonymous Coward]

The SPEC Web benchmark is useful for measuring one very narrow type of workload. The vast majority of servers 90+% are not running that type of workload. It is best to select a benchmark that best represents the workload that your systems are running. Because System Z actually publishes "in the field" power consumption numbers, it is very simple to calculate work/watt. LSPR is a great System Z benchmark as it represents a variety of commercial application running the system above 90% utilization. This is what a typical System Z does.

SPEC Web measures a system running at 100%, with a workload that fits almost entirely in memory. This is going to be the most energy efficient use of this system. Power supplies, fans, etc. are most energy efficient at full load. Most web servers run at less than 20% utilization. Power supplies, fans, etc. are typically very inefficient at these workloads. For example, a typical white box system will have a power supply that is somewhere around 50% efficiency at 20% draw.

I've gotten a bit off track, but I hope it is clear that energy efficiency isn't as simple as a simple benchmark measurement.

Re:Nice to get a watt/CPU

[Lookin4Trouble]

Well, since the compressor motor is on the outside of the server room, and is the primary consumer of energy, I'm going to say the answer is: "A small portion."

ermm... not exactly. Remember we are talking 31-year-old tech here. Compressor motor == in the main cabinet (takes up about 6 racks worth of floor space)

The only pieces not in the room are piping to/from and the radiator grid.

Re:Nice to get a watt/CPU

[RabidChipmunk]

Well, then you have an additional variable. Pulling the compressor out of the room could let you put more servers in without changing the wattage. You'll have to calculate that change yourself though.

Re:Nice to get a watt/CPU

[Frank T. Lofaro Jr.]

Watts per Application User is the real important benchmark, unfortunately you can only figure it out for specific cases.

Re:Nice to get a watt/CPU

[imsabbel]

Every single component shows how much it will heat up your place. Its this little number with the "W" at the end.

Why the hell you guys need two tollay different units for the same (Power) still eludes me, but i guess you just cant drop one with "british" in it...

Re:Nice to get a watt/CPU

[Lookin4Trouble]

Every single component shows how much it will heat up your place. Its this little number with the "W" at the end.

Why the hell you guys need two tollay different units for the same (Power) still eludes me, but i guess you just cant drop one with "british" in it...

So very, very wrong. W for Watt just shows how much power draw. What I'm interested in is BTUs of Heat output. A processor that uses 20W with 100% efficiency will put out 0 BTUs of heat, while a processor that consumes 70W with 20% efficiency (think early Pentium 4 processors) will put out roughly 1KBTU/hr at full load. Where does the line lie, and how can I measure the total output?

Re:Nice to get a watt/CPU

[j79zlr]

I am a mechanical engineer and design HVAC systems. I do not specialize in data centers, but most hospitals, schools and offices nowadays have data rooms. What we would typically do is provide cooling for the amount of power you have. For example, if you have ten 20A 120V receptacles feeding your equipment, then the cooling system should be designed for 10 X 20A X 120V X 3.41 BTU/hr per Watt equaling 81,912 BTUH or around 7 tons.

Make sure you go with a new refrigerant if you going to replace an old unit. R-22 is close to end of life, contractors are trying to sell what they have.

On this subject, my brilliant idea is a data rack with a chilled water manifold, then have rack-mount equipment which would just plug into the manifold as required. You could use small modular chillers and increase as necessary. The problem would be getting the computer manufacturer's to install water cooling inlet & outlet connections on their equipment. Moving air is inefficient and much more costly to install and operate than a few very small pumps and a couple small chillers.

Re:Nice to get a watt/CPU

[j79zlr]

It is basically all converted to heat. There is very little energy consumed in data center equipment for light, sound or mechanical movement. Electrical energy in = heat out.

Re:Nice to get a watt/CPU

[datadigger]

Every single component shows how much it will heat up your place. Its this little number with the "W" at the end.

The Watts on the component is nominal load, you can ony use it to calculate start-up currents, and the rating of the fuses.
The typical load during use (even at full speed) is usually much lower.

energy efficiency has been tackled already

[QuantumRiff]

By these guys back in 2000. The potato powered web server.. We could help our farmers, and power our data centers with beuwolf clusters of potatoes!

Re:energy efficiency has been tackled already

[veganboyjosh]

...if only the sysadmins weren't constantly dismantling the power supplies, dropping them in hot oil, and eating them, that might work!

