"Every Watt Counts: AMD E-350 vs. Intel Core i3-2100T" article from xBits

"The variety of components for small energy-efficient systems keeps growing day by day. In this review we are going to talk about energy-efficiency processors: AMD E-350 (Zacate) and Intel Core i3-2100T (Sandy Bridge). We will also discuss new Mini-ITX mainboards: Gigabyte E350N-USB3 (AMD Brazos platform) and Zotac H67-ITX WiFi (for LGA1155 processors)."
http://www.xbitlabs.com/articles/cpu/di ... 2100t.html


"It turns out that miraculous energy-efficiency of the AMD Brazos platform is in fact a myth created by smart marketing team at AMD. As we have just seen with our own eyes, it is more energy-efficient than ION2. But if you are all about saving energy, then you may want to go with an even better option. Core i3-2100T processor on Sandy Bridge microarchitecture with 35 W TDP allows you to build a system that will be more energy-efficient than AMD Brazos in most cases, and will be several times more superior than the latter in terms of performance-per-watt."
http://www.xbitlabs.com/articles/cpu/di ... html#sect0

You really will believe anything you read won't you? I don't know why you bother to spout this stuff.

I'll just quote someone elses rebuttal to this as it covers it nicely

The 2100T was tested using the notoriously ineffecient Zotac mother board. Apparently the particular AMD E-350 motherboard used in this test was not the most power efficient on the market either. They certainly can't be tested on the same board. All the tested platforms used the same power source.

I think you are missing the point. A 2600K idles within a few watts of 2100T.

That a 500 hp Ferrarri (2100T or 2600K, take your pick) is capable of sipping gas to within even a few miles per gallon of a Toyota Prius at all... is simply amazing.

Getting all worked up that the difference is 1 MPG or 5 MPG is missing the point. It doesn't matter. That the Sandy Bridge is even within hollering distance of the AMD E-350 is the point and is what amazes.

If I were the engineer who designed the Sandy Bridge, I would be very proud. If I were the engineer who designed the AMD E-350, I would be not so proud and would be searching for excuses, much as the fanboys from whom you quote are doing. Though my excuse would be that I didn't have the resources available to me that the Intel engineer did.

Apparently I am not alone in this perspective... it isn't the AMD Brazos that is miraculous... it is the Sandy Bridge that is miraculous:

What we know about computers-

I think some of this was actually mentioned in the thread you (CES) started-

1) for idle power consumption, the motherboard consumes much more power than the CPU does.
2) Motherboards vary widely in power consumption from one to the next.
3) Power supplies are not efficient at the low end of their capacity.

The very fact that they used an 880W power supply for this test pretty much makes the results irrelevant. The different motherboards is obviously a big factor in power consumption measured. What they should have done is measured power from the motherboard power header using a multi-meter.

If every watt counts, why did they use an SSD (which aren't that efficient). They should have used a moderate notebook hard drive.
If every watt counts, why did they use an inefficient 880 watt PSU?
If every watt counts, why did they use 2 stick of RAM instead of 1?

How does pointing out test results (could be called fact, but they are subjective) from other Xbit reviews considered being a "fan boy"? I would think that in the name of science we should try to be objective and critical of results, making sure the methods are pure to eliminate bias. Don't you agree? I'm not saying Brazos is great or not great- I honestly don't know right now. You could call me a Sandy Bridge "fan boy" if you want because I like my CPU. But that doesn't change the fact that the conclusions were drawn on this article from questionable methodology.

All of these reviews should be taken with a grain of salt as most authors seem to have an agenda they wish to push. Grab the data for your own use and make your own decisions.

That said, each of these platforms have their own pluses and minuses. I wouldn't use the E-350 for a desktop PC, but it might make a splendid HTPC or nettop. Also, note the price difference between these two builds. From NewEgg (using the lowest priced Asrock mobo's):

Asrock E-350 and mobo = $130

i3-2100T = $135
Asrock H67 mobo = $90
Total = $225

I refuse to disagree with you,

Unfortunately we are limited to the data that is available. And we can often delay buying decisions.... but in most instances at some point you have to make them, usually with imperfect information.

I have certainly made my fair share of purchasing mistakes based on imperfect information.

Can you honestly say that there isn't something there in the xbits results... as imperfect as the testing was?

Very well put.


That comes to $225/$130 = a $95 differential which is a 73% increase in cost.

Shopping at Microcenter I get a slightly lower differential
H61M-P23 (B3) LGA 1155 H61 mATX Intel Motherboard $57.99
http://www.microcenter.com/single_produ ... id=0359807
Intel Core i3 2100 LGA 1155 3.1GHz Boxed Processor SKU: 369124 $99.99 (the 2400 costs $50 more)
http://www.microcenter.com/single_produ ... id=0359809
That comes to $157.98/$130 = $28 differential which is a 22% increase in cost.

