Questions on "Heka" Phenom II X3s

theycallmebruce · Post by **theycallmebruce** » Wed Jun 24, 2009 7:36 pm

Hey everyone.

I'm thinking of upgrading my Brisbane 4800+ and the Phenom II X3 (705e, 710, 720) CPUs look like a reasonably priced upgrade.

I see that the 720 comes in a black edition with an "unlocked multiplier". What exactly does this mean, and does it imply that the multiplier cannot be adjusted on the 705e and 710? I'm confused because I've been able to reduce the multiplier on my 4800+ with no issues.

I'd like a 705e, but here in Australia, I can only find the 710 at the moment. I read somewhere that they are the same CPU, different binning. Can anybody verify? Then I might get a 710 and see how low I can undervolt stably.

ascl · Post by **ascl** » Thu Jun 25, 2009 12:33 am

I am not 100% sure this is correct for AMD CPUs, but I think it is. "Unlocked" means you can adjust the multiple above the default. Locked means you can adjust down but not up.

ie, my Q6600 has a multiplier of 9 and is "locked". I can still adjust this anywhere from 6-9, but not above 9. Of course I can still adjust the FSB

Mats · Post by **Mats** » Thu Jun 25, 2009 3:31 am

You can lower the multiplier with all Phenom CPU's, but with the Black Editions you can also raise it, which is great when overclocking and especially good if you want to keep the low idle power draw.

If you don't care about overclocking then I'd suggest you get any of the three models. I'd go for the 720 BE, you never know when you need the extra performace.

The difference in TDP is usually not reflected in real world use, since the TDP is a limit and not an absolute value.
(SPCR's latest 65 W 9550s review shows the difference between that and a regular 95 W 9550, and I don't think AMD is much different.)
Lostcircuits.com measured 56 W power draw (CPU only) during load with the 720 BE.

quest_for_silence · Post by **quest_for_silence** » Thu Jun 25, 2009 3:52 am

Mats wrote:Lostcircuits.com measured 56 W power draw (CPU only) during load with the 720 BE.

Nonetheless it's a rather questionable statement.

Regards,
Luca

Mats · Post by **Mats** » Thu Jun 25, 2009 3:59 am

quest_for_silence wrote:Nonetheless it's a rather questionable statement.

Why?

LodeHacker · Post by **LodeHacker** » Thu Jun 25, 2009 4:06 am

Mats wrote:
quest_for_silence wrote:Nonetheless it's a rather questionable statement.
Why?

Simply because he can? Just read this "conversation" I had with this guy: viewtopic.php?t=54242

Pure ignorance? Judge for yourself. Also notice how he cuts my posts to the first sentence and quotes that, as if he didn't read the posts completely. Me angry

quest_for_silence · Post by **quest_for_silence** » Thu Jun 25, 2009 4:49 am

LodeHacker wrote:Simply because he can

I mean you must state on topic sir.

Regards,
Luca

quest_for_silence · Post by **quest_for_silence** » Thu Jun 25, 2009 5:13 am

Mats wrote:
quest_for_silence wrote:Nonetheless it's a rather questionable statement.
Why?

While Lostcircuits gives detailed information about how they measure the cpu power consumption (to be honest, I'm not able to fully understand the correctness of their method), I was not able to find exhaustive information about how cpus are loaded. We all know that it's a crucial point.

Moreover, as you may mind, I've also criticized in a recent past Lostcircuits for some of their results, which seemed incongruent with reference to Anandtech, Xbit Labs or SPCR's ones: you also stated that their results were a bit confusing.

Eventually there is also an historical reason: all the power figures from 2006 up to now show that Intel cpus are far more energy efficient than their AMD counterparts; just to quote our guest (SPCR), if the E7200 was an AMD cpu, prolly it would has been rated as a 45W one from AMD itself.

Therefore I mean a bit strange that an AMD 95W cpu (and a BE one!) actually shows a load consumption barely halved with reference to its declared "TDP": there's no (black) magic in electronics!

As said, currently I think some of their findings are rather "questionable" (so I mean different that straightly falses).

Despite of my above expressed thoughts, I currently buy even AMD cpus: in october I've bought a 4850e, while right now a 705e, and I am happy with them.

Regards,
Luca

P.S.: some "visual" examples of my point:

Post by **NeilBlanchard** » Thu Jun 25, 2009 6:01 am

Please people,

Do not make this personal -- we are all civil human beings, right? Also, please be aware that there are probably many things that obscure other people's intended meaning -- so don't let things get under your skin...

