Intel has $55.9B record year, ships 46M tablets

[Abwx]

Intel partners with Asus & Lenovo, AMD partners with BungBunggame.

"Death by BungBunggame"

Like intel partnering with some obscure chineses firms.?.

Comsumption delta is up to 10W, this say that the Z3740 consume as much as 7-8W when fully loaded, when gaming for instance...

http://www.notebookcheck.net/Review-Asus-Transformer-Book-T100TA-C1-GR-Convertible.106219.0.html

I noticed that Notebookcheck did surely receive a check...

The T100 consume more than what is displayed on this site, they say that max power drain is 11.8W and that the adapter is 20W but this latter is obviously a 10W adapter with 18% margin, it manage to deliver 11.8W to the T100 which pump eventualy the rest of what it need for the benches from the battery...

 

[dahorns]

Like intel partnering with some obscure chineses firms.?.



I noticed that Notebookcheck did surely receive a check...

The T100 consume more than what is displayed on this site, they say that max power drain is 11.8W and that the adapter is 20W but this latter is obviously a 10W adapter with 18% margin, it manage to deliver 11.8W to the T100 which pump eventualy the rest of what it need for the benches from the battery...

I have no idea what this sentence/paragraph means. Also, commas =/ periods.

Edit: Please do not construe this post as an attack. I know English isn't the first language for many on this site.

 

[Abwx]

I have no idea what this sentence/paragraph means. Also, commas =/ periods.

Edit: Please do not construe this post as an attack. I know English isn't the first language for many on this iste.

The review graphs states 11.8W max power comsumption for the device and the reviewer say that the power adaptator has 20W max power output, yet it is clearly displayed on the adaptator that it s a 10W only model.

When the reviewer test power drain he measure it at the main and the adaptator will not supply more than what it s designed for, in this case 10W plus a margin, since they measure 11.8W we can conclude that the 10W adaptator is capable of supplying 11.8W minus about 10% losses, that s 11W or so.

But this is certainly not the comsumption of the Transformer 100, only what the adaptator can provide it, if it cant supply all the required power the device will extract the needed difference in the battery, probably what happened when they did their tests.

As such their max power comsumption is totaly irrelevant and likely well below the real value.

 

[frozentundra123456]

So if you cant in fact measure the power consumption (for the sake of argument I am accepting your contention), how do you know it is more than 11.8 watts? Use circular logic much?

 

[mrmt]

So if you cant in fact measure the power consumption (for the sake of argument I am accepting your contention), how do you know it is more than 11.8 watts? Use circular logic much?

Why nobody in the industry is calling Intel's bluff?

 

[dahorns]

As such their max power comsumption is totaly irrelevant and likely well below the real value.

That's a bit of a leap. Those values are consistent with their reviews of other Bay Trail devices. Granted, the chargers all appeared to be similarly limited, but at least one device came in under 10W max.

http://www.notebookcheck.net/Review-Asus-VivoTab-Note-8-M80TA-Tablet.115101.0.html

And, here (http://www.notebookcheck.net/Review-HP-Omni-10-5600eg-F4W59EA-Tablet.108702.0.html) is a 10 inch Z3770 that draws 12.8W with a 18W charger. The 1W higher value is likely a product of other components of the device. This can be seen by the ~1W higher idle. Also note, the device has notably less battery life than competing Bay Trail products. Again, suggestive of components other than the processor causing the higher draw.

 

[Abwx]

So if you cant in fact measure the power consumption (for the sake of argument I am accepting your contention), how do you know it is more than 11.8 watts? Use circular logic much?

Because the power adaptator was obviously overloaded, at 11.8W at the main its DC output is providing 90% of this amount, that s 10.6W, the power adaptator protection will limit the output power at this level if drain is constant.

And why is it pumping more.? Because the case reach 45°C, wich point to quite an amount of power dissipated, indeed the 10.1W delta translate to about 8W at the full SoC level, and that s assuming that the battery is not filling the gaps, we are far from the official 4-5W TDP figure, as is the case with about all Intel mobile wich are more often than not out of spec TDP wise, now wonder that it didnt throttle during the tests...

