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

[Enigmoid]

I have a hard time believing that these numbers are actually due to the cpu and not something else. I mean mobile intel cpus (ivy bridge) usually idle between 3 and 4.5 watts. I would expect the desktop versions to be slightly higher perhaps 5-7 watts (for a dual core i3). I don't think trinity can idle at 7 watts below that.

 

[Exophase]

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

I didn't make a strawman out of your argument, it's really quite the other way around - I said that CPU idle power consumption was so low now it barely matters. You then came in and started using these full system idle numbers to try to contradict my statement. If you think I'm against motherboard manufacturers and other system integrators lowering their idle consumption then you have the wrong idea, because I'm not.

I didn't take a 2W number from SB vs IB, it was just a rough figure based on numbers I've read from various sources at different times past. Yes, it's hard to measure the power consumption of the CPU because the rails are split and shared with other things. But measuring the 12V line to the CPU will give you a much closer number than measuring at the wall. This for instance shows 5-6W for i5-2500K/i7-2600K:

http://www.lostcircuits.com/mambo//...ask=view&id=103&Itemid=1&limit=1&limitstart=6

This doesn't include some parts of the CPU drawing from 3V and 5V, but it's also inflating the number because of inefficiency in the voltage regulators coming off from 12V. Suffice it to say, the idle consumption of ultrabooks and tablets quickly invalidates that the full APU chip - using the same die - are taking anywhere close to the whole system figures presented in this thread. They're obviously measuring power consumed by a bunch of other stuff, and the inefficiency of the PSU converting from 110V AC.

 

[know of fence]

I have to admit that I did misread your initial comment. It wasn't clear to me which one of the Idle numbers is the negligible one - the former or the latter, foolishly I assumed that it's the one the OP is referring to. Thank you for the link, BTW.

 

[Fjodor2001]

So if the IB and Trinity CPUs themselves idle at around 2-5 W, then what's the rest of the ~30-35 W system idle power consumption being wasted on?

What would be a reasonable teardown analysis of how the wattage is being spent between the different hardware components?

And how can some motherboards waste as much as 40 W more than other ones at idle (as ShintaiDK mentioned here)?

 

[ShintaiDK]

So if the IB and Trinity CPUs themselves idle at around 2-5 W, then what's the rest of the ~30-35 W system idle power consumption being wasted on?

What would be a reasonable teardown analysis of how the wattage is being spent between the different hardware components?

And how can some motherboards waste as much as 40 W more than other ones at idle (as ShintaiDK mentioned here)?

All kinds of crap added. 3rd party SATA, 3rd party NIC(s), discrete soundcard chip. CIR, Bluetooth, 3rd party USB, nForce/PLX bridges, wifi, firewire, EPU, TurboV, more phases and whatever crap they else invented and/or added.

 

[pelov]

I have a hard time believing that these numbers are actually due to the cpu and not something else. I mean mobile intel cpus (ivy bridge) usually idle between 3 and 4.5 watts. I would expect the desktop versions to be slightly higher perhaps 5-7 watts (for a dual core i3). I don't think trinity can idle at 7 watts below that.

AMD spent a lot of time on getting their idle power consumption down. This isn't a Trinity-only thing, as Bulldozer had very low idle power consumption as well. It's actually one of the very few redeeming aspects of the architecture. Under load it's absolutely atrocious, but at idle it does very well.

And, yea, the other parts on the PCB chew up a whole lot of power. The chipset itself can chew up 3-5watts. The WiFi or ethernet in the 1>6W range, then there's the SATA, PCIE, etc.

One of the ways to tone down power consumption is essentially to get rid of these entirely or integrate them into the CPU via SoC design. It's the reason the Atoms are able to get to such low wattage and AMD's Hondo can reach the 7W-8W range.

 

[Fjodor2001]

All kinds of crap added. 3rd party SATA, 3rd party NIC(s), discrete soundcard chip. CIR, Bluetooth, 3rd party USB, nForce/PLX bridges, wifi, firewire, EPU, TurboV, more phases and whatever crap they else invented and/or added.

But even systems built using low spec motherboards aren't able to go below ~30 W at idle. They don't have most of the stuff you mention. So what are those 30 W wasted on then?