[/moderator's hat]

Mats · Post by **Mats** » Thu Jun 25, 2009 6:24 am

Yeah we've had this discussion before, and I agree that it's a bit confusing. But since different methods are used it's still impossible to make any conclusions when comparing reviews. I can't say lostcircuits is wrong simply because of the consequences of a different method.
In general I think many people underestimate the consequences of different motherbards/chipsets used, and at the same time follow the TDP blindly.
I'd really like to see more tests done like lostcircuits do it, but as long as they're the only one doing it, it's hard to tell what's going on. I think SPCR reviews are just as confusing in a different situation: whenever they use a Phenom II the idle power draw jumps up about 20 W compared to a K8.

The TDP is still a higher value than what we can reproduce with stress tests. It's a limit, and the real world power consumption is lower than that.
If AMD made the E7200 it would be a 45 W part IF it was sold as an -e model. Otherwise it would have a higher rating.

The X3 720 is rated for 95 W, and the only thing that tells us is that it will use that much or less power. Lostcircuits test of the i7 965 shows it using 92 W, while it does have a TDP of 130 W.

The Opteron 2381 HE is a quad core 45 nm server CPU rated for 55 W ACP. ACP is supposed to reflect real world power consumption better than TDP. It's pretty much a server variant of the 905e which have a 65 W TDP. While I do think these CPU's are binned for low power draw, it also shows what's possible. SPCR's latest Q9550s review is another a good example of how accurate the TDP is.

Keeping the TDP higher than needed is also a way for both AMD and Intel to make sure manufacturers designs motherboards that actually works in the long run, otherwise you'll end up with that AM2 - K10 situation.
Even though a failing motherboard isn't necessarily AMD's fault, it will give them bad reputation and hurt them in the long run. That's one of the reasons why you don't see more accurate desktop TDP ratings from either company.

jhhoffma · Post by **jhhoffma** » Thu Jun 25, 2009 6:37 am

quest_for_silence wrote:(to be honest, I'm not able to fully understand the correctness of their method)

Maybe that's part of the problem. First of all, TDP does not equate directly to actual power consumption. If you need a rundown of the term, check here.

Second, I don't think you are making it very clear what your point actually is. You say the results are "questionable". But how do you mean? Simply because two different websites use different methods to test, does not mean one of them is wrong if the results differ. Most websites use a standard (to them) method to compare results among their own tests, not against another website's results.

The two pictures you included show this point well. The results on the left are the same as the ones on the right, except the first picture is rounded off. Also, you do realize that these results are total system power consumption (AC), don't you? Which means that power supply, motherboard, video card, sound card, NIC, and other peripherals are all added into this number. Therefore if you are comparing the 95W TDP of the C2Q Q8200 with the 185W power draw from the charts above, then that is the source of your confusion.

quest_for_silence · Post by **quest_for_silence** » Thu Jun 25, 2009 9:27 pm

jhhoffma wrote:First of all, TDP does not equate directly to actual power consumption.

I've never state so: on the contrary, have you check how LC measures power consumption?

jhhoffma wrote:But how do you mean?

As said, first of all LC's findings seem to point out to sort of better efficiency of current AMD lineup, which contraddicts historical and technological facts; then that the 56w figure (as quoted by Mats) cannot be used straightforwardly for any comparison even for the reason that there isn't an exhaustive information about how LC loads the cpus, nor it's easy to compare it to other methods.

jhhoffma wrote:Simply because two different websites use different methods to test, does not mean one of them is wrong if the results differ.

I've never stated so: I state that their results seem to be different (in these aspects) from those obtained by every (sorry, a lof of) other ones, like Anandtech, TechReport, Xbit Labs, SPCR...

jhhoffma wrote:The results on the left are the same as the ones on the right, except the first picture is rounded off.

No, in the picture on the left there are more cpu and, above all, they point out to two different articles (you know, those things called hyperlinks...).

jhhoffma wrote:Also, you do realize that these results are total system power consumption (AC), don't you?

If you have just clicked on those hyperlinks, you really wouldn't need to put me such a question.

Anyway, yes I do.

And do you know I'm talking about relative and not absolute figures? If you don't know, now you know.

Therefore if you are comparing the 95W TDP of the C2Q Q8200 with the 185W power draw from the charts above, then that is the source of your confusion.

No, I'm comparing the relative efficiency of Intel and AMD current architectures, and then wondering how LC's findings (I've gone to read the articles quoted by Mats, do you know?) are so different from a lot of reputable sources (including our guest, SPCR).

Regards,
Luca

quest_for_silence · Post by **quest_for_silence** » Fri Jun 26, 2009 2:53 am

Mats wrote:I can't say lostcircuits is wrong simply because of the consequences of a different method.