Let's assume $200 for a case, PSU and 4G-8G of memory It isn't that hard to spend $100 to $150 on just one of those items.
(and we will even assume the software is pirated) and you maintain a stock of extra keyboards, mice, fans and video displays.
How about another $200 for a current generation SSD boot drive (once you get addicted you can't go back).
That comes to $557.98/$530.00 = $28 differential which is a 5% increase in cost.

You can pick away at my assumptions. But most people will end up spending between $500 to $900 before they are done. Even if they are using parts from their personal stash... they weren't free when they were purchased. And if you have to buy a key board, a mouse and an LCD display.... those costs add up.... especially if you have refined tastes.

Bottom Line: The cost differential isn't that much in comparison to the total investment.

Color me a Sandy Bridge FanBoy

Microcenter actually had the i3-2100 for $99 with a FREE motherboard 2 weeks ago.

Last week they had the i3-2100 + motherboard for $119.

I almost bought one just to have it. And yes, you can color me a Sandy Bridge "fan boy" as well. Hey, it's like magic on a chip.

I do think there is something to the Brazos. The integrated GPU is the best yet.

Me too... but I think if I am going to do it, I should get at least the 2400.... and maybe just hold my breath for a few mores weeks for the Z68 motherboards. Why not?

As I understand the criticism of this review, it has two parts.

1. It used a very large PSU that was operating at very low levels of output and efficiency.

That probably favored neither chip, though it probably didn't help accuracy if at lower levels of current the PSU would deliver more erratic efficiency. If the efficiency was stable, it shouldn't matter.

2. The AMD was used on a non-efficient motherboard as was the Sandy Bridge.

You can't use the same board and it is somewhat difficult to control for the effect of the board. Users are stuck selecting for the boards that are available. I guess one way to control for this would be to use the most efficient motherboard available for each chip.

For the Sandy Bridge, that would be the Intel. One of the critical commentators suggested a specific board for the AMD.

===============
If we used a Pico PSU and used more efficient boards for each of the chips, is there reason to believe that the AMD energy performance would have improved more than the Intel energy performance would improved... or vice versa?

Are there any other testing issues that I am overlooking?

We live in the era of the die-shrink, and the benefits of Intel's 2nd generation Sandy Bridge 32nm technology are fairly obvious. The AMD E350 is not however a 32nm part, it is 40nm http://www.cpu-world.com/news_2011/2011010501_AMD_launches_low_power_2011_platform_with_4_new_APUs.html. While this is a step on from the existing 45nm AMD CPUs it is too small a step in my opinion, limiting both the performance and energy saving ability of the processor. This is why according to most reviews the E350 falls between Intel Atom and CULV in terms of performance. Sure the E350 is a low power processor but unless/until AMD release a 32nm version it simply isn't in the same league as Sandy Bridge.

Well then, would be reasonable in your opinion to say the personal attack that I got for the temerity of posting this for discussion "You really will believe anything you read won't you? I don't know why you bother to spout this stuff." was out of bounds instead of me?

Yes, I think it would have improved the AMD performance and made it "look better". Imagine if you will, those results with AMD -6 watts across the board, and sandy bridge -2 Watts. That would change things wouldn't it? That would make the idle and load values for the AMD brazos platform look a lot closer to expectations.

Brazos- ultimately I think with a 32nm die shrink this would really shine as a tablet or netbook APU. The graphic performance blows everything else away. The actual computing power, number crunching and encoding power, is weak. What devices tend to avoid those types of operations? Tablets and netbooks. I would even go so far as to say that in certain circumstances it's better to give up some CPU power for added GPU muscle.

Both Sandy Bridge and Brazos are amazing, just in different ways. Brazos is cheaper (and I think will get a lot cheaper than this suggests. Zotac full featured boards are expensive) and smaller, sandy bridge is bigger but surprisingly affordable too.

The idea that I got an i5-2400 for $150: With performance people used to spend $700 on... and the low heat output- it's amazing.

My only problem with Sandy bridge is that I run too cool. @1.06 volts, my idle temps are 18-19C. 30-32 at load. I think my CPU is running too cool.

Intel is definitely still ahead of AMD in power efficiency and performance per watt. What helps a lot with this is their aggressive schedule for die shrinks, for a long time they've been like 12-18 months ahead of AMD in this area, which obviously really puts AMD at a disadvantage. Next-gen Brazos on TSMC 28nm should be more competitive with Intel's 32nm chips.

The reason most people are so excited about Brazos is because it has a really good balance between cost, performance, and efficiency for netbooks/ultraportables and mini ITX boards. For example, if money wasn't an issue and I was in the market for an ultraportable, I'd probably get something CULV based because they tend to offer better performance than Brazos and still have comparable or better battery life. But good luck finding a CULV ultraportable for $400-500, normally they're far more expensive than Brazos laptops are. Brazos is priced to compete with ION laptops and mobos, and against these it does pretty good.

Please note X-bit Labs' methodology:

As opposed to the SPCR metodology they measure the power consumption directly on +3.3, +5 and +12V DC rails after PSU. It means the efficiency of PSU is completely irrelevant.

I conclude from their previous articles that they used Allegro ACS713-30T sensors on DC rails as described here.