 

[coercitiv]

Why nobody in the industry is calling Intel's bluff?

Nobody in the industry called SDP a bluff?

In other news the tablet just arrived, will get to play with it a bit and see how the chip fares.

 

[Abwx]

That's a bit of a leap. Those values are consistent with their reviews of other Bay Trail devices. Granted, the chargers all appeared to be similarly limited, but at least one device came in under 10W max.

http://www.notebookcheck.net/Review-Asus-VivoTab-Note-8-M80TA-Tablet.115101.0.html

Let s start with this asus vivotab :

Because of the frugal components and the highly effective CPU/GPU combination, the power consumption is always extremely low. The included power adapter supplies a maximum of 2 A at 5 V and thus supplies 10 Watts, which is just sufficient.

Same trick, set apart that the adaptator is not as good as the one in the review i checked, this one is capable of 9.7W only at the main, barely 9W at the DC output, yet another purely artificial number...

Notice that it reach 46°C...

And, here (http://www.notebookcheck.net/Review-HP-Omni-10-5600eg-F4W59EA-Tablet.108702.0.html) is a 10 inch Z3770 that draws 12.8W with a 18W charger. The 1W higher value is likely a product of other components of the device. This can be seen by the ~1W higher idle. Also note, the device has notably less battery life than competing Bay Trail products. Again, suggestive of components other than the processor causing the higher draw.

This one is throttled , fortunately since the case reach 55°C :

At this point, throttling kicks in, and the clock speed drops to 1,088 MHz

But for the performances :

This throttling did not occur in the previous tests, which is why the Omni 10 scores so well in the CPU benchmarks.

Because when the tests starts, at about full frequency :

The test model briefly sucks in 15.1 Watts, but it quickly goes down to 12.8 W again.

Also all thoses Baytrails are supposedly 4W while even thoses doctored power measurements point to twice, if not more, than this value :

http://www.notebookcheck.net/Intel-Atom-Z3740-Tablet-SoC.101427.0.html

 

[dahorns]

Also all thoses Baytrails are supposedly 4W while even thoses doctored power measurements point to twice, if not more, than this value

My understanding, lay as it may be, is that the thermal dissipation of a chip isn't the same as the amount of energy the chip consumes. Presumably, only a portion of the energy used by the chip is converted to heat. The rest hopefully goes to doing all that switching that we want our transistors to do. Also, isn't it expected that Intel chips will exceed TDP in short bursts when the thermal headroom is available? I mean, that is a design feature, not a flaw. Finally, their numbers come from running a stress-test. TDP isn't (usually) based on these type of unrealistic work scenarios.

 

[ShintaiDK]

My understanding, lay as it may be, is that the thermal dissipation of a chip isn't the same as the amount of energy the chip consumes. Presumably, only a portion of the energy used by the chip is converted to heat. The rest hopefully goes to doing all that switching that we want our transistors to do. Also, isn't it expected that Intel chips will exceed TDP in short bursts when the thermal headroom is available? I mean, that is a design feature, not a flaw. Finally, their numbers come from running a stress-test. TDP isn't (usually) based on these type of unrealistic work scenarios.

Essentially 100% of the energy is turned into heat in chip. Reality is like 99.999something% and the rest in radiation.

 

[III-V]

My understanding, lay as it may be, is that the thermal dissipation of a chip isn't the same as the amount of energy the chip consumes. Presumably, only a portion of the energy used by the chip is converted to heat. The rest hopefully goes to doing all that switching that we want our transistors to do. Also, isn't it expected that Intel chips will exceed TDP in short bursts when the thermal headroom is available? I mean, that is a design feature, not a flaw. Finally, their numbers come from running a stress-test. TDP isn't (usually) based on these type of unrealistic work scenarios.