Well, Mats: you are swedish, I'm italian, so it could occurr that the english language may lead both to some misunderstanding.
My "questionable" is not equal to "plainly wrong": have I been clear in my state?

Mats wrote:In general I think many people underestimate the consequences of different motherbards/chipsets used, and at the same time follow the TDP blindly.

Maybe: mine it's not the actual case. I doubt that those 56W pointed out by LC are plainly meaningful. On the contrary, this finding worths a lot of considerations (the questions which properly belong to the adjective questionable). Here on SPCR, with the "same" configuration a "95W" X3 720BE pulls from the wall around 150w AC at load, while in the same condition a somewhat/somehow comparable "65W" E7X00 should pull maybe 60w less than the AMD baby: I'm really wondering how much it will measure at LC labs.

Mats wrote:I'd really like to see more tests done like lostcircuits do it, but as long as they're the only one doing it, it's hard to tell what's going on.

Me too: meanwhile I will consider their precious data not extremely useful.

Mats wrote:I think SPCR reviews are just as confusing in a different situation: whenever they use a Phenom II the idle power draw jumps up about 20 W compared to a K8.

I agree with you again (and other valuable sources as TechReport seem to contraddict somehow SPCR findings on AMD K9/K10 at idle).

Mats wrote:SPCR's latest Q9550s review is another a good example of how accurate the TDP is.

Even if previous SPCR articles show clearly that what Intel declare as TDP is not what TDP typically stands for. To quote Intel's words:

...Analysis indicates that real applications are unlikely to cause the processor to consume maximum power dissipation for sustained periods of time. Intel recommends that complete thermal solution designs target the Thermal Design Power (TDP) indicated in Table 26 instead of the maximum processor power consumption. The Thermal Monitor feature is intended to help protect the processor in the unlikely event that an application exceeds the TDP recommendation for a sustained period of time...

Anyway, computerbase.de has reviewed the same Q-S cpus and their findings seem to be a bit different than those from SPCR: in the same "AC from the wall" scenery they find out a lesser power usage of about 37W (almost 50w if undervolted), with reference to standard Intel' quad cores (unfortunately the verdict is nearby the same: the newer ones cost too much compared to their elder brothers).

Mats wrote:Keeping the TDP higher than needed is also a way for both AMD and Intel to make sure manufacturers designs motherboards that actually works in the long run

As far as I know Intel does not keep the TDP higher than needed, but lower than it should be: anyway, I mean we're going too much off topic dealing with TDP data.

Regards,
Luca

jhhoffma · Post by **jhhoffma** » Fri Jun 26, 2009 8:34 am

Pardon if you took offense to my post. It wasn't meant to accuse, just point out any possible misunderstanding among the posts.

I actually didn't see that the pics were links, but you said you were giving a visual example of the difference. I didn't see any difference other than newer processors were included on one of the charts, likely because the latter article has included more CPUs - not sure why that's a problem.

I don't really care what's in the articles; I have no way of knowing if you've read or comprehended the articles in entirety. I'm simply responding to your own posts, which I found a little vague, and was attempting to account for all possibilities instead of asking a bunch of questions and having to post multiple times to respond.

I don't think anyone would argue that AMDs CPUs are more efficient than Intel's at this point. However, AMD is much closer now than they have been since the Core-series was introduced.

In the end, it seems you have some issue with LC's method, which is fine, and see a discrepancy there. I look at it as one more data point to use in my comparisons. One website can hardly account for all eventualities, and even many of the popular sites all use very similar methods, so there is little to differentiate the results. For me, I want the largest spread of conditions possible, to give a more accurate portrayal of the product as a whole, not just during gaming vs. officework or burn mode vs. idle.

Mats · Post by **Mats** » Fri Jun 26, 2009 8:55 am

quest_for_silence wrote:Here on SPCR, with the "same" configuration a "95W" X3 720BE pulls from the wall around 150w AC at load, while in the same condition a somewhat/somehow comparable "65W" E7X00 should pull maybe 60w less than the AMD baby: I'm really wondering how much it will measure at LC labs.

About 30 W, maybe even less. Their own review of the QX9650 shows that it uses 65 W. That CPU is basically two E8400 chips in one package, faster and with more cache than the E7400.

quest_for_silence · Post by **quest_for_silence** » Sun Jun 28, 2009 7:07 am

jhhoffma wrote:Pardon if you took offense to my post. It wasn't meant to accuse, just point out any possible misunderstanding among the posts.

I beg your pardon for have been upset.

jhhoffma wrote:I didn't see any difference other than newer processors were included on one of the charts, likely because the latter article has included more CPUs - not sure why that's a problem.