Re:energy efficiency has been tackled already

[argiedot]

So, the internet is a series of tubers then?

Re:energy efficiency has been tackled already

[dpiven]

Brings a new meaning to the phrase "server farm".

Reason To Buy A CPU

[bostons1337]

Just curious what you guys think about how this relates to buying a CPU. Do you think individuals and companies are going to take a big look at the CPU Energy Use when deciding on buying CPUs? I personally don't think it will become a deciding factor, like processor speed or L1 and L2 cache size, but I think it definetly helps in making a decision.

Re:Reason To Buy A CPU

[foobsr]

"The plan includes new products and services for IBM and its clients to sharply reduce data center energy consumption, transforming the world's business and public technology infrastructures into "green" data centers.

The savings are substantial -- for an average 25,000 square foot data center, clients should be able to achieve 42 percent energy savings. Based on the energy mix in the US, this savings equates to 7,439 tons of carbon emissions saved per year."

http://www-03.ibm.com/press/us/en/pressrelease/21524.wss (emphasis mine :)

I think otherwise.

CC.

Re:Reason To Buy A CPU

[Cecil]

Absolutely it's important. I bought an Athlon64 4400+ EE instead of an Athlon64 4400+ because the energy efficiency means it's energy efficient (obviously), low heat (less obviously) and therefore my system is much quieter (even less obviously), and uses even less energy to run fans. Energy efficiency has a lot of benefits for a computer. Once people start realizing this and trying it, they will see the light. There's absolutely no reason for your computer to sound like a jet engine (or, as the fans wear out, like an angry lawnmower). It means less dust, less maintenance, longer life for components, it's silly not to consider energy efficiency when buying a CPU.

Now I'm just waiting for GPUs to go down the same road. I'm pretty sure my Radeon X1900 XT puts out ten times more heat and noise than the rest of my system combined.

Re:Reason To Buy A CPU

[bmajik]

My newegg order just showed up last nite. I wanted a machine that was as silent as possible, so I got an AMD BE2350 (the 45W TDP dual core Athlon @ 2.1ghz), an MSI k9 platinum (heatpipe cooling for the chipset), and a Gigabyte "silent-pipe" 8600 GT card.

For a power supply i got a seasonic 330w S12 (variable speed ballbearing fan).

My computer is entirely fanless except for the stock AMD CPU fan and the Seasonic power supply fan. There's not even a case fan. System and CPU temps seem to be stable around 40C.

My vista "index" is 5.0, with the 5.0 being the lowest number and coming from the CPU.

I wanted a really quiet machine. That meant eliminating fans. That meant buying energy efficient parts (the CPU and the Seasonic PS are both spendier than equivalent parts that don't stick to a tighter energy budget). But the machine _is_ quiet. I've got a kill-a-watt at home that I haven't tried out yet but I hope to see less than 100w of consumption. My old socket 754 machine is 5w sleep, ~100w booted but idle.

I'm also going to be consolidating my "always-on" applications (file serving, possibly BT) onto a Windows home server machine so that i can have my other boxes power-save as much as possible without any real service interruption. Having a few songs here, a few videos there, etc means that I can't keep the majority of machines sleeping the majority of the time (WOL is pretty spotty IMO.. if you configure WOL such that a machine "can" wake, it usually will stay awake from other network noise)

One of the other things i bought with this order was a new UPS. Sticking to a smaller power budget has other interesting effects -- like you can get away with a smaller (and cheaper) UPS to get the same amount of uptime.

Re:Reason To Buy A CPU

[RenderSeven]

With 12 systems running 24x7 my power bill takes a pretty good hit. But every time I think about reducing it, it occurs to me that every watt consumed ends up as heat, and I heat 9 months of the year this far north, and since I heat with electric, its a wash. Cut computers, more baseboard heating, same btu's and same electric bill. If I could switch to more efficient heating system then CPU efficiency might matter, but right now I cant see how it helps. Instead all I care about is MIPS per BUCK (everything is relatively quite these days).

Re:Reason To Buy A CPU

[Spoke]

One of the biggest power draws these days are graphics cards. Often, graphic cards will draw as much power as the rest of the system. It's typical for the cheapest graphics card to add 10 watts to the power draw of the system when idle, with 25-50+ being common for medium to high end cards. Unfortunately, current graphics cards don't do much in the way of reducing power draw when idle.

