Xbit Labs: AMD-> Improvements of Next-Generation Process Technologies Start to Wane.

[CLite]

Relative power consumption has greatly increased, as now the computing part is more and more power-conserving, so the percentages are increasing. Also the absolute power has been increasing, as we are further moving toward a RISC CPU which reads every year bigger CISC ops.

You are perhaps misunderstanding me. Relative power consumption of course greatly increases as you scale down a design using the same technology. However, what I'm saying if you start with a small design and then improve it going from 32nm to 22nm you will most likely pack some more computing power into it which decreases the relative consumption of the decoder. As more people want their phones to run videos and pack more computing power then the impact of the decoder will decrease.

I have no real horse in this race, but I think discounting Intel's financial capital and desire to produce 14nm atom's in 3 years is a mistake. They have an ARM license but they are obviously more interested in cornering the market with the more exclusive x86. It of course has downsides, but for you to assume that you have more wisdom and foresight than the entire company of Intel paints you as naive and arrogant.

 

[ncalipari]

Not try and use it to draw comparisons between architectures. Again, without knowing how the systems were set up it's hard to draw conclusions, but if you were measuring a maximum of 8W draw on your Origen before the PS, I'm guessing you probably didn't have an external hard drive attached through the USB interface. Did the E350 system have a HDD? More than 1GB RAM?

no they were both using SD CARD, and the same ram amount. The radio part of ARM boards is just much more optimized,.

It would be easy enough to show a comparison where an x86 system and ARM system are within 100% of each other for power but the x86 blows the performance of the ARM system out of the water, especially when buying COTS designs. That doesn't mean that x86 dominates ARM either.

please do. I would be very curious to see it. Please find an ARM system in the 1-10W range that get a lower per watt performance than an x86 cpu within a 20W envelope. I would love to see that.

 

[ncalipari]

You are perhaps misunderstanding me. Relative power consumption of course greatly increases as you scale down a design using the same technology. However, what I'm saying if you start with a small design and then improve it going from 32nm to 22nm you will most likely pack some more computing power into it which decreases the relative consumption of the decoder. As more people want their phones to run videos and pack more computing power then the impact of the decoder will decrease.

People first of all want long battery life. I couldn't care less about a 2500k in a smartphone, if my battery weights 1 kg and last 1 hour.

Then we have computational power.

Anyhow consider that lately the computational power has not been increasing much. The goal is to stack the same computational power in a smaller power envelope, so the decoder is again eating more and more of the available TDP.

I have no real horse in this race, but I think discounting Intel's financial capital and desire to produce 14nm atom's in 3 years is a mistake. They have an ARM license but they are obviously more interested in cornering the market with the more exclusive x86. It of course has downsides, but for you to assume that you have more wisdom and foresight than the entire company of Intel paints you as naive and arrogant.

It's not a question of wisdom and foresight. It's a question of physics. The theoretical lower limit of the power needed by a decoder is VERY HIGH.

Either you defeat physics law, or drop x86.

 

[MrTeal]

no they were both using SD CARD, and the same ram amount. The radio part of ARM boards is just much more optimized,.



please do. I would be very curious to see it. Please find an ARM system in the 1-10W range that get a lower per watt performance than an x86 cpu within a 20W envelope. I would love to see that.

Sure. To keep it within the realm of parts mentioned, that MSI E350 board vs the Origen. The MSI with a proper low power PSU would run around 20W. The application benchmark is USB copy speed. I don't have the systems available to test, but I'm going to go out on a limb and say that with two good USB 3.0 sticks you're going to be several times faster on the X86 platform.

What does that have to do with ARM vs X86? Nothing. It's not much less valid than your comparison though, since you aren't willing to share the details on your test other than to say that in java int performance x86 gives twice the performance of ARM at 3x the cost and 10x the power.

 

[Fox5]

for a customer I made a benchamark of AMD e350 vs Origen, based on java integer performance.


Samsung Origen: 1142

Power consumption Max: 8 Watt

Idle: 2 Watt

Price: 190 Dollars


Amd e350: 2137

Power Consumption MAX: 57 Watt

Idle: 12 Watt

Price: 600 dollars

Amd e350 is not the latest tech available, but is much faster than any atom. Performance is within the same order of magnitude, while power consumption both is one order of magnitude greater.

