AMD FX: Energy Efficiency Compared To Eight Other CPUs

[pschmid]

We've already seen AMD's Bulldozer architecture come up short in the performance benchmarks. However, the company also claims it made important improvements to power consumption. Can FX-8150 at least score some points in the energy efficiency department?

AMD FX: Energy Efficiency Compared To Eight Other CPUs : Read more

 

[compton]

The low idle and load power consumption numbers of the SB K series are why I love them so much. Less power = less heat and noise, and SB is certainly worth it for me. BD, on the other hand, is just a strange bird. Someone out there could probably find a way to leverage it successfully, and that one person is going to be very happy. Maybe Bulldozer makes a lot more sense in its server configurations -- but I really wish AMD had just given the Phenom II a slight dust-off and die shrink. Everyone was pulling for AMD to do something great with BD, and the efficiency results are just abysmal. If you got great performance, but dis-proportionally high power consumption, that would be okay as well. With BD, you get the worst of both world, and not much of a saving grace. Perhaps Trinity will do something with this albatross that is BD and make it respectable, because the efficiency comparison is embarrasing.

 

[de5_Roy]

thank you tom's. this kind of article (performance-efficiency analysis) is one of my favorites. i've been waiting eagerly for an article like this from reviewer sites, tom's beat everyone else.
the benchmarks with real world softwares(and not some specialized highly threaded synthetic benchmark that gives biased results) are the ones that matter to me. i use some of the softwares occassionally (blender), some more frequently (winrar, 7zip, lame encoder) and this article helped me a lot when i choose my next pc.
did you guys see the ridiculous tdp number on cpu-z screenshot of fx8150? 223 w what the !@#$. i wonder which one got it wrong, amd or cpu-z.
amd-fans-in-denial can argue as much as they want, but the reality didn't change. the efficiency numbers pretty much mirrored the bd review - bd isnt power efficient. even the ph ii 980 - the most power hungry of phenoms is more power efficient than fx 8150. and people who don't care about power consumption should care about the cooling and maintenance bd would need along with a power hungry high performance gfx card. imagine running an air-cooled fx 8150 @ 4.7 ghz with nvidia gtx 580 or radeon hd 6990.
i can use any kind of acronyms like 'lol' or 'lmao' on bd's laughable power efficiency(even lynnfield beat it!) and performance but i am really sad and disappointed.
if amd can't compete with intel, intel will keep selling their cpu at a high(and higher) price - avg users like me will be the loser.

 

[tacoslave]

everytime i read a BD article i die a little inside. Plus what we all knew would happen already started Intel already raised the K series prices a couple bucks.

 

[compton]

Geez, the 2700K is creeping up on $400. Thanks a lot AMD. You're off my Christmas list.

 

[dragonsqrrl]

[citation][nom]compton[/nom]Geez, the 2700K is creeping up on $400. Thanks a lot AMD. You're off my Christmas list.[/citation]
Ya, the MSRP is $332, but the price on newegg is $370. Even for a brand new processor that's a huge premium over MSRP. It'll stabilize to the $330 price range eventually, but this initial price hike is no doubt related to the Bulldozer launch.

 

[de5_Roy]

@compton: phenom might get a die shrink with the llano upgrade. according to the latest trinity leak, llano's new 'husky' core will feature a phenom ii class cpu with amd 6xxx class gpu. this is just a rumor though.

 

[soccerdocks]

Thank you very much for including Matlab in the benchmarks. Its a really informative benchmark for those in engineering.

 

[Guest]

"Everyone was pulling for AMD to do something great with BD, and the efficiency results are just abysmal."

Not really, for the most time everyone was aware that BD was not going to be a SB killer, AMD themselves had hinted at it, then their PR department (propaganda office I would say) started pumping up the hype.

 

[amk-aka-Phantom]

And this is exactly why AMD fanboys should STFU about Bulldozer being an "excellent server CPU". You don't want high power consumption on a server.

 

[amk-aka-Phantom]

[citation][nom]dragonsqrrl[/nom]Ya, the MSRP is $332, but the price on newegg is $370. Even for a brand new processor that's a huge premium over MSRP. It'll stabilize to the $330 price range eventually, but this initial price hike is no doubt related to the Bulldozer launch.[/citation]

2700K is BS... 100MHz extra is definitely not worth it. 2600K and 2500K remain best bang for buck right now.