No problem: the pictures show no difference between themselves, they suggest what I feel to point out, that LC's findings are not (or seem to me not to be) self explanatory.

jhhoffma wrote:I don't think anyone would argue that AMDs CPUs are more efficient than Intel's at this point. However, AMD is much closer now than they have been since the Core-series was introduced.

I think that stating

Mats wrote:The difference in TDP is usually not reflected in real world use, since the TDP is a limit and not an absolute value... Lostcircuits.com measured 56 W power draw (CPU only) during load with the 720 BE

is misleading for several reasons (ranging from as those watts are comprehensive of VRM leakages, to how they load the cpu, and last but not least because here power consumption is often taken out of "AC from the wall" scenery), all with reference to the overall complexity (LC's article on i7 power plays gives some good point).
Even the preamble is misleading since AMD's TDP is a limit, while Intel's one no.
As you noted I have some issue with LC's method and to the possibility of use their findings to wisely choose among different AMD cpus.

In the end, those 56w may be a very interesting value if we deal according to SPCR rules but, if we change the rules (as in LC's lab), not definitely so.

jhhoffma wrote:In the end, it seems you have some issue with LC's method, which is fine, and see a discrepancy there. I look at it as one more data point to use in my comparisons. One website can hardly account for all eventualities, and even many of the popular sites all use very similar methods, so there is little to differentiate the results. For me, I want the largest spread of conditions possible, to give a more accurate portrayal of the product as a whole, not just during gaming vs. officework or burn mode vs. idle.

If you know how to deal with those measures, please tell me: here I see that a "45w TDP" 4850e showed almost 40Â°C (fourthy!) less than a "95w TDP" 720BE on the same cpu's power circuitry, and that a "65w TDP" Intel E7200 IGP system sucked almost 60w (sixty!) less than a "measured 56w" 720BE IGP system doing the same things (but - as already said - I bet even one million euros that E7200's power consumption isn't that negative, even in LC's lab, nor near to a very few w).

Regards,
Luca

Refreshment · Post by **Refreshment** » Thu Jul 02, 2009 4:59 pm

Greetings.

Have a question here.
Has anyone bought a Phenom 2 X3 705e from Amazon?
Ask this because they say they have stock, yet there aren't any user reviews there.

Also slightly off topic. Does the phenom 2 x2 draws a lot more power than an athlon 64 5000+ BE? I'm interested in the Phenom X2 550 but the power consumption has me worried. Thats why im considering the 705e.

Thanks for your time.

Nighthound · Post by **Nighthound** » Sun Jul 05, 2009 7:34 pm

I just bought the 705e X3 for an HTPC build. I have an Antec Fusion case which I managed to snag for $55 on close out. I am wondering if it is a foolish idea to try to cool the 705e passively. Apparently the Fusion is very nice for passive cooling because of the placement of the 120mm fan right next to the CPU. I have a Mini Ninja ready to go. Thoughts?

theycallmebruce · Post by **theycallmebruce** » Sun Jul 05, 2009 9:26 pm

I'd say try it, and let us know

Just keep an eye on your temperatures. The machine will shut itself down automatically if it gets too hot anyway.

My guess is you will probably get away with it at idle, but you will probably need a fan blowing on it at load.

JamieG · Post by **JamieG** » Sun Jul 05, 2009 11:24 pm

Nighthound wrote:I just bought the 705e X3 for an HTPC build. I have an Antec Fusion case which I managed to snag for $55 on close out. I am wondering if it is a foolish idea to try to cool the 705e passively. Apparently the Fusion is very nice for passive cooling because of the placement of the 120mm fan right next to the CPU. I have a Mini Ninja ready to go. Thoughts?

While the AMD 4850e that I'm using a passive Ninja Mini on in my NSK2480 (the non-VFD version of the Fusion from memory) is probably nowhere near the heat levels of a 705e X3, I've encountered no problems with heat issues with only a single 800rpm fan at 7V (approx 500rpm or so) as an exhaust right next to the Ninja Mini. This is with a passive Asus 3450 as well in that case, which heats it up a bit. I've run Prime95 on the CPU for a while with no errors, so you probably should be fine with the 703e at normal CPU loads in a HTPC. (No guarantees though

).

Hope that provides you some comfort.

Nighthound · Post by **Nighthound** » Mon Jul 06, 2009 6:59 pm

OK, thanks. I'll give it a go and see what happens. I will likely replace the Antec exhaust fan with something a little more robust and quieter. Since this is 45nm part I am thinking I will be OK with the Ninja Mini.

This is probably the wrong forum but what is a good lightweight utility for everyday monitoring of system temps (I don't need fan control). I thought AMD would have something but I can't see that they do...