http://www.hexus.net/content/item.php?item=30964&page=9

no undervolting, but everything else looks pretty fair.

Fair? I doubt it. Sensible? Surely not: we know Sandy Bridge systems can consume less than half that! Credible? Not at all in my opinion. There are inefficient 1155 boards (among other tricks) you can use to make the platform look bad. There's lots of transparent dishonesty on tech sites. Recall how Zacate was hyped!

It'd be great if AMD made CPUs that idle lower than Intel's. In spite of all the hype about Sandy Bridge, it's obvious idle power consumption could be better and that it would be a material improvement for people who have small heatsinks or loads of cores. But the big potential savings when it comes to idle power consumption have been about othere components than the CPU for a long time.

Some odd choices in that comparison but even so it seems to be valid.


First odd choice was to use a GT430 GPU for the Intel while the AMD uses ~6550D, why not pop a 5550DH card in there? its supposed to be near identical to 6550D right? Second there is no info about what motherboard was used with the Intel platform other than it was Z68 on some of the tests. this will affect power draw. I've seen i3 systems draw less power than what the review shows. it may have been on H67 chipset. I wouldn't call this a complete idle power comparison of the two platforms, only a specific observation done while testing this particular AMD APU.

Still the power consumption numbers look fairly good for AMD especially when using the integrated GPU - the 6550D, and it looks like a decent performer for what it is.
If you read between the lines it seems like yet another recycle of the same old tech from AMD, something I wouldn't touch myself. But as a budget/low power/HTPC system it is an interesting choice.

ah.
wasn't paying too much attention to the motherboard models, though i'm aware of the possible differences.

list of the test platforms:
http://www.hexus.net/content/item.php?item=30964&page=3
looks decent to me, only the motherboards and cpus are different and triple channel ram where available.
they probably didn't choose the most efficient intel motherboard, but who can say the amd a75 one is any better?

29w idle with no undervolting and a 400w psu is quite good i think.

Their numbers do! That or they rigged the test in other ways. There can't that much difference between the CPU's.

Your power consumption could be under 15W with a good PSU (there are good high-wattage PSUs nowadays). Without undervolting, maybe 16W is the limit but it doesn't make much difference at idle. People have done that kind of thing with Clarkdales as well as with this year's Intel crop. All you need is to pick the right desktop mobo (that ain't going to be an Asus). I'm saying "desktop" because, if you choose parts well, mobile gear goes under 10W idle.

HFAT, you could be right, but considering my D945GSEJT with extremely minimal equipment (1 stick of low voltage RAM, 1 fairly low power USB drive) idles at ~9W AC (and 11w with a mobile HDD, more ram, etc), I find it hard to believe that and SNB equipment (mobile or otherwise) idles this low. Even my eeePC (N280/945GSE based) idles (with the screen off) at ~9W. I could be wrong, though.

Sure. But is it better than Sandy Bridge? I get 18W, with a picoPSU but still.

This is why we need SPCR - that Hexus review has a different audience in mind, testing with a 1000W PSU and stepping down to 400W for the IGPs..

looking forward to the spcr review indeed!
the spcr test platform has the i3 2100 at 38w idle including a 9400gt and using an asus motherboard.

i will make use of the integrated gpu so i'm getting one either way.
but it's interesting how a lot of the reviews i have skimmed through have the amd a8 consuming less at idle.

The chipset which comes with these boards was outdated when it was released, never mind now. Intel picked it because they could make it cheap I guess. Its power consumption only looks good when compared to desktop parts (it's a mobile chipset with a mobile CPU).
There are (more expensive) mobile Atom boards which consume less than 5W at idle and less than 10W on load.
There are reportedly Macs which idle lower, stock. And they're not fitted with Atoms but with proper CPUs. That's because they use expensive mobile parts.

On a budget, I don't think you can break 13W or even 14W with conventional CPUs.


I bet none of these reviews have tried to measure the CPU's power consumption. Modern CPUs do not consume a whole lot at idle.
You can make boards with AMD chipsets which consume little power at idle. I have such a board (custom HP board, available for over $400 as a spare part).
They're probably simply testing boards which are more efficient than the boards people typically use (or even more efficient than any board that's going to be available for retail purchasers). The same thing happened earlier this year with new AMD CPUs as I mentionned. How many times can you fool a lot of people?