 

[silverblue]

[citation][nom]de5_roy[/nom]@compton: phenom might get a die shrink with the llano upgrade. according to the latest trinity leak, llano's new 'husky' core will feature a phenom ii class cpu with amd 6xxx class gpu. this is just a rumor though.[/citation]
But that's what Llano actually is - four Husky cores and a 6xxx GPU. I'm not sure what you're getting at.

[citation][nom]amk-aka-phantom[/nom]And this is exactly why AMD fanboys should STFU about Bulldozer being an "excellent server CPU". You don't want high power consumption on a server.[/citation]

And you don't know if AMD purposefully binned better quality chips for the server market. I admit, it's an assumption, but a reasonable one, wouldn't you say? Let's wait for the reviews and see.

I'm liking the article, but considering all the comments about the ASUS board being power hungry, plus the touted motherboard round up, I'm finding it strange that you'd use the CVF again. Still, good article, thanks very much.

 

[4745454b]

Company may say its coming, but I don't think there is much they can do about IPC. Pipeline too long, latencies to the cache are too high, and it might be to light on the floating point power. Only a chip redesign will fix those issues.

 

[Martell1977]

And yet the 8120 and 8150 are both sold out or out of stock on Newegg. I guess there are alot of people who don't care for power and effeciency as long as it has the AMD tag.

Interesting article I'll stick with my I7-950 for a while. Was on a P4 and skipped over 3-4 proc generations before needed to upgrade. Hoping to get the same longevity from my 950 as well.

 

[The Greater Good]

We'd argue that no desktop machine runs at full throttle 24x7, making idle behavior incredibly important, too.

As far as the average user goes, yes. This is correct. I run BOINC so my desktop is always at 100% CPU load 24/7 365; 4 cores/8 threads always maxed.

 

[de5_Roy]

[quotemsg=9521795,12,267802]But that's what Llano actually is - four Husky cores and a 6xxx GPU. I'm not sure what you're getting at.



And you don't know if AMD purposefully binned better quality chips for the server market. I admit, it's an assumption, but a reasonable one, wouldn't you say? Let's wait for the reviews and see.

I'm liking the article, but considering all the comments about the ASUS board being power hungry, plus the touted motherboard round up, I'm finding it strange that you'd use the CVF again. Still, good article, thanks very much. [/quotemsg]
you're right. i mixed up amd's naming schemes and thought llano was made of athlon class 'stars' cores with radeon 'sumo' 6xxx igp. i missed that the whole thing was called a husky core.

kitguru did a power consumption comparison with a gigabyte 990fx board. they also compared how much power fx 8150 and others use at the same oc speed. the results are not different, it was the fx 8150 that was power hungry.

 

[silverblue]

http://www.hardwareheaven.com/reviews/1285/pg15/amd-fx-8150-black-edition-8-core-processor-vs-core-i7-2600k-review-power-temps-overclocking.html

I would have to agree on that basis; 68 more watts for the 8150 under full load as compared to the 2600K at default clocks. I was hoping for a bit of good news on that front, but sadly it's not to be. I'd like to see more as regards disabling every even numbered core (and what it does to power), however buying an octo-core to use it as a quad-core is somewhat, well, counter-productive.

On another topic, I find it amusing that AMD went from an anaemic 2MB L3 cache in Phenom to 6MB with Phenom II, yet with no L3 cache, Llano is noticeably faster than Phenom - is it memory related? Just ignore SYSMark 2007 (hell, if you compare the 8150 with the Core2 Extreme QX9770, the latter wins...).

http://www.anandtech.com/bench/Product/403?vs=21

 

[Quaddro]

FAIL..

 

[amk-aka-Phantom]

[quotemsg=9521803,17,267802]Just ignore SYSMark 2007 (hell, if you compare the 8150 with the Core2 Extreme QX9770, the latter wins...). http://www.anandtech.com/bench/Product/403?vs=21[/quotemsg]

:lol: AMD fanboys are so hilarious. Just ignore all benchmarks where Bulldozer loses... Core 2 Quad Extreme QX9770 was an extremely powerful CPU targeted at multimedia production; I can easily believe it beating Bulldozer in certain tasks. It is not an error.