For that very application, and with that synthetic benchmark, ARM is much better. If you are plugged to the power, have an unlimited amount of cash and don't care about heat then x86 is better. Anyhow the smartphone world is much more similar to my use case than the latter.

x86 simply cannot play in the 1-10 Watt league.

The E-350 is not AMD's lowest power processor. That would be the Z-01, which draws ~5W max, so total platform power should be under 10W. However, it's only 1Ghz, to the E-350's 1.6Ghz. Assuming that benchmark scales linearly with clockspeed, the 1Ghz -z01 would score about 1336 on that benchmark, so still faster than the samsung origen by a bit. Intel has atoms that do a bit better in power consumption.
I have a 10" tablet with the z-01 processor, and average power consumption of the entire system is under 10W.

IMO, the 2W-10W battle field (ie, tablets and ultraportables) is one that x86 is fully capable of conquering, the challenge will be doing it at a lower cost than ARM. Getting competitive in cell phones (where 2W needs to be roughly the max power consumption of the SOC, and average needs to be around half a watt) is going to be much tougher.

Intel's ULV platforms are going to own the sub 10W market in performance. Trigate plus spending transistors on power gating is going to create an unmatched level of performance/power consumption. Heck, Intel is already unmatched as is, they're only besting themselves.
But the cost of an ULV chip is probably at least 50% more than it's ARM competition, even if power is vastly in its favor. Atom and Bobcat exist not because of power efficiency, but because of low cost of production. They're comparable in cost to ARM solutions. They may be able to beat ARM in power efficiency as well, but it's close, Atom and Bobcat have the lead at current power levels, but the A15 core is putting more focus on performance. Current ARM designs are optimized for absolute low power usage first and foremost, which means they probably left a bit of efficiency on the design table in order to hit lower power targets for phones. The very profitable smart phone and tablet market is also increasing demand for high end, pricier ARM socs, so ARM can start throwing transistors at their designs to increase performance if need be.

The theoretical lower limit of the power needed by a decoder is VERY HIGH.

It's high for a phone platform, but once you start measuring power in watts, it's not that high. Tom's hardware did a test years back with the Athlon XP Mobile chips, a chip with barely any power management features, and they got the chip down to about 3W power consumption at 300Mhz. Now at about the same power consumption, AMD can put out a 1Ghz dual core chip with graphics on it.
ARM cpus use a decoder too, btw.

 

[Idontcare]

People who defend your point of view usually have a poor understanding of the history of x86, and its architecture. Or, like Idontcare, refuse to accept reality.

Excuse me, did I crap in your wheaties or something?

I have no idea what you are going on about in terms of how it relates to me specifically but member callouts do violate the posting guidelines.

 

[ncalipari]

Sure. To keep it within the realm of parts mentioned, that MSI E350 board vs the Origen. The MSI with a proper low power PSU would run around 20W. The application benchmark is USB copy speed. I don't have the systems available to test, but I'm going to go out on a limb and say that with two good USB 3.0 sticks you're going to be several times faster on the X86 platform.

from 55 watt to 20 watt only with PSU? LOL.

What does that have to do with ARM vs X86? Nothing. It's not much less valid than your comparison though, since you aren't willing to share the details on your test other than to say that in java int performance x86 gives twice the performance of ARM at 3x the cost and 10x the power.

You can very well run the java benchmark yourself. I'm not sharing the details, but java integer benchmark are available to everybody.

 

[ncalipari]

Excuse me, did I crap in your wheaties or something?

I have no idea what you are going on about in terms of how it relates to me specifically but member callouts do violate the posting guidelines.

well Idontcare when I signaled you much more evident violations you just laughed and or ignored, so I thought that saying that I disagree with you wasn't such a serious violation/.


By the way I disagree with your opinions, as I think they are very simple minded, and I disagree that a public figure comes out to call other people idea "crap".

If you say that I can't disagree with you, I will immediately stop that.

 

[ncalipari]

The E-350 is not AMD's lowest power processor. That would be the Z-01, which draws ~5W max, so total platform power should be under 10W. However, it's only 1Ghz, to the E-350's 1.6Ghz. Assuming that benchmark scales linearly with clockspeed, the 1Ghz -z01 would score about 1336 on that benchmark, so still faster than the samsung origen by a bit. Intel has atoms that do a bit better in power consumption.
I have a 10" tablet with the z-01 processor, and average power consumption of the entire system is under 10W.

well sorry to notify you that power consumption is not linear with frequency.