Inefficiency in semiconductors is only exhibited as heat. It's just electrical resistance. If we built them out of superconductors, implementation issues aside, they would consume zero power.

Essentially 100% of the energy is turned into heat in chip. Reality is like 99.999something% and the rest in radiation.

Radiation just slams into other matter and gets "absorbed" as heat, same with light. So it's all heat in the end.

 

[witeken]

If we built them out of superconductors, implementation issues aside, they would consume zero power.

I'm not going to search for a source now, but I doubt that would be the case. The whole chip can't be superconducted, IIRC, at least not in current implementations. One of the biggest benefits of superconductors would be its very high speed.

If anyone wants to do some digging themselves:

Superconductivity: http://spectrum.ieee.org/semiconductors/design/superconductor-logic-goes-lowpower
http://spectrum.ieee.org/biomedical/imaging/superconductivitys-first-century
http://spectrum.ieee.org/semiconductors/design/superconductor-ics-the-100ghz-second-generation

 

[sckmcck]

Inefficiency in semiconductors is only exhibited as heat. It's just electrical resistance. If we built them out of superconductors, implementation issues aside, they would consume zero power.

That's not true. All CPUs today are irreversible so by virtue of thermodynamics, they must consume power to computer.

EDIT: Some links for perusal

http://en.wikipedia.org/wiki/Landauer's_principle
http://en.wikipedia.org/wiki/Reversible_computing

 

[Phynaz]

The ADF continues to threadcrap in a thread about Intel's stellar corporate performance. I wonder if they will tolerate that this afternoon in the AMD thread?

Well, that one will be about AMD's terrible corporate performance.

 

[elemein]

Let s start with this asus vivotab :


Same trick, set apart that the adaptator is not as good as the one in the review i checked, this one is capable of 9.7W only at the main, barely 9W at the DC output, yet another purely artificial number...

Notice that it reach 46°C...




This one is throttled , fortunately since the case reach 55°C :

But for the performances :

Because when the tests starts, at about full frequency :

Also all thoses Baytrails are supposedly 4W while even thoses doctored power measurements point to twice, if not more, than this value :

http://www.notebookcheck.net/Intel-Atom-Z3740-Tablet-SoC.101427.0.html

My question would be; isnt the power draw figure for the whole platform, not just the chip? Why blame the consumption of the whole platform (screen, speakers, SoC, memory, etc.) on just the SoC?

Or am I wrong and they're measuring just the chip as it runs?

 

[TreVader]

Inefficiency in semiconductors is only exhibited as heat. It's just electrical resistance. If we built them out of superconductors, implementation issues aside, they would consume zero power.
Radiation just slams into other matter and gets "absorbed" as heat, same with light. So it's all heat in the end.

You can't build superconducting semiconductors, it's not possible. Superconductors lose nearly nothing to electrical resistance but by definition will only provide that as a benefit while in a superconducting state. A quench is when a superconductor becomes a semiconductor/insulator and that will cause massive heat. So, to my understanding of physics and chemistry, it's not possible.

I'm sure others here are more knowledgable and can expand on this.

 

[witeken]

You can't build superconducting semiconductors, it's not possible. Superconductors lose nearly nothing to electrical resistance but by definition will only provide that as a benefit while in a superconducting state. A quench is when a superconductor becomes a semiconductor/insulator and that will cause massive heat. So, to my understanding of physics and chemistry, it's not possible.

I'm sure others here are more knowledgable and can expand on this.

You can start by reading my links, it's already been done:

Researchers have demonstrated simple digital frequency dividers with Josephson junctions that have 0.25-µm minimum features and that operate at data rates up to 770 Gb/s. They rely on rapid single flux quantum (RSFQ) circuits, whose speed grows as junction sizes shrink. At 0.25 µm, the junctions are intrinsically nonhysteretic, a good sign for the future of complex chips operating at clock frequencies of 100 GHz.

 

[TreVader]

You can start by reading my links, it's already been done:

This stuff is decades away! I stand corrected about the superconducting transistor. However, this paper is talking about creating a handful of transistors total, not even a basic analog to a CMOS circuit. That's just not viable in the near future.

