Chipmakers Admit Your Power May Vary 138
Dylan Knight Rogers writes to mention a News.com story discussing the realities of chip power consumption. From the article: "Assessing only pure performance is passe. The debate these days is about performance-per-watt, which seems like it should be a simple miles-per-gallon type of calculation. However, miles are miles, and gallons are gallons. There's no one simple way to measure processor performance, and measuring the amount of power output by today's chips is proving just as difficult."
Re: (Score:2, Insightful)
Re: (Score:1)
Re:Transmeta thought so to! (Score:2)
How is this news? (Score:5, Insightful)
Re:How is this news? (Score:3, Insightful)
Re:How is this news? (Score:1)
Possible? Sure. Easy? Well, modern CPU power converters have several physically-distributed power outputs that don't share the load equally, they drive multiple load pins that don't share the load equally either, and they only tolerate a fraction of a milliohm of added resistance. There can also be a big question about
Re:How is this news? (Score:3, Informative)
Say what? They're not [i]that[/i] intolerant. Otherwise the overclockers wouldn't be playing around with increasing the voltage. Normal power supplies would have to be far better, and motherboard power compensators far more expensive. Besides, if your measurement device adds that much resistance, you simply increase the voltage of the rail a smidgen to compensate.
Now, I am talking about doing all this in a lab, for best results.
The true diff
Re:How is this news? (Score:1)
Yes, they are. Consider a high-end CPU that draws 100 watts at 1.2 volts. That's 83 amps of current. If you add a one milliohm series resistor for measuring the current, you've dropped the CPU voltage by 83 millivolts. By comparison, I just looked up an Opteron and it was only specced for a +/- 50 mV change.
Certainly the measurement is doable. I'm just saying that the cheap and dirty approach
Re:How is this news? (Score:2)
As for determining what sort of load to run under, how about do some research and see what people use the chips for? It's good for other aspects of the business too.
All in all, I don't buy the argument that Intel dosesn't know
Re:How is this news? (Score:2)
Re:There are other ways to measure current... (Score:2)
No resistance? Even a hundred micro-ohms would be significant, and it takes a pretty impressive chunk of copper to get that low. Using a Hall effect sensor in the midst of a 100 amp circuit is tricky, too.
NO issues? You mean besides the custom adapters built just for testing, and $250k ATE machines. That's a f
Re:How is this news? (Score:2)
Well, modern CPU power converters have several physically-distributed power outputs that don't share the load equally, they drive multiple load pins that don't share the load equally either, and they only tolerate a fraction of a milliohm of added resistance.
So what? Use a known set of peripherals with known power load and a known PSU, then measure power usage at the wall. Simple.
Re:How is this news? (Score:2)
It's not as simple as you might think. How do you know how much power your "known" set of peripherals use? I'm sure that there are power specs for just about anything, but I'm sure they just like the specs we see for CPUs at the moment - close but still an approximation. Trying to get an exact measurement of power used for something like a graphics card or a motherboard is going to
Re:How is this news? (Score:2)
How do you know how much power your "known" set of peripherals use?
Measurement. You can minimize error by testing in a known setting - a C800 with integrated video, for instance. Add a PCI card as a secondary display and measure the difference. Same goes for disks.
Trying to get an exact measurement of power used for something like a graphics card or a motherboard is going to have the same problem of measuring CPU power usage
Separating motherboard power from cpu power is problematic. If you can under
Re:How is this news? (Score:2)
Re:How is this news? (Score:2)
Re:How is this news? (Score:1)
its still and always will be an estimate.
Meaning take the published specs and don't assume they are maximum.
Its like car manufacturers telling us we will get 40mpg but when I'm dragracing I get closer to 5.
Of course, YMMV.
Re:How is this news? (Score:1, Troll)
The Slashdot Community,
http://slashdot.org/ [slashdot.org]
10 Jun 2006
Re:How is this news? (Score:2, Insightful)
it should be a simple miles-per-gallon type of calculation. .
It is. That's the problem. Mr. Krazit seems to be utterly clueless. I defy him to predict the milage I get the next time I go out for a drive.
Hasn't he ever noticed, like most of the rest of us have, that the milage he gets is not actually the same as the EPA test "prediction"?
