Why is AMD Trinity idle power consumption lower than for Intel Ivy Bridge?

[Fjodor2001]

See:

http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/8

Total system idle power consumption:

AMD A8-5600K/A10-5800K: 32 W
Intel 3570/3220: 39.6 W

Isn't that strange? I know Intel 3570 is 4C, but 3220 is 2C/4T so it should be comparable to the AMD CPUs. And the cores should be power gated anyway, so at idle they should consume very little power.

Any explanation to this? Is an AMD Trinity based system simple better than Intel Ivy Bridge at achieving low idle power consumption?

 

[djshortsleeve]

See:

http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/8

Total system idle power consumption:

AMD A8-5600K/A10-5800K: 32 W
Intel 3570/3220: 39.6 W

Isn't that strange? I know Intel 3570 is 4C, but 3220 is 2C/4T so it should be comparable to the AMD CPUs. And the cores should be power gated anyway, so at idle they should consume very little power.

Any explanation to this? Is an AMD Trinity based system simple better than Intel Ivy Bridge at achieving low idle power consumption?

Less power, lower power

 

[lehtv]

Less power, lower power

It's 22nm vs 32nm though

 

[Khato]

Any explanation to this? Is an AMD Trinity based system simple better than Intel Ivy Bridge at achieving low idle power consumption?

Simple - differences in motherboard power consumption. Unfortunately the Intel DZ77GA-70K used in the review isn't present in other Anandtech motherboard reviews, but it's pretty easy to cross-reference and get a basic idea of how it compares. Namely, compare the Anandtech chart which has a range of 69W to 113W idle between the different motherboards - http://www.anandtech.com/show/6170/...747-featuring-gigabyte-asrock-ecs-and-evga/22 - to this one - http://hothardware.com/Reviews/Z77-Motherboard-RoundUp-MSI-ASUS-Gigabyte-Intel/?page=12 - which shows the Intel board used to compare against Trinity being 17 watts above the MSI GD65.

Oh, here's another nice point for comparison - http://www.anandtech.com/show/6360/asus-f2a85v-pro-review-a-look-at-fm2-with-a85x/6 - that shows a similar 31 watts for the Trinity platform, but only 22 watts for an i3-3225... Yeah, makes Trinity look a tad bit less impressive eh?

 

[Exophase]

There are other reviews which show it the other way around. Whole system idle consumption isn't a good thing to compare CPU idle consumption with, but these days it's going to be so low that it'll be a drop in the bucket compared to what it consumes while you're actively using it.

The real purpose of the idle measurements are to give you a rough estimate of how much the CPU alone uses at full load, by removing almost all of the CPU's contribution so you can subtract the two numbers.

 

[Idontcare]

Why is AMD Trinity idle power consumption lower than for Intel Ivy Bridge?

Don't confuse platform power usage with that of an individual chip within the platform.

Those two platforms you referenced do have idle power usage that favors the platform which contain a Trinity chip, but that doesn't mean the Ivy Bridge chip is the reason that specific Intel platform consumes more power at idle.

Don't you remember how awful the power usage of the initial Atom platform was because the chipset itself consumed something like 3X the power that the atom CPU consumed?

 

[know of fence]

Whole system idle consumption isn't a good thing to compare CPU idle consumption with, but these days it's going to be so low that it'll be a drop in the bucket compared to what it consumes while you're actively using it.

Objection. For most people idle consumption is by far the most important, followed by stand-by consumption. PCs spend most time idle or close to idle (internet, video, work), unless use is dominated by certain well known habit-forming 3D games, or distributed computing like F@H keeps the processors busy.
Full load may increase consumption 10x but people may spend 1000 fold the amount of time in idle.

 

[Ferzerp]

Full load may increase consumption 10x but people may spend 1000 fold the amount of time in idle.

Hyperbole doesn't help your point. A used pc is under load far more than a single minute per 16 2/3 hours.

In fact, I'd say that it's far closer to 5:1 or so (or even more), especially on these low powered devices. (The AMD option will spend more time under load than the Intel as it takes longer for it to do the exact some thing). Any other time it would be in hibernate or sleep which are different power states entirely.

 

[Abwx]

In fact, I'd say that it's far closer to 5:1 or so (or even more), especially on these low powered devices.

Have you ever looked at a task manager checking CPU utilisation.??..

(The AMD option will spend more time under load than the Intel as it takes longer for it to do the exact some thing).

That s the typîcal empty argument..

How much longer in what average user task.??.

Browsing ?..Youtubing ?...

Likely that the average working time is meaningfull..

Any other time it would be in hibernate or sleep which are different power states entirely.

If it is not empty enough for you , imagination will push you further...