IMO, the 2W-10W battle field (ie, tablets and ultraportables) is one that x86 is fully capable of conquering, the challenge will be doing it at a lower cost than ARM. Getting competitive in cell phones (where 2W needs to be roughly the max power consumption of the SOC, and average needs to be around half a watt) is going to be much tougher.

that's simply not possible

Intel's ULV platforms are going to own the sub 10W market in performance

only for the CPU. You're forgetting GPU, radio and HW acceleration. Plus motherboard, memory, ....

It's high for a phone platform, but once you start measuring power in watts, it's not that high. Tom's hardware did a test years back with the Athlon XP Mobile chips, a chip with barely any power management features, and they got the chip down to about 3W power consumption at 300Mhz. Now at about the same power consumption, AMD can put out a 1Ghz dual core chip with graphics on it.
ARM cpus use a decoder too, btw.

Yes but a 300mhz chip is not 1/3 as fast as a 1ghz chip. Again, you fail in the hypothesis of linear scaling.

 

[Idontcare]

well Idontcare when I signaled you much more evident violations you just laughed and or ignored, so I thought that saying that I disagree with you wasn't such a serious violation/.


By the way I disagree with your opinions, as I think they are very simple minded, and I disagree that a public figure comes out to call other people idea "crap".

If you say that I can't disagree with you, I will immediately stop that.

I seriously have no idea what you are going on about here.

Are you feeling OK?

 

[Fox5]

Sure. To keep it within the realm of parts mentioned, that MSI E350 board vs the Origen. The MSI with a proper low power PSU would run around 20W. The application benchmark is USB copy speed. I don't have the systems available to test, but I'm going to go out on a limb and say that with two good USB 3.0 sticks you're going to be several times faster on the X86 platform.

from 55 watt to 20 watt only with PSU? LOL.




You can very well run the java benchmark yourself. I'm not sharing the details, but java integer benchmark are available to everybody.

MSI Windpad 110W with AMD Z-01
http://www.techspot.com/review/441-msi-windpad-110w-windows-8/page6.html

The system idles at 8W and uses 14W while stressed, but that's powering the processor, hard drive, 10" IPS LCD screen, and so on. And it'll outperform just about any ARM SOC out there, and the Z-01 has similar costs to the high end ARM processors.

For comparison, the ipad 2 (which probably has similar hardware costs all around) idles at about 3W, but draws up to 10W under load. Still, at least x86 is in the same ballpark as the competition, higher power consumption, similar cost, higher performance.

 

[ncalipari]

The system idles at 8W and uses 14W while stressed, but that's powering the processor, hard drive, 10" IPS LCD screen, and so on.

It has a much lower clock than the e350, so I really doubt that it can come any close to the e350 platform.

And it'll outperform just about any ARM SOC out there, and the Z-01 has similar costs to the high end ARM processors.

I really doubt that.

 

[Fox5]

It has a much lower clock than the e350, so I really doubt that it can come any close to the e350 platform.




I really doubt that.

It is 62.5% the speed of the e-350. 62.5% of the e-350's benchmark score (that you posted) is still higher than the ARM benchmark score you posted.

I've run all the major browser (javascript and more) benchmarks on my tablet, and it outperforms all the ARM tablets out there, sometimes by as much as 80% depending on the benchmark.

I'll be happy to run any freely available benchmark on it. Remember, it the AMD Z-01 also has more memory bandwidth than any current ARM SOC, so it's not just cpu performance at play. The GPU also has more GFLOPs and fillrate than any mobile GPU, and with much higher transistor density. (GPUs seem to be AMD's strongest point by far, even nvidia's designs aren't as dense, and nvidia beats everyone else)

 

[ncalipari]

It is 62.5% the speed of the e-350. 62.5% of the e-350's benchmark score (that
you posted) is still higher than the ARM benchmark score you posted.

yes, it's faster by less than 10%, but consume 400% more power in idle, and 100% more power at full load. a deal breaker I guess.

I've run all the major browser (javascript and more) benchmarks on my tablet, and it outperforms all the ARM tablets out there, sometimes by as much as 80% depending on the benchmark.

Care to share which CPU were you testing?

I'll be happy to run any freely available benchmark on it. Remember, it the AMD Z-01 also has more memory bandwidth than any current ARM SOC, so it's not just cpu performance at play. The GPU also has more GFLOPs and fillrate than any mobile GPU, and with much higher transistor density. (GPUs seem to be AMD's strongest point by far, even nvidia's designs aren't as dense, and nvidia beats everyone else)

I see that nowdays most smarphones are CPU-bound.

