What are the Athlon II X4 925 and 945 on page 1? I'm not seeing them on the AMD site. Are they the Phenoms 925 and 945? I was under the impression that only Phenoms had L3, so that would make sense.

I would have been interested in seeing some dual cores CPU's in the test.

Typos. Corrected.



X2's and X3's another day.

Once again, a SPCR review on a high level - thanks

It's good to be updated in the CPU field, though, I dont have an itch to buy one of those tested;

The 910e seems to be a better regarding consumption, than my current Phenom II 925, but not enough to justify an upgrade:

My Phenom II X4 925 in a Gigabyte GA-MA78G-DS3H idles @ 54-55 watts and consumes 125 @ load.

Good benchmarks & analysis. However, I missed an important area regarding these low-power CPUs: are they worth the extra money for the tweaker? For example, could SPCR's X4 630 undervolt/clock to match the 610e? Running in spec is essential for a lot of users, but this forum has a lot of hobbyists willing to trade 30 minutes for $50...

Last line in article:



Once again, the 'energy-efficient-way-more-expensive' processors fall short. Here's a quick calculation that may interest some people: (would have been cool to see it in the article but it's ok spcr!)

the X4 910e system used 34.7 W-hrs during the tests, while the much better performing X4-635 used 38.6 W-hrs. If one were to run these tests 24/7 on both computers, running the X4 910e would save you roughly .0936 kWh per day. Say your electric bill is about $ .11 per kWh (US national average)..this means it would take you 681 kWh's to recover the extra money you spent on the energy efficient processor. Given you are only saving .0936 kWh per day, you would have to run this system 24/7 for 6818 days or 18.68 years to break even on the increased cost of this processor.

Is it worth it? Unless you are on a mission to save the environment from the evils of silicon valley...no.

Amazing how in the fast-paced world of consumer electronics, this story hasn't changed in 4 years. Who buys these AMD 'e' chips?

I have some questions about this in the Benchmark Power Consumption results:

"Our timed benchmark results multiplied by the average power consumption during each test gives us an estimate of how much total power was used during testing."

It seems to me that there could be a flaw in the logic here. The above statement assumes that the system will be turned off when a user is not performing the activities in the test (virus scan, iTunes, etc). I think it is more reasonable to assume that both computers will remain on for same amount of time, because that is the way it would work in the real world. The computer that finishes first may be idle while the other one is still working, but I doubt it would usually be turned off.

Also, I don't think that a quad core is necessarily the most efficient processor for these tests since not many applications can use all the cores at once. That was one reason why I wanted to see some dual cores in the test, and testing of dual cores against the quad cores (in addition to testing between the dual cores).

Do we actually know what the street prices are for these CPU's (based on ads from retailers)? Sometimes manufacturers will adjust prices before they hit they streets based on the competition.

No flaw that's how efficiency is calculated while doing work; when it's idling it's just a space heater.

It's important to know the power used under those two conditions so one can evaluate the CPUs for different applications.

It's surprising to see that the worst possible performing motherboard in idle/load power was chosen for the Intel CPU tests, whereas the AMD CPU tests were performed using a motherboard that's not the worst possible for that platform, as shown by previous reviews on SPCR. I would've thought the Intel DH55TC to be a better choice for the Intel CPU tests, considering it's more normal idle/load power numbers.

The conclusions drawn about the total amount of power used in a 24x7 interval should not use that particular calculation. It is a serious logic bust.

I also disagree with you about how efficiency is calculated (or should be calculated). In the real world our computers are used as space heaters whether we like it or not. Obviously, a 2.8 GHz processor is going to finish tasks quicker than a 2.6 GHz processor, but the machine does not shut off immediately when the task is done so it is erroneous to assume that the machine that finishes first will not use any power in idle mode while the other one takes a little longer.

I'm wondering about the AMD mobo; Was the IGP or the Asus 9400 + disabled IGP used? If it's the latter - how efficient is the mobo at shutting down the IGP?

Cant seem to find the info in the article, have I overlooked it?

Is it well known that the 1156 board we chose is very inefficient? We only have two 1156 boards and for all we know the DH55TC is the exception rather than the rule. We made the choice because we have had fewer problems with the Asus board and most people don't use Intel branded boards.