That's because the EPA test only gives valid results for. .
Not indicative of real world performance (Score:5, Funny)
benchmarks (Score:4, Insightful)
Is this perfectly scientific? No.
Is it practical? Hell yes!
Re:benchmarks (Score:1)
Re:benchmarks (Score:2, Insightful)
Some miles are up hill and some are down hill... (Score:5, Insightful)
BTW, EPA mpg are measured without using real mile on real roads.
Re:Some miles are up hill and some are down hill.. (Score:4, Insightful)
No offense, but nothing is perfect. This is why we have a thing called "standard deviation".
Me hitting the letter "e" will probably not take the same amount of energy to process twice. But I bet over 1000 e's the standard deviation could be found and would indicate that 66% of the time it's "x J +/- y" and so on...
So you sample something like "building the linux kernel to a ram drive" 100 times, find the deviation and use that. The tighter AND lower the better. The wider and higher the worse.
Tom
Re:Some miles are up hill and some are down hill.. (Score:2)
There's a reason I ask people what they intend to do with their PC before selling them one -- do they need more ram, or more drive space or more drives or a bigger video card
Very rarely does CPU speed come into the equation; the slowest CPU available at retail is quite fast enough for most people, most of the tim
Re:Some miles are up hill and some are down hill.. (Score:2)
Why?
Me hitting the letter "e" will probably not take the same amount of energy to process twice. But I bet over 1000 e's the standard deviation could be found and would indicate that 66% of the time it's "x J +/- y" and so on...
But with different usage-patterns you *will* get consinstant differences..so sure, you'd get data, and they'd be valid, but that doesn't mean GP needs a stat-class. Regardless of how much people refuse to believe it, even in todays 'massage-the-da
Re:Some miles are up hill and some are down hill.. (Score:2)
Most likely with the CPU/memory under full load the Wh deviation is less than 10% of the mean usage. On a typical desktop the Wh rating is about 200-250 at full load. If you see a variance of more than say +/- 20Wh something is wrong or the test isn't reproducible. If you think things like differing occurences
Re:Some miles are up hill and some are down hill.. (Score:2)
Because, as the parent post pointed out, the problem of how to describe the measurement of a varying quantity does not actually pose a problem.
But with different usage-patterns you *will* get consinstant differences.
I agree the "hitting E" example seems a bit odd, but I would guess the parent just needs a computer science course badly.
Instead, do the same with a few typical real-world usage patterns (arranged into a repeatable suite). Then divide the performance value by the watts value,
We have a nation of SUV's (Score:2)
Our nation is one of conviences, not of caring if our grandchildren have conviences.
Re:We have a nation of SUV's (Score:2)
Re:We have a nation of SUV's (Score:3, Insightful)
Let's see... processor running full steam instead of low power mode when idling probably amounts for a waste time of more than 90% (unless you work/live at the box).
Opteron at full == 95W, at low == 35W, diff 60W. price per KWh is about 7 to 10 cents. Let's say 8.5 to be close to middle. 60W * 24 * 31 * 0.085 = $3.80 per month. Probably double that once you factor in power supply inefficiencies and cooling costs. So you sp
Re:We have a nation of SUV's (Score:2)
Sure maybe if you ALREADY HAVE said box might as well try to get more mileage out of it. I wouldn't buy a lower rated part for that purpose though...
Tom
thanks Intel for the Pentium 4 (Score:2)
Re:We have a nation of SUV's (Score:1)
Our nation is one of conviences, not of caring if our grandchildren have conviences.
You are obviously not running a server farm composed of hundreds or thousands of CPUs running at full load 24x7. Not only do you have to power those procs, but you have to cool them as well. For almost any server, the cost in energy over five years is going to be more expensive than the chip itself. Energy requirements are cal
Re:We have a nation of SUV's (Score:2)
The average desktop consumer certainly doesn't. However, performance per watt is very important to two segments of the population:
1.) Laptop users. A high performance per watt, and more importantly a low wattage in general, means you can get more things done on a single charge. The average consumer doesn't care about money, but they do care about time.