If you don't do any serious gaming, sticking with the onboard graphics will often reduce power draw significantly. If your mobo doesn't have onboard graphics, picking an inexpensive fanless graphics card will draw the least power.

If you were using onboard graphics, I would expect your system would idle around 55w (+-5w or so). Peak power draw would be less than 100w. With the GPU you're using, I'd guess that it adds add 10-20w at idle and another 50w at peak. It'd be interesting to see what the actual numbers are.

Something people often forget is that a good PSU with active power correction will also significantly reduce the apparently load on a UPS (as well as the grid if you don't have a UPS), not to mention that PSUs with APC are normally significantly more efficient. For example, if your system draws 100w but your PSU has a power factor of .5, you are actually pushing twice as much current through the AC line as a system drawing 100w but a power factor of 1. This ends up doubling the load on your UPS if you have one.

These days it's fairly easy to build a system which idles below 50w as long as you're informed. A bit more research will get you something in the 30-35w idle range if not lowre. I do wonder what you had in your old Socket 754 machine which caused it to idle at 100w. I suspect it had an inefficient PSU and a mid-high end graphics card or wasn't using Cool'n'Quiet. All recent AMD systems I've seen which support Cool'n'Quiet idle at 60w or less unless you have a power sucking GPU.

Re:Reason To Buy A CPU

[timeOday]

Do you think individuals and companies are going to take a big look at the CPU Energy Use when deciding on buying CPUs? I personally don't think it will become a deciding factor

I'd say we've already turned that corner. Intel aborted the netburst architecture (P4) because there was no easy way to dissipate more than a couple hundred watts, in that way power became the limiting factor. I don't think we'll see desktops reverting to 10W processors, nor do I think a 10% difference in consumption between competing products will be a big deciding factor, but I do think the exponential growth in CPU power requirements seen for the 20 years leading up to 2005 or so are over, simply because heat dissipation becomes an annoying issue as you approach 1KW PSUs. So I think attention has turned to how to get increasing performance from basically a fixed amount of power.

Re:Reason To Buy A CPU

[tksh]

FWIW, I have a 754 with four hard drives and a fanless card. It idles just below 100W and at 28C. I also have a Mac mini that idles at 30W.

Re:Reason To Buy A CPU

[Spoke]

Sounds about right. Each 7200rpm hard drive will add close to 10w during idle.

WD recently launched some low power drives, but they don't spin at 7200rpm. Performance is quite good considering and power draw is about half of a comparable 7200rpm drive.

Is this true?

[Anonymous Coward]

So, now computers run on gas/oil? Did I just wake up after 20 yrs in coma?

Re:Is this true?

[nschubach]

Well, unless your positive that your local energy company is nuclear, solar, hydroelectric, thermal, or wind power... then your computer runs on gas, oil, or coal.

How big a fraction?

[columbus]

This looks like a positive development.

It seems that the computer industry on the whole has become more concerned with energy efficiency over the last few years. I'm glad to see it. As a discipline, computer science is always looking for ways to eke out more efficiency, whether it is at the algorithmic level or at the level of chip manufacture. It seems to be a be a natural fit to extend this thinking further into energy consumption as well.

But I have to wonder, how much of a difference can we make? I think that the energy consumption involved in the field of computers - through the whole lifecycle: manufacture, operation & disposal - is relatively low compared to the energy consumed in other areas of the economy (transportation, heating, lighting, manufacture).

Would we be better off spending our time optimizing energy consumption in other industries?

Re:How big a fraction?

[Seenhere]

The best estimate I've seen recently is reported here: http://uclue.com/index.php?xq=724

(Doesn't include energy for manufacture or decommission; just operation.)

Result: Roughly 10% of total energy consumption is due to computation.

Re:How big a fraction?

[thatseattleguy]

I seem to recall statistics (noted on Slashdot, where else...) that came out of the Department of Energy a week or so ago that the energy required to operate the internet, computing, and communications infrastructure in the US was something on the order of 10% of the total power output.

Even if that's wrong by a factor of 2, it's still significant - especially as it's by far the fastest-growing segment of power consumption in this country (and, presumably, in most others).

Re:How big a fraction?

[DamonHD]

More than a factor of four in my case (more than 20x by some narrow measures):

http://www.earth.org.uk/low-power-laptop.html

and

http://www.earth.org.uk/saving-electricity.html

Basically I replaced my entire 670W rack of Web-facing servers at home with a single Linux laptop that uses ~18W off 12V DC (+7W wasted in the mains adaptor), which sometimes now runs off-grid on solar PV so that 12V DC power figure is meaningful.