Yes, this does assume that the system is turned off after using it. However if we add idle time, how much? How long does the average PC stand idle? This varies from person to person, so it's just easier to assume that the user is very energy conscious.

You can also say the test suite isn't representative of average use and throw the power numbers out altogether.



Yes, the IGP was disabled. The difference can be calculated easily since we know the graphics card's idle power. I'll test it next time I'm in the lab.

[quote="Lawrence Lee]Yes, this does assume that the system is turned off after using it. However if we add idle time, how much? How long does the average PC stand idle? This varies from person to person, so it's just easier to assume that the user is very energy conscious.

You can also say the test suite isn't representative of average use and throw the power numbers out altogether.[/quote]
No human can possibly be that energy conscious.

Multiplying the watts used in the tests times the elapsed time of each test is extremely misleading. If you want some sort of total energy consumption, then just assume some total amount of time the computer running CPU intensive tests, for example that might consume about 10-25% of the time for the (for the slowest PC) and then balance is idle time (the idle time will vary depending on how long each PC takes to complete the test), and then calculate a total energy consumption with idle time (and idle watts) included, instead of assuming the computer is turned off. Even 25% would only be true for a ridiculously energy conscious person who doesn't think, read, or type much (even typing is close to idle power). I would bet that real world idle CPU intensive percent is closer to 5% for most users.

The problem is that the person in a post above who calculated the total watts and the cost made erroneous calculations based on the formula used.

Looks like I'd associated the Asus P7H55D-M EVO as one of the first boards that appeared in the early Core i5 661 reviews that had much higher than usual idle wattage, but it was in fact the Asus P7H57D-V EVO I was thinking of.

In any case, there's still quite a significant difference between the Asus P7H55D-M EVO and Intel DH55TC, as shown in these links:

http://www.silentpcreview.com/article1019-page5.html
http://www.xbitlabs.com/articles/mainbo ... evo_8.html

Edit: Looking at the xbitlabs link, I agree the Intel board is probably the exception here, with the Asus boards being the two highest. If only you had a Gigabyte board for a middle of the road number

For some reason Intel boards are reasonably priced and fairly easy to find in Australia. Features like undervolting are not as useful these days, and overclocking isn't nearly as necessary as it was back in the Celeron 300A or P4 days. Speedfan support is nice to have, but with these low power CPUs and amazing aftermarket heatsinks and low speed 120mm fans, I haven't had any use for Speedfan.

It's very difficult to find accurate power testing data anywhere as most sites don't seem to know what they're doing, or don't care about the data they're recording as long as it fits on a graph and looks pretty, then have 10+ pages of graphs devoted to games that never ever vary more than a couple of frames between each board, but are then critical about a product if there is a 0.2% difference found.

Thankfully SPCR is different and the reviewers know what they're doing and care about the data being presented.

Thanks

I'm a little excited to hear the result, and I'll try to dig up how much power is saved by disabling the IGP - initially, I think the Phenom II X4 910e + Asus M4A78T-E combo is quite some more less power hungry, than my Phenom II X4 925 in a Gigabyte GA-MA78G-DS3H - but still not enough reason to upgrade

There's a few jokers in my comparision, PSU etc.

My PC probably only runs at 80-100% load less than 10% of the time, so idle power is my main interest. And usually, I'd like to spend a little more, to get lower overall power consumption - but 40% extra for the 910e vs. the 925, is not worth it.

Have you actually seen a retail price for the 910e? Until you do, don't assume anything.

The 'e' chips are nice for business and OEM type applications where you don't want to worry about instability or other possible issues from undervolting. Such as if a software undervolt is required due to hardware limitations. Usually that requires the software to be running which doesn't work well when you have multiple people logging on or no-one logged in for large chunks of time.

Now, is it worth it for the price? Not usually on the newest CPUs. But eventually the prices come down and the difference for the 'e' CPUs is negligable in which case I would always pick the 'e' over the normal. The other possibility is that maybe you don't need a 3.0GHz CPU and 2.4GHz is plenty. Yeah, you could get more CPU for the same price, but you probably will do very little that could use it or that would make a noticeable difference in processing time.