2.) Corporations, especially those with large serve
Re:We have a nation of SUV's (Score:2)
Big corps really hate "new" stuff. Despite the fact most boxes are already capable of idling a lot cooler few [that I've seen] actually use it. Nothing like 200 quad processor Opterons running at full speed because enabling a kernel module is "hard".
Tom
Just as well... (Score:1)
May Heisenberg will protect us!
Re:Just as well... (Score:5, Funny)
I have no idea what direction you are going with this. I have, however, determined exactly how fast you are going with this.
Re:Just as well... (Score:1)
Re:Just as well... (Score:1)
Dang it, that's what happens when you revise too quickly... er... I mean, the superposition didn't resolve!
Well... (Score:4, Insightful)
Re:Well... (Score:3, Informative)
This takes in slow city traffic, faster freeway traffic and top speed travelling, approximate to an average consumers car usage.
It's not just the CPU (Score:5, Insightful)
The best proof that modern software makes modern hardware suck is that, back in the mid-eighties, I used an Atari ST to do desktop publishing, and it wasn't all that different from what I can do now with a simple PC that would look like a supercomputer back then.
Re:It's not just the CPU (Score:5, Interesting)
Because of advances in engineering and design, engines are far more powerful and efficient now than they were in the early 90s. Cars have far better aerodynamics. However, gas mileage has not improved. In many cases it's gotten worse since the 80s. Likewise 0-60 times haven't improved much.
So what happened? Instead of improving the performance of your average family sedan, auto makers have added better armor, more air vents, more lights, DVD players, and more plush materials. Everything is safer and more comfortable now than it was in the 80s and 90s.
My 86 Camry will beat your 2007 Camry in a drag race and it will get better fuel mileage. But for a 500 mile trek across California or a bad accident? I know which one I'd prefer.
Likewise, my Pentium 4 has 16000 times more ram than my first computer (a C64,) and 256 times the ram of my first 486 (side note: how long before someone informs me of the amount of ram my 486 had?)
My 486 could write a document just as easily and with as much style as my P4. But it couldn't write a document while I was watching a subtitled MP4 movie in another window, listening to music, burning a DVD, and downloading hot lesbian pr0n from bit torrent. And it certainly couldn't do all that on dual 20 inch widescreen flat panel displays.
Sure, software is more bloated. But like the 2007 Camry (available wherever fine cars are sold,) after a long day your ass is going to be a lot more comfortable.
Re:It's not just the CPU (Score:2, Funny)
My ass has never been as comfortable as it was in my '67 Triumph GT6. Don't ask about my right leg resting against the tranny case though.
When I wear shorts I just tell people it came from shrapnel at Anzio.
KFG
Re:It's not just the CPU (Score:1)
KFG
Re:It's not just the CPU (Score:2, Informative)
I recently read an article noting that the 2007 Camry (with 250HP or whatever) will out drag race most sports cars from the 1980s, much less the sedans. Although, greater point taken about bloat.
Re:It's not just the CPU (Score:2)
Re:It's not just the CPU (Score:2)
Re:It's not just the CPU (Score:2)
Re:It's not just the CPU (Score:1)
Re:It's not just the CPU (Score:1)
Re:It's not just the CPU (Score:2)
Re:It's not just the CPU (Score:2)
when I was young....
TTL was hot and fast
CMOS was cool and slow
a fave joke was about the Russians designing the worlds largest micro chip
these days its not a joke its reality
CMOS is hot enough to cook an egg
hard drives shut down from over heat
components accelerate their aging when running hot which they always are
in summer you have to be in air con or have multiple fans
I now always have a 4" fan on my cpu
V8's are becoming dinosaurs
the next generation is 128bit p
Re:It's not just the CPU (Score:3, Insightful)
Right now, you have machines that will do amazingly powerful things, especially with the Next-gen coming out. So what do you do with that power?
Quite simply, one of the things you can do is optimize less.
For example, early FPS games were written largely in assembly in an attempt to eke out every bit of power from the system. It worked, but it was really expensive financially and broke
Re:It's not just the CPU (Score:2)
Re:It's not just the CPU (Score:2)
Yep - here's the same software running on a more recent machine [nvg.org]. My old ST ran Thing, qed, Papyrus, CAB and so on just fine - although in 640x400 monochrome. Multitasking with Geneva worked very nicely, and there was always MiNT for all the UNIXy stuff. WYSIWYG was more than possible with NVDI, which let me use Truetype fonts in all GEM applications.