 

[Fox5]

yes, it's faster by less than 10%, but consume 400% more power in idle, and 100% more power at full load. a deal breaker I guess.




Care to share which CPU were you testing?



I see that nowdays most smarphones are CPU-bound.

1. Up to 80% faster, depending on the benchmark.
2. 250% more power consumption in idle. But it's also Windows versus iOS, 4GB of high speed ram, and the AMD cpu does not have as advanced power management capabilities as current low power ARM chips. It's gen 1, gen 2 should close the gap. Less than 200% greater power consumption at load, for up to almost 200% greater performance at times.
3. It's the AMD Z-01, the power optimized version of AMD's C-50 APU. It's a netbook processor shoe-horned into a tablet market. Next year will bring real tablet optimized processors from AMD and Intel. Will they get the battery life of ARM? Probably not, but to play in the tablet market form factor, but ARM will have to scale power usage up to match the performance of x86, we haven't reached the crossover point yet where ARM outperforms x86 at lower power consumption.
4. The tablet has Android 4.0.1 ported to it as well, so it may be possible to run some like-for-like Android benchmarks on it. (although windows is probably better optimized for x86 than android is at this point: case in point, on windows the AMD gpu has beastly h.264 decode, on android it's not accelerated)

Edit: I got the cpu benchmark test of Antutu to run. In that, the dual core AMD chip scores about as well as the quad core Tegra 3, and both beat everything else by a landslide.

 

[ncalipari]

1. Up to 80% faster, depending on the benchmark.

again: on which cpu?

2. 250% more power consumption in idle.

2 Watt vs 12 Watt is a 500% increase

But it's also Windows versus iOS, 4GB of high speed ram, and the AMD cpu does not have as advanced power management capabilities as current low power ARM chips.

you're right on the last part. x86 lacks the power gating ability of arm. Both due to lack of investment, and due to inherent shortcomings of x86

It's gen 1, gen 2 should close the gap. Less than 200% greater power consumption at load, for up to almost 200% greater performance at times.

I think you should either work more on your fraction skills, or reconsider your expectations about intel

3. It's the AMD Z-01, the power optimized version of AMD's C-50 APU. It's a netbook processor shoe-horned into a tablet market. Next year will bring real tablet optimized processors from AMD and Intel. Will they get the battery life of ARM? Probably not, but to play in the tablet market form factor, but ARM will have to scale power usage up to match the performance of x86, we haven't reached the crossover point yet where ARM outperforms x86 at lower power consumption.

market shown us that people prefer battery over performance

 

[CLite]

It's not a question of wisdom and foresight. It's a question of physics. The theoretical lower limit of the power needed by a decoder is VERY HIGH.

Either you defeat physics law, or drop x86.

I don't know the absolute theoretical limits of the x86 decoder as compared to ARM decoders, beyond the fact that it is less efficient. However, I'm going to guess a company as advanced as Intel has a roadmap for the next 5-10 years that does not involve violating physical laws.

If you, ncalipari, know for a fact that they'd have to violate physical laws to compete with ARM then I'd highly suggest shorting the hell out of their stock and take advantage of your superior knowledge.

 

[ncalipari]

I don't know the absolute theoretical limits of the x86 decoder as compared to ARM decoders, beyond the fact that it is less efficient.

ARM doesn't have a decoder. At least not in the x86 sense. That's why it is so efficient.

However, I'm going to guess a company as advanced as Intel has a roadmap for the next 5-10 years that does not involve violating physical laws.

yet doesn't guarantee to have more of 5% of the most lucrative market at the moment: smartphones and tabled

If you, ncalipari, know for a fact that they'd have to violate physical laws to compete with ARM then I'd highly suggest shorting the hell out of their stock and take advantage of your superior knowledge.

I prefer not to invest in stocks. I prefer to sell my superior knowledge to paying companies.

I'll make another example: you can invest as much as you want in vacuum tubes, but transistor are still going to to rock them.

 

[MrTeal]

from 55 watt to 20 watt only with PSU? LOL.

55 Watts from the review you linked. The MSI board was tested here...
http://www.anandtech.com/show/4134/the-brazos-review-amds-e350-supplants-ion-for-miniitx/7
32 Watts while running Cinebench, and from that review...