And you don't know if AMD purposefully binned better quality chips for the server market. I admit, it's an assumption, but a reasonable one, wouldn't you say? Let's wait for the reviews and see.

Ummm.... wtf do you mean "AMD purposefully binned better quality chips for the server market"? So you mean, the Bulldozers that will make it to the servers will SUDDENLY have much lower power draw? Yeah, right.

There's a separate Bulldozer server line which came before FX series. FX series was NOT meant to be a server chip. It was made for gaming and failed in that area. The only area where it might be useful is the workstation. BD scores top benchmarks in Photoshop - it will be great for a light Photoshop user (a heavy user who earns money with Photoshop will just get a Xeon). It also falls in between 2500/K and 2600/K in some other multimedia creation software, both price- and performance-wise. That's where these chips must logically end up.

Well, at least good to see that the fanboys have settled down with their "patches" and "fixes" that are allegedly supposed to set BD's performance straight...

 

[iam2thecrowe]

didn't we already know it was a power hog and no faster overall than a phenom x6? I dont believe the server chip will be a lot better. It might be a bit more power conservative due to lower clock speeds, but it will still perform no better than a phenom x6. So you pay more and get less...oh wait, you get the adjustable TDP cap crap, which just lowers its performance in return of power savings.

 

[vitornob]

In the single-thread page, wouldn't be more precise to separate the generations?
It would be like this:

- Fx-8150 uses about 40% more power, measured in watt-hours than 1st core gen.

- FX-8150 uses about 164% more power, measured in watt-hours than 2nd core gen.

- FX-8150 uses about 11% more power, measured in watt-hours than AMD last generation.

 

[alidan]

where i live its 11 cents a watt
100 watts costs 96.36$ to sustain over a year

lets say this, every build lasts what, 4-5 years, lets go 5 years on this

75 watt = 72.27 = 5 year = 361$
100 watt = 96.36 = 5 year = 481$
125 watt = 120.45 = 5 year = 602$
just some numbers

 

[Onus]

This goes beyond sad, beyond pathetic, into the realm of WTF was AMD thinking when they released this chip? IF there is a niche for whom it does very well (perhaps [some] low-density server, or a specific [type of] application), then AMD needed to form a partnership with a system integrator and target that specific market with vertical solutions; this general release does nothing but make the company look incompetent. Release [a lot of] engineering samples, and say "this is what we're working on, but it clearly isn't ready;" people could focus on the few areas where it shows promise; but now, it just screams FAILURE to the whole world. I see a lot of reviewers trying to be polite, but there is no escaping this CPU is a misbegotten failure. The emperor has no clothes.
The only reason I don't write off my 990FX rig as an experiment that failed is, quite frankly, the 970BE on it now meets my needs quite nicely.

 

[de5_Roy]

i wonder if the chipset and cpu combo had anything to do with the power consumption.
may be amd should have done an intel and built a backwards incompatible chipset focussing on performance and efficiency improvements from ground up. backwards compatibility might have brought some performance loss along with power drain. all the am2, am3 cpus seemed too power hungry(compared to intel's) for the tasks they do.
amd built brazos and llano platform which don't use am3 and both seem more power efficient (they're slower than phenom ii) than am3+.

 

[vitornob]

[citation][nom]alidan[/nom]where i live its 11 cents a watt100 watts costs 96.36$ to sustain over a yearlets say this, every build lasts what, 4-5 years, lets go 5 years on this75 watt = 72.27 = 5 year = 361$100 watt = 96.36 = 5 year = 481$125 watt = 120.45 = 5 year = 602$just some numbers[/citation]

Let me made some minor corrections.
1 kWh (kilowatt per hour) costs 11 cents.
Using multi-thread applications for 6 hours in FX-8150. To have the job done we would use:

AMD Wh -> 264.0 Wh
2600k Wh->131.4 Wh
365 days/yr means
Amd kWh/yr -> 96,36 kWh -> Psu 80% Eff -> 120,45 kWh
2600k kWh/yr -> 48,05 kWh -> Psu 80% Eff -> 60,06 kWh

Difference btw, in about 60 kWh a yr -> $6.60
Please correct me if I'm wrong in any step