I have no reason to believe that my situation is really exceptional, and I'm running a fairly standard mix of servers (NTP, SMTP, DNS, HTTP both flat-file and Java-based)...

Rgds

Damon

Useful for consumers as well

[Kelbear]

I had concerns about power draw when setting up a new PC to replace my aging setup. I've already piled on a considerable amount of electronics in the small room I live in. Having both A/Cs activate on the same breaker causes the breaker to trip, so even the added burden of a computer upgrade was worrisome.

Having the additional information would have taken off a good bit of stress, and would help a bit in calculating how much headroom I needed in the PSU to keep the PC itself running smoothly.

Tax benefit

[slackoon]

The government gives tax benefits for driving hybrid vehicles and I believe they should do it for energy efficient computers as well. "According to the Computer Industry Almanac Web site, at the end of the year 2000, there were 168.84 million computers in use. The projection for the end of 2001 is 182.24 million." So just imagine how many there are now! With that many computers, many of which are never turned off, the energy savings could be enormous.

Re:Tax benefit

[Yartrebo]

Not having a computer is better in terms of energy demand than even having a very efficient computer. Same for the hybrid car. For this reason I find these tax breaks perverse.

Giving tax breaks for efficient items penalizes those who conserve the most by not even having the item or by using less. A business that invests money into writing more efficient software and using less servers should not be penalized vis-a-vis a business that invests the money into more efficient servers.

Re:Tax benefit

[Lord Ender]

Not having a computer is better in terms of energy demand than even having a very efficient computer.

You're saying that it is more energy efficient to have to drive to the bank than to do your banking online?

Owning a computer is critical to functioning efficiently in modern society. Offering incentive NOT to own a computer is like offering incentive NOT to learn how to read and write; it's totally counterproductive.

Re:Tax benefit

[slittle]

People who "conserve" by not buying things are not contributing to the economy, tax revenue or technological progress, so why the hell should they get tax breaks?

Expensive toys fund the economy, the government and advance the state of the art so that everyone can enjoy a better life while using less resources. Your theory on conservation is simply useless (at best).

Re:Tax benefit

[Yartrebo]

I was talking about computer servers, where many tradeoffs can be done between software and hardware. Giving up a home server (massive energy hog for the services rendered) is a lot easier than giving up a primary PC. Likewise, it is much easier for a company to improve its software and buy less servers than to make do with fewer PCs.

PS: Probably the way with the least negative side effects to improve efficiency would be a blanket tax on all dirty, nonrenewable, or almost fully developed energy sources (in my opinion: coal, oil, natural gas, peat, bitumen, oil shale, geothermal, nuclear fission, hydropower). Minor energy users (such as computers) will be lightly taxed, while energy hogs (such as airlines, furnaces, autos, heavy industry, and ACs) will be heavily taxed.

Re:Tax benefit

[gedhrel]

I don't quite understand your jump from economic growth to a better life for everyone. The poverty gap increases; everyone (for the proper meaning of the word) clearly aren't all having a better life; and economic growth is a poor straw man used by politicians, compared to measures of actual welfare.

Re:Tax benefit

[slittle]

Well, I don't live in the USA so perhaps I take the lack of corporatist corruption for granted...

True, economic growth doesn't necessarily make life better for everyone, but it at least makes it possible. Stagnating doesn't do society or the environment any good.

Re:Tax benefit

[Zeality]

Hey, I couldn't find a PM option, so I'm hoping you get e-mail notification of comments to you, Yartrebo: I ran across an old post of yours at the Game Dev forums in which you spoke of ripping models via AdriPSX in 2002: http://www.gamedev.net/community/forums/topic.asp?topic_id=93782 The post's links are dead (not surprising since it's 5 years old). At the Chrono Compendium, a bunch of us are obsessively pulling out all the stops to map Cross's file structure and rip its assets, but the community isn't the most active one and so we're not getting much expert help. If you remember any details, could you provide any advice? http://www.chronocompendium.com/Forums/index.php/topic,3684.0 Huge thanks for considering.

Performance?

[Speare]

Okay, just stating the wattage is like stating MPG for a car or the energy usage for a fridge. But every year, car performance stays about the same or gets worse, and the fridge ain't getting more full. There doesn't seem to be a single useful energy metric that can drive informed purchasing decisions.