Science is full of these prophetic announcements. Intel was talking about 10Ghz Pentiums 12 years ago. Here we are in 2015 and can't even reach half that.

Also like to mention that in my cursory glance in appears only the interconnects are superconducting and the gates themselves get switched on and off between superconducting and semiconducting states. That, or they stay semiconducting and just the interconnect is lossless.

 

[III-V]

I'm not going to search for a source now, but I doubt that would be the case. The whole chip can't be superconducted, IIRC, at least not in current implementations. One of the biggest benefits of superconductors would be its very high speed.

If anyone wants to do some digging themselves:

Superconductivity: http://spectrum.ieee.org/semiconductors/design/superconductor-logic-goes-lowpower
http://spectrum.ieee.org/biomedical/imaging/superconductivitys-first-century
http://spectrum.ieee.org/semiconductors/design/superconductor-ics-the-100ghz-second-generation

"Implementation issues aside"

Good grief guys, that disclaimer was there for a reason.

 

[sckmcck]

Ok, let's say that we're looking at a Haswell i7 operating at approximately 3.5GHz. This chip has approximately 1.4 billion transistors. Let's just say that for the purposes of this thought experiment, at each clock cycle, 1% of those transistors switch and the chip is running at 50C.

So, using Landauer's principle:
3.5x10^6 * 1.4x10^9 * .01 *10^-23 * 323 * Log[2] = 1.5x10^-7 J / s

So that's approximately 150 nanowatts that's consumed by this processor to switch 1% of it's transistors per clock cycle. This is completely independent of interconnects or anything else. This is a purely entropy consideration.

 

[TreVader]

Ok, let's say that we're looking at a Haswell i7 operating at approximately 3.5GHz. This chip has approximately 1.4 billion transistors. Let's just say that for the purposes of this thought experiment, at each clock cycle, 1% of those transistors switch and the chip is running at 50C.

So, using Landauer's principle:
3.5x10^6 * 1.4x10^9 * .01 *10^-23 * 323 * Log[2] = 1.5x10^-7 J / s

So that's approximately 150 nanowatts that's consumed by this processor to switch 1% of it's transistors per clock cycle. This is completely independent of interconnects or anything else. This is a purely entropy consideration.

Lol 2005 and 2 posts. Talk about a lurker.

 

[elemein]

Lol 2005 and 2 posts. Talk about a lurker.

I don't know why that is relevant to his post, which seems pretty knowledgeable (though questionable due to lack of voltage inclusion.)

 

[Excessi0n]

I don't know why that is relevant to his post, which seems pretty knowledgeable (though questionable due to lack of voltage inclusion.)

Voltage doesn't matter. What he is calculating is the amount of energy required to affect the change in entropy involved in non-reversible computation in an environment of that temperature. Actually reaching that number would require everything in your computer to be literally perfect.

A superconducting computer could get a whole hell of a lot closer to Landauer's Limit than a semiconducting one could, but it wouldn't be perfect. Superconducting transistors do require a non-superconducting or weakly-superconducting component, so there will be very small losses due to resistance. On the flip side, the limit would be much lower for a superconducting computer since they are, by necessity, incredibly cold.

 

[Abwx]

Ok, let's say that we're looking at a Haswell i7 operating at approximately 3.5GHz. This chip has approximately 1.4 billion transistors. Let's just say that for the purposes of this thought experiment, at each clock cycle, 1% of those transistors switch and the chip is running at 50C.

So, using Landauer's principle:
3.5x10^6 * 1.4x10^9 * .01 *10^-23 * 323 * Log[2] = 1.5x10^-7 J / s

So that's approximately 150 nanowatts that's consumed by this processor to switch 1% of it's transistors per clock cycle. This is completely independent of interconnects or anything else. This is a purely entropy consideration.

Shouldnt the first factors be 3.5 x 10^9..?.