I eventually saw sense and bo
News? (Score:5, Informative)
I've got two x85 class Opterons sitting here at 1Ghz most of the time. That's ~35W vs. ~95W. AMD seems to care about power. Intel is no worse off with the Pentium M and "core" series (netburst was a mistake).
Tom
Re:News? (Score:1)
Re:News? (Score:2)
Tom
What a day. (Score:2, Funny)
Looks like everyone's coming out of the closet today.
Re:What a day. (Score:2, Funny)
Re:What a day. (Score:2)
Re:What a day. (Score:1)
I'm only the 18th nervous breakdown!
Waaabababababa pbbbbt! Hooo!
YMMV (Score:1)
I don't want it to be like buying industrial lighting where you have to compare 60W bulbs with 800 lumens to 75W bulbs with 1000 lumens.
I'm not confused by simple linear ratios: We just don't need naming conventions and measurements whose only purpose is to obfuscate easy
comparisons that would allow for simple commodity pricing of a consumer good (which is what y
I would like to know... (Score:3, Insightful)
Re:I would like to know... (Score:3, Interesting)
Raise the clock and the charge time on the traces goes down, means you need a higher voltage. Think of filling a bottle with a small hose. If you want to fill a single bottle faster you have to increase the pressure [voltage]. Also raises the current overall if you keep it up. This is why overclockers often have to raise the voltage of the part they are OC'ing.
Raise the clock and more transistors are swi
Re:I would like to know... (Score:1, Flamebait)
Here's the short answer to the GP:
A fixed amount of energy is needed for any computation (dividing a number, or flipping the output of an inverter), the amount of energy depends on the architecture or process, but just pick any value for now.
Power is the integral of energy over time.
As the frequency increases, that same amount of energy mentioned before is needed in a shorter amount of time. Hence pow
Re:I would like to know... (Score:2)
You're right, I'm not an EE. But I have worked at enough hardware firms to know that raising the clock does more than "raise the work per time period".
Tom
Re:I would like to know... (Score:2)
Raise the clock speed without altering the chip: more work is done per unit time.
Therefore energy requirements per unit time increase (by definition).
Therefore power requirement increases (by definition).
This extra capacitance crap, etc., explains non-linear increases. But the OP wasn't asking about rate of increase, he was asking about increase, period, which can be answered with grade 9 physics.
Re:I would like to know... (Score:2)
Yeah, the simple answer is "more shit is happening for a constant unit of time". The more accurate answer is the circuit is less efficient requiring more energy to operate at a higher frequency.
That also explains why you can't scale indefinitely without the chip melting. If was a matter of work you could just duty cycle it. Sure your thr
Re:I would like to know... (Score:2)
Well...
Frequency = 1 / Period.
Period is duration.
Assuming a circuit could function at twice the frequency, the SAME amount of energy is required for each edge transition, assuming a static digital CMOS circuit: each capacitor must be charged to create a field to create a channel to charge the load cap, etc etc etc... Putting aside topology, or the entirety of electrical engineering for that matter, this is essentially Feynman's lecture on the thermodynamics of com
Re:I would like to know... (Score:2)
It depends on the specific values of the clocks, chips have an optimum point. If you take about a chip with the optimum at 1GHz and run it at 500MHz, at the same voltage, it will need more energy mostly because of leackage. Now, if you put the same chip at 2GHz, you'd still need more energy, mostly because of the increased resitence of the components
Re:I would like to know... (Score:1)
http://forums.anandtech.com/messageview.aspx?catid =50&threadid=1867448&STARTPAGE=1 [anandtech.com]
I remembered seeing it a few days ago.
Re:I would like to know... (Score:3, Informative)
When not changing state, a CMOS device dissipates almost no power. But each CMOS gate has a tiny capacitance that must be charged or discharged each time it changes state. This requires energy. The energy dissipated for each transition is essentially constant, but the number of transitions in a given time can vary. Since power=energy/time, the mor
Re:I would like to know... (Score:1, Informative)
This happens because of the way CMOS logic works. Being Compilimentary Metal-Oxide Semiconductor logic, every gate has p-type and n-type transistors. n-type transistors can only drive a '0' but p-types can only drive a '1' so both ar
For the laptop users (Score:1)
It seems some want to eliminate the time component from speed measurements, so you'd only care that one machine got to 110,000 calculations versus another getting to 120,000 calculations.