I will note that every single review I've seen comparing idle power seems to be using an H67 ASUS board, and I admit that they are not known for their power efficiency... But then, we don't really know what ideal A75 boards look like yet, either. Idle power comparisons seem a bit moot, at the moment, until we get a larger sample.

cpu power consumption alone is moot, i will put 'system' into the topic title.

unfortunately the zacate test system was proved to be more efficient than retail versions.
this was disappointing but the engineering sample reviews were marked as such so i can understand that, and nettops such as the zotac zbox did manage to get as low as 11w in idle, so that's something.

as for the a8 reviews retail motherboards were used, made by msi, gigabyte, asus or asrock.

http://techreport.com/articles.x/21207/10
this compares 3 different motherboards from msi, asus and gigabyte, the i3 system in this case consumes less power.

Probably need to wait a month+ for more mobo's to hit retail and the frenzy of Bios updates to settle and then look at some reviews for more practical power results.

That said, Llano would make for a decent HTPC/light gaming rig. Looks like the HQV results are decent and it puts up some nice fps for lighter duty games at reasonable resolution.

The thing to watch as reviews come out over the coming months are the ram sticks used. It doesn't matter much if you are using an add in video card but if you are going APU/IGP and are measuring power draw and performance you get into these issues:

llano APUs benefit from faster ram on the motherboard more so than a traditional CPU + discrete video card. Early reviews have shown that even DDR3-2000 shows a benefit and recent price searches have shown that DDR3 1866 and DDR3 2000 are close in price enough that US residents don't have a good reason to buy DDR3 - 1333 or DDR3 1600 for use with llano.

Many ram kits run with higher voltages than considered base/normal/spec however you want to think of it. So I could buy DDR3 2000 at 1.5V and test it for power draw and performance and some web site can grab some DDR3 1333 and run it over spec at say 1.8V and skew the results slower while still using more power. Or just as easily another site could mention faster ram but suggest that it requires higher voltage even though you can buy faster ram at the same voltage.

Someone once said something about the 3 kinds of lies: Lies, Damn Lies, and Statistics. Apparently that comment is from the 1800s so it shouldn't come as a surprise anymore.

small mistake earlier, it should be 27w idle not 29w, my bad.

and here is the 65w a8 using only 16.2w at idle!
http://www.xbitlabs.com/articles/cpu/display/amd-a8-3800_21.html#sect0

16W? DC? AC? What components? Numbers without a methodology are pointless. These latest numbers seem plausible and are intriguing but are they credible? It's hard to say without details. Like everyone else, I'd like to see a SPCR review.


It wasn't only the engineering samples. Look at this piece of dishonest reviewing for instance, using a retail mobo:
http://www.anandtech.com/show/4134/the- ... -miniitx/7
Now compare to a clear-headed and competent review:
http://www.silentpcreview.com/article1167-page5.html

Which zbox idles at 11W?

Here's an analysis of the power consumption of a desktop Clarksdale (close to Sandy Bridge) showing why the numbers you posted in the first post are totally useless for the purpose of estimating idle CPU power consumption:
http://ssj3gohan.tweakblogs.net/blog/61 ... r-(english).html
The blogging software ate the period: it's not 85W but 8.5W! If you like extreme tweaking, I think you'll have fun checking out this link even if you don't care about the power consumption analysis.

http://www.xbitlabs.com/articles/cpu/di ... 00_21.html
Z68 + 2100 ~25 watts Idle
the asus Z68 is inefficient, it probably draws 5-7 more Watts just for itself by overvolting and ovecumbering the pcb with useless phases and chips.
there is also the 1.65V Dram and hdd 5-7 watts and PSU adds to that number.

http://www.xbitlabs.com/articles/cpu/di ... html#sect0
H67 + 2100T consumes ~10 watts with ssd drive

The h67 mini itx is how it's done probably but if you want efficient mbd take intel intel and compare that..
2100T consumes the same as usual 2100 cpu (same voltage in idle)

zbox review, it was 13w idle not 11w, pretty close for something from memory though!
http://hothardware.com/Reviews/AMD-Fusion-Hits-Retail-Zotac-and-Gigabyte-E350s-Tested/?page=8

i'm not saying all the reviews are perfect, but i presume that most sites take steps so that comparisons make sense. that anandtech review isn't very informative in this regard, but all the reviews i have linked to provide details on their test methodology and from what i can tell they all use the same components except for the motherboard and cpu.