 

[ShintaiDK]

But even systems built using low spec motherboards aren't able to go below ~30 W at idle. They don't have most of the stuff you mention. So what are those 30 W wasted on then?

I think you confuse mobos and the entire platform as I linked. The only difference between those tests are the mobos. There are still discrete card(GTX260) and a HD(Raptor X) etc. The title says system power consumption for a reason.

 

[Fjodor2001]

I think you confuse mobos and the entire platform as I linked. The only difference between those tests are the mobos. There are still discrete card(GTX260) and a HD(Raptor X) etc. The title says system power consumption for a reason.

I was referring to the system idle power consumption mentioned in my original post, see here. Then there was no discrete GFX card etc.

Basically a minimal system will only consist of CPU + Motherboard (low spec) + RAM + SSD. But even such a system cannot go much below ~30 W. The CPU should idle at around 2-5 W, and the SSD at a couple of 100 mW. So it's only RAM and motherboard (with the individual HW components it consists of) left.

I.e. the question still remains: Where are the 30 W idle system power consumption wasted on such as system?

 

[ShintaiDK]

I was referring to the system idle power consumption mentioned in my original post, see here. Then there was no discrete GFX card etc.

Basically a minimal system will only consist of CPU + Motherboard (low spec) + RAM + SSD. But even such a system cannot go much below ~30 W. The CPU should idle at around 2-5 W, and the SSD at a couple of 100 mW. So it's only RAM and motherboard (with the individual HW components it consists of) left.

I.e. the question still remains: Where are the 30 W idle system power consumption wasted on such as system?

You forgot PSU efficiency at that load. It can easily be 10-20%. If I am not mistaken AMD still uses their 1200W PSU.

Also you can see the PSU efficieny must increase rapidly. Unless the change for a i3 3220 is 24W AC side from idle to load.

You can find other Atom/Bobcat/VIA tests also showing very high idle loads due to PSU efficiency.

Doesnt take much to guess where that graph will go for say 10W or 20W DC load.

 

[know of fence]

AT uses 110 VAC with presumably a moderately high powered PSU on their test platform, so their efficiency is probably ~70% at those low idle numbers.

TechReport has Trinity at 24W idle, which only makes sense if measured after the power supply.

Source: http://techreport.com/review/23662/amd-a10-5800k-and-a8-5600k-trinity-apus-reviewed/7

CPU needs <5W which includes 12V to Vcore conversion, but doesn't include L3 cache (for intel CPUs only) or memory or graphics. We know roughly about at how much discrete cards idle. So 5-15 W for the IGP+Memory? Mind you Trinity graphics still isn't the same as their 28nm Graphics Core Next cards, but VLIW4 shrunk to 32nm.

There is also a cooler on top of the Southbridge (A85 chipset or whatever it is called), which we know is traditionally built on older process nodes. And of course there is everything else, though I don't see a reason why the I/O-controllers would need much power when idle.

Anyways for PCs just like at home, power consumption is the result of many little things, but mostly it's voltage conversion and graphics, which eat up power. So Intel's promise to reduce cpu power by 20x with Haswell among other things, maybe won't affect system idle power all that much, probably not even cut it to half.

 

[alexruiz]

The Gigabyte A85x FM2 board used in Anand's review is second worse only to the MSI A85x FM2 as far as power consumption. The ASUS uses less electricity, and big interrogant is the Biostar A85 MW that I suspect should be the best. Even more, any current FM2 mobo using the A85x will use quite a bit more than a A75 based FM2 board simply because the available mobos are not stripped down; in fact, quite the opposite, all the available A85x based FM2 boards are feature rich, and as others have already said, mobos have a big difference in consumption even between same socket and manufacturer depending on chipset and features.

Furthermore, successive BIOS revisions always help to bring the mobo consumption down. All those A85x FM2 boards will consume less in a few months after newer BIOS are released.

AMD did a good job on the idle consumption, but as some others have also mentioned it, it was already there since BD.