Once the internet page is loaded it enter hibernation ?..

Or does it do it while i m reading my mailings..?..

Or it exit hibernation just before posting these words
and will return in this state once it s uploaded..?..

 

[Ferzerp]

There is a language barrier here or something. I cannot really divine what you mean by the characters appearing on my screen enough to respond.

Sorry.

 

[Phynaz]

I think what he is saying is you are wrong about the amount of idle time, and race to the finish is a figment of your imagination.

You could link him to Tech Report's task power calculations.

 

[Exophase]

Objection. For most people idle consumption is by far the most important, followed by stand-by consumption. PCs spend most time idle or close to idle (internet, video, work), unless use is dominated by certain well known habit-forming 3D games, or distributed computing like F@H keeps the processors busy.
Full load may increase consumption 10x but people may spend 1000 fold the amount of time in idle.

Full load power consumption for the CPU alone can be 50-100x idle load, not 10x. And if you really keep your computer at idle 99.9% of the time then you aren't using it. Someone who is playing games an hour a day will have it out of idle way more than that.

These days both CPUs are likely using under 2W when idle. But you can see the system idle consumption grossly outstrips this. So yes, it's still a drop in the bucket - if you are so concerned about saving power that you're counting mW turn your computer off.

Just for the curious, here are some cost numbers - let's say you use that full 2W 24/7 - computer always on and always idle. That's about 17520W over the course of a year. In the US that'll cost around $1-3 for the year. If that price concerns you vs the amount you paid on the computer then you're being too frugal. If that amount concerns you then you should start unplugging your TV when it's not on, because chances are it's using more power than your CPU does while idle.

Fighting to optimize idle consumption matters on laptops. It matters on tablets. It especially matters on phones. On desktops it has hit diminishing returns. But I think people still like to bring it up because they're obsessed over trying to make one processor look better than another in any way possible.

 

[LogOver]

No, it's exactly opposite. You will hardly notice 5-6W diff. in your montly bill. But even if your system spends only 0.001% of time under full load, you still need to provide sufficient cooling solution for your cpu. More noisy and bulky for trinity.

BTW, Trinity does not consume less than IB in idle. Idle system power mostly depends on motherboard.

 

[know of fence]

To answer OP's question with a wild, largely unsubstantiated guess. When scaling down the frequency there is a barrier, a minimum voltage. There is little to be gained.
So It must be the IGP, even though Intel has EIST to scale frequency and voltage for the CPU, its integrated graphics runs at a high, fixed frequency all the time (at least on my 32nm Clarkdale CPU). Whereas AMD likely has some power saving circuitry build in and also probably auto adjusts the frequency for Trinity graphics.
This conjecture could be verified by comparing idle results for Trinity + discrete graphics, if the absolute difference is smaller for idle power, then the savings come from integrated graphics.

Hyperbole doesn't help your point. A used pc is under load far more than a single minute per 16 2/3 hours.

In fact, I'd say that it's far closer to 5:1 or so (or even more), especially on these low powered devices.

Full load power consumption for the CPU alone can be 50-100x idle load, not 10x. And if you really keep your computer at idle 99.9% of the time then you aren't using it. Someone who is playing games an hour a day will have it out of idle way more than that.

These days both CPUs are likely using under 2W when idle. But you can see the system idle consumption grossly outstrips this. So yes, it's still a drop in the bucket - if you are so concerned about saving power that you're counting mW turn your computer off.

The numbers linked in the OP are 32 W and 39W. idle; 11x load.

I didn't say 2 W difference is all that important. In the early 2000s all you could hope for were 60+ W idle with some low power office Celeron system. I'm glad we are at about half of that, now.

My statistical claim is this: power(idle)*time(idle) >> power(load)*time(load)

My point about 1000x or a 100000x if you will, was to call attention to the fact that people often underestimate time, which has the potential to factor in much more heavily than almost everything else. Still larger factors/multipliers only come with the number of users, when hundreds of millions of office PCs worldwide save 20-30 W or just 2 W, that's significant.

 

[ShintaiDK]

Simple - differences in motherboard power consumption. Unfortunately the Intel DZ77GA-70K used in the review isn't present in other Anandtech motherboard reviews, but it's pretty easy to cross-reference and get a basic idea of how it compares. Namely, compare the Anandtech chart which has a range of 69W to 113W idle between the different motherboards - http://www.anandtech.com/show/6170/...747-featuring-gigabyte-asrock-ecs-and-evga/22 - to this one - http://hothardware.com/Reviews/Z77-Motherboard-RoundUp-MSI-ASUS-Gigabyte-Intel/?page=12 - which shows the Intel board used to compare against Trinity being 17 watts above the MSI GD65.