So how do you deal with CPUs that are twice as powerful in the next product cycle? The wattage will be about the same, but the amount you can get done with that chip will be much higher. It's like next year's car suddenly weighs twice as much, or goes twice as fast, or seats two whole families, while getting the same mileage. You can't even consider it in two tiers like "passenger cars vs truck frames" because you have to deal with 2008, 2009, 2010, 2011 performance tiers... they change all the time. How can someone make an informed decision from this?

Re:Performance?

[DMUTPeregrine]

Watts per MFLOP. Or MIPS. Or Watts per Point, where point is an average on some benchmarking system. Just giving watts for a computer is like just giving gallons for a car. You don't know how many miles it can go on those gallons, and so the figure is useless.

Re:Performance?

[Uerige]

For a typical desktop system? Try getting the lowest absolute power x86 compatible cpu (speed doesn't matter). It certainly won't idle any slower than any other one, but it might do it more silently. These kind of metrics aren't meant to be interpreted by anyone and their mom.

Now if you're building a datacenter or, possibly a gaming rack, take a look at all of the processors that are actually fast enough for your demands and then pick the one that uses the least power under your actual expected load.

CNN is not "reporting" anything.

[fm6]

Read TFA carefully. It's an IBM press release! CNN is just passing on content they got from http://marketwire.com/>MarketWire, which is a PR channel, not a news organization.

*REAL* programmers...

[Anonymous Coward]

*Real* programmers eat microwave popcorn. However, they use the heat from the CPU to cook it instead of using an actual microwave. *REALLY* good programmers can even tell which process is running by the rate of popping.

Re:*REAL* programmers...

[PrescriptionWarning]

Maybe you're referring the the programmers which throw poop and are just as likely to write a shakespeare as they are to write a decent program.

thats right. codemonkeys.

Re:*REAL* programmers...

[FireBreath]

you have something against us poop-flinging shakespeare coders? :)

Yeah but Apple Trounced them

[clare-ents]

We colo AppleTVs. Why? 1Ghz Core Solo, 18W. We also do Mac Minis. Why? 2x2Ghz Core 2 Duo, 40W. Let's put 125 Mac Minis, up against the IBM mainframe and see who's faster.

http://www.mythic-beasts.com/appletvdedicated.html

What goes around, comes around

[Dorceon]

One of the top posts on that discussion was an Al Gore joke of a completely different nature than the modern Al Gore joke.

What unit are you measuring the watt cost of?

[Anonymous Coward]

With cars, you measure gallons of fuel cost per mile of movement.

As most of the purpose of cars is moving miles, there is little debate over the unit (miles). But little doesn't mean none - for example, part of the purpose may also be to move multiple passengers, so even if you have a car with a very high mpg, people would object and call the measurement inappropriate if the car e.g. only took a single passenger.

What's the unit of cost for CPUs? Gigahertz? If so that means an overclocked P4 has the same watt-independent goodness as a Core2duo. Is that right? Number of movements in a database per second? Maybe?

Out of genuine interest in how meaningful these numbers will be, is there a standard "unit of work" to apply to CPUs?

miles per gallon?

[oyenstikker]

How about "burning library of congresses"

"It's what the EPA wants" !!??

[scottyokim]

The most interesting quote of the story: "Over time every vendor is going to be asked to provide typical energy use numbers for their equipment. It's what the EPA wants, ..." This is a new use of the word "asked" that I'm not familiar with ... And it's not about saving the customers money, it's about what the EPA wants??

Interesting, but what does it mean?

[pclminion]

These benchmarks are interesting, but are they relevant to real life? There are too many factors to say.

Consider a 20 watt CPU which sits idle 99% of the time. Then imagine a 40 watt CPU which is loaded to 100% all the time. Which is "worse?" I'd say the 20 watt CPU is worse, because it's 20 watts of completely USELESS power.

Or imagine that a corporation has a cash-cow application. They can make $10 million per year if they run it on server X which draws 2000 watts. Or, they could make $5 million if they run it on server Y which draws 1000 watts. Is server Y really "better" than server X just because it draws less power? The company would need two Y boxes to get the same performance, and then they'd be back up to 2000 watts again.

Measuring things like "cycles per watt-hour" or even just pure power usage don't really say anything about whether a CPU is preferable in any particular circumstance. Let's invent a unit called the "benefiton," which is a single unit of "benefit to humanity." We really want to optimize "benefitons per watt-hour," not cycles per watt-hour, dollars per watt-hour, or anything else. But defining what a benefiton really is is almost impossible.