BTW, has anyone seen a Trinity review performed on an A75 or A55 based board? I speculated that the retail launch was more a showcase of the A85x chipset, and so far, all the reviews have been using mobos based on a85x, hence making me think that my speculation was correct

 

[sm625]

For most people idle consumption is by far the most important, followed by stand-by consumption.

I would say for most people, power consumption while a flash game is running in the foreground is the most important metric. People leave those games running all the time, not thinking about how horribly wasteful they are with system resources. It is really quite aggravating.

 

[Fjodor2001]

You forgot PSU efficiency at that load. It can easily be 10-20%. If I am not mistaken AMD still uses their 1200W PSU.

Also you can see the PSU efficieny must increase rapidly. Unless the change for a i3 3220 is 24W AC side from idle to load.

You can find other Atom/Bobcat/VIA tests also showing very high idle loads due to PSU efficiency.

Doesnt take much to guess where that graph will go for say 10W or 20W DC load.

Why would you want to use a 1200 W PSU for a ~100 W max TDP system?

It would be better to use a picoPSU + PowerBrick which would have an ~85% efficiency even in the ~30-40 W range (when the system is idle), see:

http://www.silentpcreview.com/article601-page4.html

 

[ShintaiDK]

Why would you want to use a 1200 W PSU for a ~100 W max TDP system?

It would be better to use a picoPSU + PowerBrick which would have an ~85% efficiency even in the ~30-40 W range (when the system is idle), see:

http://www.silentpcreview.com/article601-page4.html

Ask Anand. But unfortunately its the same on most sites. The point is you can reuse numbers everywhere since the platform stays the same. Even tho the results actualy ends up misinformative.

 

[LuckyKnight]

Would a Mini ITX board consume less than a mATX board?

The boards that I am looking for with my HTPC was originally ITX, but since none of them are compatible with the case I was looking at (heatpipes in the way) - I was looking at mATX.

http://www.overclockers.co.uk/showproduct.php?prodid=MB-545-AS&groupid=701&catid=1903&subcat=2399

http://www.overclockers.co.uk/showproduct.php?prodid=MB-098-AK&groupid=701&catid=1903&subcat=2399

http://www.overclockers.co.uk/showproduct.php?prodid=MB-099-AK&groupid=701&catid=1903&subcat=2399

Also - I haven't seen a power consumption comparison between the dual core and quad core variants yet (65W)

Question is - which one has lower power consumption - if different anyway. I don't mind paying literally 3 pound more for SATA 6GB, but not at the cost of much higher power consumption, as it will be run passively using heatpipes.

I was looking at ASRock simply because they are the only ones doing CIR headers for HTPC. The lack of decent mITX boards/availability is very disappointing.

 

[Rvenger]

Would a Mini ITX board consume less than a mATX board?

The boards that I am looking for with my HTPC was originally ITX, but since none of them are compatible with the case I was looking at (heatpipes in the way) - I was looking at mATX.

http://www.overclockers.co.uk/showproduct.php?prodid=MB-545-AS&groupid=701&catid=1903&subcat=2399

http://www.overclockers.co.uk/showproduct.php?prodid=MB-098-AK&groupid=701&catid=1903&subcat=2399

http://www.overclockers.co.uk/showproduct.php?prodid=MB-099-AK&groupid=701&catid=1903&subcat=2399

Also - I haven't seen a power consumption comparison between the dual core and quad core variants yet (65W)

Question is - which one has lower power consumption - if different anyway. I don't mind paying literally 3 pound more for SATA 6GB, but not at the cost of much higher power consumption, as it will be run passively using heatpipes.

I was looking at ASRock simply because they are the only ones doing CIR headers for HTPC. The lack of decent mITX boards/availability is very disappointing.

An ITX board could essentially have less power consumption than a MicroATX or ATX. Again, it depends on how many features that mITX board actually has. Many of them have Wireless adapters, nice audio codecs for HTPCs so they very well could consume more power than a stripped down MicroATX board.

 

[LuckyKnight]

I narrowed it down to the ASRock FM2A75 Pro4-M.. simply because it's the only mATX ASRock have with a HDMI port!!

I wish more manufacturers would adopt CIR. There was a nifty looking Zotac ITX board a few months ago, but no sign of it. Still, doesn't have CIR anyway.