With desktop machines, just hook up each computer to a 1000VA battery backup UPS and see how FAR each gets ... not whether one got to 100,000
Thought it was "chipmunks" (Score:1)
This is just marketing (Score:1)
Re:This is just marketing (Score:2)
The problem before hand was that the designs were just inefficient. It took your K6-2 or P2 running at full tilt to keep up with demand. Scaling didn't make too much sense. Now a 500Mhz K8 can cope with most usage, playing mp3s takes less than 1% of the cputime where it used to take more than 80% on a 486...
I wouldn't call the power savings as a "new scam" or trick
Re:This is just marketing (Score:2)
Semprons don't, which came as a bit of a surprise when I tried to configure clockspeed control and found that it wasn't working.
Re:This is just marketing (Score:2)
Tom
Re:This is just marketing (Score:2)
The BIOS recognizes it just fine. For whatever reason, AMD either disabled clockspeed control in Socket 754 Semprons or didn't include the necessary circuitry for it (depending on whether Semprons are Athlon 64s with certain functionality (such as half of the cache) turned off or whether they're a completely different design.)
Re:This is just marketing (Score:1)
Only the server guys ever cared about CPU usage. That's why a SCSI operation utilizing 20% over 60 seconds, beat an IDE operation using 30% for 50 seconds (in this case 3 seconds total time saved)
Re:This is just marketing (Score:2)
Only when we started getting into designs like the K7 and Core processors did the speed become excessive. A 2Ghz K7 core was way more than capable of playing mp3s or video files while not killing the box.
Tom
Re:This is just marketing (Score:2)
A: Companies started caring more, since they have to pay twice the electricity (once in their server farm, and again to remove the resulting heat).
B: Portable computing keeps gaining in popularity, including cellphones and PDA type devices. Less power demand increases battery life and reduces weight.
C: CPU's just started getting so hot that more and more elaborate measures were needed to cool them. Reducing power d
Re:This is just marketing (Score:2)
1. When the TDP of chips hit over about 80W, heat became a problem and frequency scaling helped abate that.
2. The tech is needed in laptop chips. With the exception of the Pentium 4, all notebook chips are just modified desktop parts. So the manufacturers made one core and saw no reason to disable that feature, especially since #1 is true.
So frequency scaling just piggybacked its way onto the desktop. What's good for the goos
Samples of same chip vary (Score:2)
Bad Analogy (Score:2)
Sure, both performance and power usage of a chip will vary depending on what you do with it, so any simple one-number power:performance measure won't tell you much useful. Of course, the same thing is true with cars, too; both gas mileage and other aspects of performance (including whether it will go where you
assuming PPW is accurate... (Score:2)
Re:assuming PPW is accurate... (Score:2)
Re:assuming PPW is accurate... (Score:2)
The Solution (Score:1)
A standardised code segment with broad instruction type usage and a long time to complete(to minimize differences/errors). If everybody is using the same reference instruction sequence on all processors for the same amount of time, no debate ensues. Right?
Re:The Solution (Score:2, Insightful)
Am I missing somthing? (Score:1)
The foundation's already there, in water cooling systems. A rudimentary system could be built by dropping a thermometer in the reservoir, and turning off the radiator at the beginning of each test.
Granted, you're only measuring waste heat, but how much power does a CPU pump through data busses?
Power capacity of server rooms (Score:1)
The first real problem is that blade servers are so small now, but require so much power, that companies can easily fit way more compute power in a server room than can be reasonably cooled. So they need more power-efficient servers to use their server space effectively.
And the problem isn't that power can't be measured--it can be measured just as easily as performance. Which is the problem hinted at in the article--firms focusing on the positive results they hav
Use a battery (Score:2)
Re:Power can't be measured? Yeah right. (Score:1)
As per my other post under this article the author is confusing issues of measurment with issues of prediction.
KFG
power companies (Score:1)