I don’t have any pico PSUs or anything super efficient readily available so don’t expect any of the numbers to be particularly impressive, but what they are is comparable to one another. I hooked up each one of the systems I’d been using to the same PSU and measured power in three conditions: idle, full CPU load (Cinebench 11.5) and while playing a 1080p H.264 video.

He doesn't list what power supply he's running, but given the text I would assume it's probably a standard desktop model. Even 80Plus gold supplies in the 500W range tend to have terrible efficiency running at 20W output. You'd lose a couple watts on the output by not running 4GB of RAM, and a few watts running the OS off a thumb drive or SD card as opposed to a hard drive. I would be surprised if you couldn't easily run 100% CPU utilization within 20W input power with a properly specced and efficient 50W supply. You probably could push the power higher during gaming with the GPU fully utilized, but I don't see why you'd need the GPU running during USB copying.

You can very well run the java benchmark yourself. I'm not sharing the details, but java integer benchmark are available to everybody.

It's not the java benchmarks that are my issue of contention. It's the unverifiable claims regarding price and power you made. I can buy an E350 based ZBox AD10 for $315 that is a fully complete system with RAM and HDD, so when you present a claim that the E350 system if $600 that's obviously because of your implementation of it. Likewise, the power draw claims are meaningless: The MSI Wind U270 is E350 based @ 1.6GHz and ships with a 40W supply. I doubt that if you're running a game while plugged in the battery is dying, so if you're drawing more than that it's implementation.

If you want to compare x86 to ARM
1) Don't do it with an APU.
2) If you want to compare the cost, look at the actual chip or chipset, not a completed system.
3) If you want to look at power, don't compare the draws of the complete system.

If I came here and said that Llano is way more efficient than Nehalem because my Llano system idles at 43W, cost $700 and scores 75% as well in Cinebench as an i7 system that cost $2500 and idles at 200W it would be just as meaningless, especially if I won't disclose that the i7 is running 580 SLI.

 

[firewolfsm]

If anyone wants to get back to the thread topic and what the chairman said...

He has a very valid point that the returns on generational leaps are actually decreasing at the ~20nm level. We don't feel this in the arm market yet because they haven't even released 32nm chips yet.

Silicon transistors are within a few generations of dead, in terms of progress. Going the way of the vacuum tube and magnetic storage. It can't continue into the next decade. People are racing to get graphene transistors and optical interconnects out quickly, but they're still in research phase and this is a problem because they're running out of time.

Do you remember EUV lithography? That was a failed attempt to continue the process of arranging matter from the top down, that is the fundamental problem with lithography. We need matter to arrange itself from the bottom up like carbon or neurons to advance processors.

 

[ncalipari]

55 Watts from the review you linked. The MSI board was tested here...
http://www.anandtech.com/show/4134/the-brazos-review-amds-e350-supplants-ion-for-miniitx/7
32 Watts while running Cinebench, and from that review...

I humbly notice that also my link was from this site.

He doesn't list what power supply he's running, but given the text I would assume it's probably a standard desktop model. Even 80Plus gold supplies in the 500W range tend to have terrible efficiency running at 20W output.

Yes he does:

http://www.anandtech.com/show/4996/asus-e35m1m-pro-review-anyone-for-fusion/4

Jeantech 250 W

You'd lose a couple watts on the output by not running 4GB of RAM,

sorry?

and a few watts running the OS off a thumb drive or SD card as opposed to a hard drive.

Actually, since there are no parts moving at 70 mph, SD and SSD consume a lot less power.

It's not the java benchmarks that are my issue of contention. It's the unverifiable claims regarding price and power you made. I can buy an E350 based ZBox AD10 for $315 that is a fully complete system with RAM and HDD, so when you present a claim that the E350 system if $600 that's obviously because of your implementation of it.

6 months ago, the price on brazos were higher, now it's not the latest CPU of its kind. Moreover price in europe are higher.

Likewise, the power draw claims are meaningless: The MSI Wind U270 is E350 based @ 1.6GHz and ships with a 40W supply. I doubt that if you're running a game while plugged in the battery is dying, so if you're drawing more than that it's implementation.

I can bet they also use some power saving features to limit the CPU and/or system

If you want to compare x86 to ARM
1) Don't do it with an APU.

Why? also ARM has gpu integrated in the CPU.

2) If you want to compare the cost, look at the actual chip or chipset, not a completed system.

why? you get the system for free?