Just measuring which processors consume certain amounts of power isn't going to help us optimize our usage of energy on this planet.

"typical" energy usage harder than it looks

[walterbays]

Yes it's great that EPA is working with industry to get some more meaningful power ratings of computer equipment. But it's not easy to give "typical" energy use. What's typical gas mileage of your car? Does a 25 MPG EPA rating mean you'll get 25 MPG? How heavy is your foot, where do you drive, how far, how hilly, how many stoplights, etc.?

From the SPEC power benchmark you'll get not just a single composite number, but all the details so you can judge what is most relevant to your own system. Is it idle most of the day? Then look at that end of the graph, and check the power management software settings to automatically suspend. Does it run flat out? Then look at that end of the graph. For even better information you can test your own systems with your own workload.

The marine diesel engine industry

[dj245]

In the marine diesel engine industry, there is a measurement of NOx (nitrous oxides), usually measured in grams per killawatt-hour (g/kW-hr). But not all engines will be used in the same service, so they won't be running at the same load. Some will run at 100% load most of the time they are on (generators, fire pumps maybe) while others will run at about 65 or 75% of full power all of the time- these are your direct-drive propulsion diesels. These different duty cycles have a dramatic effect on the numbers. So what to do?

The International Maritime Organization has created a few different cycles- E2 is Constant Speed Main Propulsion, E3 is Propellor law operated propulsion for example. You pick your cycle, run your engine at a variety of loads, then use weighted averaging on those loads to determine what the emissions would be if the engine ran at E2 all the time. Then you can say that for the E2 cycle, the engine puts out so much NOx.

For computers, someone needs to come up with some different computer cycles. There may be several of them- 50% parallelizable with 25% floating point and 75% integer math, 100% parallelizable with 100% floating point math, etc. Different architectures may take dramatically longer to do floating point or non-parallizable workloads. Only then could you run a bunch of tests and really say that under this load the computer uses this much power to do a certain amount of work in a given amount of time.

This is not new or novel stuff. This is similar to how the EPA tests cars. Some cars do highway miles much better than city miles, so they do both and weight the averages.

Elevators

[G3ckoG33k]

We were 450 people working at my previous job. More than 400 used computers. The elevators of that building used more than 10 times the current of the computers.

You use RESULTS per Watt

[Colin Smith]

The results in terms of a car are miles/kilometres travelled. In terms of computers, MIPS or MFLOPS are however not results, they are performance measurements. Using them would be like describing car efficiency in RPM per gallon. Not the results you're after.

So the first thing to do is define what your results are. The results computers produce are the "bits of information you want".

SPEC and TPC both have benchmarks which already attempt to describe the results that customers are after.

http://www.spec.org/
http://www.tpc.org/

Get a WattsUp? Pro...

[jddj]

Get a WattsUp? Pro (from think geek...).

My dual-core Athlon MythTV server uses 95 watts at the peak use I'm making of it. Start up current spike is pretty high, but the "costs more to run it than to power it on" line is complete BS. I can measure - don't have to assume.

Running my server 24x7 would be like powering a 100 watt light bulb 24x7. It's not a perfect analogy - the server uses energy in a much more complex way than the simple resistive load of a light bulb - but it is a USEFUL analogy. Gets you to the right order of magnitude.

BTW, this 95w figure for the MythTV box encouraged me to move my only 24x7 app - mt-daapd music server - to a Linksys Slug running Debian and a USB hard disk with spindown. The Slug and USB drive use 6 watts when running, 1.5 watts when inactive. Much better than 95 for the MythTV box.

Love my WattsUp? Pro, and my Slug.

Finding the greenest OS

[UberDude]

Depends entirely on what the processors are doing, just as the equivalent metrics for cars depend on the weight and speed of the car. What I'd like to see is a rating of the load that, for example, an operating system puts on the CPU. Let's say that you want to compare Vista vs Mac OSX or some brand of Linux OS, doing some basic activities (starting up, copying data, etc), and give the results in a way that translates into carbon units.

Athlon at 654MHz

[ano22000]

my Athlon Xp mobile 1800+, run at 1,5 GHZ. But if I do "powercfg.exe /change profile /processor-throttle-ac CONSTANT" with winxp, it run at 654MHz. It's enough for old games for example.