I picked up a 4050e a while back. Yeah, I probably could have got the 2.6GHz normal CPU for the same as the 2.1GHz e one. But that probably wouldn't have really done much to the overall performance and having a slightly lower-powered and cooler CPU is always nice.

I also purchased a 4050e several years ago for about $50. Recently upgraded to 240e (2.8 GHz) for about $75.

My 240e runs on a Asus VM2A mb and idles at 21C (according to Asus). This is probably not exactly accurate, since ambient temps are a little higher than 21C, but still pretty amazing since it Xigmatek S-1283 is passive (no fan), although there is the P180 Mini top fan not very far away (on low speed and inaudible with under-desk location).

Aside from the logical errors that I pointed out in previous posts above concerning measurement formulas that compare CPUs (and that I have not seen an actual retail price yet for the new AMD e chips mentioned above), I find some of this discussion about cost to be nitpicking. I got into quiet computing in early 2004 (after my old Gateway PC starting sounding like a vacuum cleaner). Since then I have built two new computers and spent boatloads of money on fans, heatsinks, fan mounts, sound deadening material for cases, fan controllers, etc, etc.

All things considered, I find that the AMD "e" processor makes quiet computing much easier, and actually cheaper in the long run compared to all the other stuff I previously spent money on to make my computer quiet. Also, unless one is into high-end gaming or advanced multi-media, I am not sure a quad-core is cost or energy efficient. In other words, IMO there is too much emphasis on total processing power, versus on the actual power draw.

Agreed, in large part -- re, the emphasis on total processing power. In general, most CPUs with dual cores that approach 3ghz speed are plenty fast enough for most apps.

I'm also using a 240e, in a htpc, w/ a basic 80+ PSU, 785 chipset, 4gb ram and 500gb notebook drive. Keeping it cool & quiet is trivial -- and it draws less than 35W at the wall in idle, maybe 10W higher in HD playback peaks.

Still, one of Larry's key points is that standard AMD chips undervolt quite well, and they may be just as CnQ as the higher priced, lower TDP parts when undervolted by the user.

Well if you're not gaming or encoding, why get a quadcore at all?

Personal rant here
AMD has disappointed me ever since K10 Barcelona design came out. It was not revolutionary in any way and is still held back by the same bottlenecks the first Athlon64 CPUs had. But with increasing core number and available memory bandwidth those bottlenecks are more obvious.
Thus all current AMD CPUs are in my mind inefficient by design. And no undervolting is going to change that.

That's the current retail price - it might lower, but who knows...

There's some (only a few?) exceptions - me? I'm running some Nikon software, which is ridicolus CPU consuming. Except for that, I dont need much CPU power - my 4850e was good in all other aspects and sadly, it's now collecting dust...

Does that Nikon software make use of more than 2 cores (or even one core) at a time? I know that some software can do this, but certainly not all software.

When I say retail price, I mean the actual price charged by a retailer such as Newegg. I don't mean what AMD says is the projected retail price. Last time I looked, there was no price listed on NewEgg (chip was not in stock).

I suspect that both AMD and Intel make adjustments to their prices based on competitive offerings.

I read this, but...

Do these CPUs undervolt better than normal ones, or are they just the same with lower starting voltage and frequency?

CPUs are invariably sorted by binning. Presumably, these CPUs do run stably at lower voltages than normal, and they are preset to run at lower voltage w/o any user intervention. From the article:

(That's quite a sentence!) The review provided all of the information (idle power, load power*, benchmark time) you need to produce the results you want. Doing that's a lot easier than backing other numbers out of the results you suggested. So, go for it! I'd be interested to see how the rankings move as you change the overall duration.

* A PC's power draw will vary during some of these benchmarks, so errors could creep in. How did SPCR determine the average consumption? Were any comparisons made to the Power Angel's KWh?

Looking at it a little more changes my opinion. The differences in idle power are so low, there simply won't be big changes. For example, the i5-750's idle is only 1W more than the 610e's. Balancing the benchmark energy would require 6.3 hours' idling. Certain combinations may show more difference, but it's simply not worth spending much time worrying IMO. Especially now that individual CPUs run at different voltages.