3) If you want to look at power, don't compare the draws of the complete system.

have you been able to run the CPU from battery and the rest from socket while on the go? sounds interesting

If I came here and said that Llano is way more efficient than Nehalem because my Llano system idles at 43W, cost $700 and scores 75% as well in Cinebench as an i7 system that cost $2500 and idles at 200W it would be just as meaningless, especially if I won't disclose that the i7 is running 580 SLI.

I'm sorry but I can't see your point.

 

[Blue Shift]

I'm sorry but I can't see your point.

I believe his point was that setting up a straw-man argument doesn't really help anyone to understand the issue at hand, or even to understand your point of view and how you reached it. Just in case you didn't notice, this was also the point of various other posts on this page.

Sorry, I really don't mean to offend, but... Could you possibly post a different example, one that's less contrived, and use numbers from that instead? For a start, I'd recommend comparing specific (please name them) devices of the same form factor, since screens are rather power-hungry.

Once you come up with new numbers, could you help explain to us why these current performance/power necessarily imply inherent flaws in necessary commonalities of x86 CPUs? I can't really follow some of your leaps:

you're right on the last part. x86 lacks the power gating ability of arm. Both due to lack of investment, and due to inherent shortcomings of x86

It's not a question of wisdom and foresight. It's a question of physics. The theoretical lower limit of the power needed by a decoder is VERY HIGH. Either you defeat physics law, or drop x86.

In case you don't realize why these statements seem confusing to the rest of us, here are some questions they raise:

Power gating is a feature that's also used in some x86 CPUs. Is the power gating seen in those cases inherintly inferior to what's seen in ARM CPUs? If so, in what ways is it inferior and why is this the case? Why can't implementation seen in ARM CPUs be adapted for use in x86 CPUs?

You've said that the theoretical lower limit for decoder power consumption is "VERY HIGH". This implies that you know an actual value for this lower limit; what is it? What is this theoretical lower limit that is so cripplingly large that power-efficient x86 GPUs are impossible to make? And how was this number reached?

 

[MrTeal]

I humbly notice that also my link was from this site.
Yes he does:

http://www.anandtech.com/show/4996/asus-e35m1m-pro-review-anyone-for-fusion/4

Jeantech 250 W

Did you actually click the link I provided? The one you just linked was a review for a different board, by a different author, 9 months apart. You might assume it's the same power supply, I'm not so confident.

sorry?

RAM also consumes power. If you go from the 4GB as tested to 1GB, you'll save a couple watts.

Actually, since there are no parts moving at 70 mph, SD and SSD consume a lot less power.

Yes, that was the point. I was exclusively referring to the test that I linked, where if you replaced the HDD used in the test with a flash drive you would further reduce the power.

Why? also ARM has gpu integrated in the CPU.

why? you get the system for free?

have you been able to run the CPU from battery and the rest from socket while on the go? sounds interesting

The GPU on an ARM SoC has nothing to do with x86 vs ARM.
The cost of the system contains the cost of the different CPU, but that's a small part of it.
The power draw of rest of the rest of the system has nothing to do with x86 vs ARM.

 

[ncalipari]

Did you actually click the link I provided? The one you just linked was a review for a different board, by a different author, 9 months apart. You might assume it's the same power supply, I'm not so confident.

I simply showed you that a reputable source showed power figures very similar to mines

RAM also consumes power. If you go from the 4GB as tested to 1GB, you'll save a couple watts.

this sites disagree with you.

http://www.anandtech.com/show/2624

Yes, that was the point. I was exclusively referring to the test that I linked, where if you replaced the HDD used in the test with a flash drive you would further reduce the power.

I used SD card on both systems.

The GPU on an ARM SoC has nothing to do with x86 vs ARM.

Why? can you mount ARM Gpu on x86 and viceversa?

If not so, then GPU is strictly related to CPU platform.

The cost of the system contains the cost of the different CPU, but that's a small part of it.

I think newegg disagree with you. Moreover you cant buy ARM CPU without the board.

The power draw of rest of the rest of the system has nothing to do with x86 vs ARM.

Following this reasoning, we should say that integrating the GPU in the CPU is a bad move.

Another example: if you need wifi, and cpu gives you wifi for free, then you're going to save power. Aren't you?

 

[Phynaz]

this sites disagree with you.
http://www.anandtech.com/show/2624

"RAM power requirements measured a constant 2W per DIMM"

So yes, dropping from 4GB to 1GB would save power as he said. It looks really bad when you get owned by your own links BTW.