test methodology link for the 16.2w a8:
http://www.xbitlabs.com/articles/cpu/display/amd-a8-3800_9.html#sect0
what they say:
The graphs below show the full power draw of the computer (without the monitor) measured after the power supply. It is the total power consumption of all the system components. The PSU's efficiency is not taken into account. The processors are loaded by running the 64-bit LinX 0.6.4 utility. We enabled all the power-saving technologies for a correct measurement of the computer's power draw in idle mode: C1E, AMD Cool'n'Quiet and Enhanced Intel SpeedStep.
the wording is a little confusing but i think they mean it's measured at the power socket, though 16.2w ac with an 880w psu seems implausible...

all power consumption comparisons do so with cpu+mb combos.
perhaps the fairest way we can hope for is with the most efficient retail boards available and using the same components and cooling?

nice yuu!
can we perhaps put some confidence in the a8 review now?

edit:
9.7w ac with an 880w psu!?!

9W, AC or not, is ridiculous. And the explanations of xbitlabs don't make sense in context of the numbers as you noticed yourself. There are ways to measure power consumption which are unreliable even if the reviewers are being honest (if a bit naive). So no, we can't put any confidence in these reviews.
Most if not all of the reviews cited with the exception of SPCR don't state how they mesure power draw so all the numbers would be suspect even if they "took steps so that comparisons make sense", an assessment I obviously don't agree with. Consider the sources, their qualifications, motivations and incentives...
In fact I wonder if this thread was a ploy to demonstrate the SPCR's value in the wake of Mike's announcement.

edit:
It's hardly practical to hope that someone is going to compare all available boards... there are more than you think!
This is the best we can hope for actually: http://www.silentpcreview.com/article1139-page4.html
That was SPCR's first Sandy Bridge review. As you can see, it's possible to do better than to compare system power consumption. There are other ways to do better. People who wanted to optimize system power consumption have published results of systematic tests in which components and settings where switched for instance. But you'll find none of this on your average review site...

being able accurately and reliably measure the dc power consumption of every separate component would be a godsend... don't think it'll happen though.

as it is we just have to use what we get, and the hopefully forthcoming spcr review would be very useful in this regard.

Outstanding SPCR review, but the OCZ low voltage is actually outdated 1.8V DDR3 marked as 1.65V probably 60/70nm DDR3 that draws more power. it would have been better to use some mainstream DDR3 based on 30/40nm.

My PC 2500K with SS-400ET. draws 49 watts AC 220V from the wall. ~ 35 Watts DC. by removing 500GB WDC and GTX460, that leaves 10 watts for the 2500K, H61 and DDR3 4GB.

It is no wonder that 2100T with HD1000 Gfx and SSD is around 10 watts on 220V, their methods are explained in this link and translated on xbit. http://www.fcenter.ru/online.shtml?arti ... ower/26716

I believe when X-Bit Labs says "...measured after the power supply. It is the total power consumption of all the system components. The PSU's efficiency is not taken into account. " means they measure the power at the PSU output, not the input.

They could have parsed it better.

Ok folks, my english in not that good (really), someone explain me this:

http://www.tomshardware.com/reviews/amd ... 975-2.html
"By default, Llano wants to be the primary output. So, when you configure your BIOS, set up Windows, and get drivers installed, AMD says use the APU’s outputs. If there’s a discrete card installed, CrossFire gets enabled automatically. At that point you have to install the latest Application Profiles patch, disable CrossFire, shut the system down, and switch over to the discrete outputs. From then on, you won’t see a video signal until Windows loads up. Enable CrossFire, reboot again, and then you should be able to use Dual Graphics. Whew."

Is it a mess to put ANY discrete card with llano? Or is it related to some low-target cards? Is it an isolated case? If i want to put my 5750 on it, is this procedure necessary?

Depends on the motherboard and BIOS. Some boards are defaulting to enable hybrid/asynchronous crossfire. Maybe you could do the process differently if crossfire is disabled until after you setup the discrete card and then enable it.

If enough people complain about this setup process they may change the BIOS defaults and that would change the steps needed to set it up.

Either way crossfire will never be as simple as just using a single video card.

Keep in mind that hybrid/asynchronous crossfire is only taken advantage of in games with DX10 or higher support. If your favorite games are DX9 based you could just disable the crossfire in the BIOS and use the 5750 by itself. If you have a mix of games that take advantage of it and that don't take advantage you can take the simple route and just use the 5750 by itself or do the research and figure out if it's worth turning on crossfire for specific games or if you have no older games that would be slowed down by using crossfire.