Oh, here's another nice point for comparison - http://www.anandtech.com/show/6360/asus-f2a85v-pro-review-a-look-at-fm2-with-a85x/6 - that shows a similar 31 watts for the Trinity platform, but only 22 watts for an i3-3225... Yeah, makes Trinity look a tad bit less impressive eh?

Perfectly shown. Specially highend boards with alot of "junk" vs a simple lowend board will have a large difference.

 

[ehume]

Interesting discussion on variations in idle power consumption. As for second-to-second power consumption, we'd have to spend hours looking at Real Temp or something similar. But if you have your cpu fans on automatic, you will hear any sustained increase in power usage as the fans spin up. The spin-up can even alert you to your cpu doing some work you didn't expect. And it's a cue as to how little time your rig spends working above idle. It's fun, actually.

 

[lopri]

Rofl this reminds me of one of those AMD CPU reviews @AT where application tests were done on a 780G board then when it comes to power consumption out of nowhere an (unidentified) 790FX board was used.

 

[VirtualLarry]

Rofl this reminds me of one of those AMD CPU reviews @AT where application tests were done on a 780G board then when it comes to power consumption out of nowhere an (unidentified) 790FX board was used.

Does this imply anti-AMD (pro-Intel?) bias on the part of AT's reviewers?

 

[Exophase]

The numbers linked in the OP are 32 W and 39W. idle; 11x load.

I didn't say 2 W difference is all that important. In the early 2000s all you could hope for were 60+ W idle with some low power office Celeron system. I'm glad we are at about half of that, now.

Are you having a problem distinguishing the power consumption of an entire system from the power consumption of just the CPU? This is about comparing CPU power consumption at idle. If you didn't read that from my post then please take the time to read them again.

 

[ShintaiDK]

I remember certain Asus boards use up to 40W more than some simple board in power consumption tests. Same socket, same CPU, same everything but the board.

All boards cept the last is LGA1156 and using P55 chipset. 41W difference idle between lowest and highest.

At load it looks abit different. But still 25W difference in what should have been the same when used in reviews against other sockets.

 

[Abwx]

At load it looks abit different. But still 25W difference in what should have been the same when used in reviews against other sockets.

So you re implying that ANAND used the most efficient Trinity MB
and compared it with a crappy 1156 MB ?..

Can you provide a link showing that the FM2 board was carefully
choosen between the less power hungry , as you ve done it
for the intel set ?...

Remember that he used a full ATX mobo for Trinity , certainly not very
helpfull for lowish power comsumption records....

 

[OCGuy]

Would someone please give me a guess at what % of the human population actually would care about those minute differences in idle power usage?

 

[LogOver]

So you re implying that ANAND used the most efficient Trinity MB
and compared it with a crappy 1156 MB ?..

Can you provide a link showing that the FM2 board was carefully
choosen between the less power hungry , as you ve done it
for the intel set ?...

Remember that he used a full ATX mobo for Trinity , certainly not very
helpfull for lowish power comsumption records....

Not sure about FM2 board, but Intel DZ77GA-70K (from Anand's review) clearly not the best in power efficiency.

http://www.silentpcreview.com/article1256-page5.html

 

[know of fence]

Are you having a problem distinguishing the power consumption of an entire system from the power consumption of just the CPU?

Actually I do, and so do many others, partly because it's difficult to measure, it's not really separable from the rest and because memory controller and graphics also share the same die.

You used an esoteric (cpu alone) number to make a strawman out of my argument. Be it genuine confusion or sophistry ...
I still would like to know where you got that number, maybe you also have a number for the PCH/Southbridge, RAM and voltage regulators in idle state.

There almost is no difference (1-2W) between Ivy and Sandy, or i3 and i7 in idle power if tested in the same board. Is this how you arrived at 2 W?
http://www.hardwarecanucks.com/foru...-intel-i7-3770k-ivy-bridge-cpu-review-22.html

 

[ShintaiDK]

So you re implying that ANAND used the most efficient Trinity MB
and compared it with a crappy 1156 MB ?..

Can you provide a link showing that the FM2 board was carefully
choosen between the less power hungry , as you ve done it
for the intel set ?...

Remember that he used a full ATX mobo for Trinity , certainly not very
helpfull for lowish power comsumption records....

I assume you mean 1155. And no, I didnt imply anything. I just said you should be careful comparing across sockets and mobos that way, since there can be huge differences between mobos. And I showed you the facts behind.

Full ATX or not doesnt really matter as you can see on the LGA1156 numbers. Not sure why you even mention that.

Furthermore it seems Anandtech reused old numbers due to different SSDs listed. Maybe they also used different PSUs.

Here is another example from LGA1155: