athlon ii x2 250 vs athlon x2 5050e power consumption review... disappointing.

[hans007]

So I bought an athlon ii x2 250 to try out on my board. I also have a 5050e.

anyhow

x2 250
2 x 1mb cache
3ghz
45nm
65W tdp


5050e
2 x 512kb cache
2.6 ghz
65nm
45W tdp.

I had expected the x2 250 to at least win at idle, its 45nm for god sakes.

Anyhow the test setup.
J&W minix-780G motherboard ITX
2 x 2GB ddr2-800 crucial CL6 sodimm
Western digital WD15EADS green drive
nec slim dvd-rw (i'm sure this draws some amount of power idle)
inwin-bm643's 120W power supply (reasonably efficient, but not as good as an earthwatts 380 by a watt at idle, previously tried this with an EA380 that i no longer have with the 5050e).

Anyhow here are the results for idle and load.

x2 250 at idle runs at 800mhz. voltage 1.018 v according to CPU-Z. The 5050e runs at 1000mhz .936V at idle.

Wattage on kill-a-watt idle for the 5050e=38 watts.
Wattage on kill-a-watt idle for the x 250=49 watts.

Orthos load.
x2-250 = 105W
5050e= 91W

Overall really disappointing for the x2 250 if you are trying to build a super low power system (such as mine).


I think its a great chip otherwise, but for this use maybe not. I guess the next chip to get will be the 600e/605e then.

 

[Fox5]

Have you considered undervolting/underclocking the 250? Even at the same mhz, it should vastly outperform the 5050e.

 

[LoneNinja]

Doesn't surprise me, there should be 45w variants of Athlon II out in the future, the 250 was released for performance because that's what AMD needed.

 

[Lonyo]

Yeah, when one is 45w TDP and the other is 65W TDP, I would expect the 45w one to win, even if it is 65nm vs 45nm.
Smaller process doesn't automatically mean lower power consumption, and when one is a specific low power CPU and the other is a regular one, then it's no surprise to me that the 45nm higher TDP chip uses more power. The fact that it only uses 11/14w more power is likely in part due to the smaller process, rather than the 20w more the TDP might imply.

 

[richierich1212]

yes you can undervolt the 250 pretty well. Try it and let us know.

 

[Jovec]

Something is up with the 250 though. AT's 250/550 review shows the 250 at 5watts lower at idle but 6 watts higher under load. Given the minor clock speed bump and L3 of the 550, I'd expect the 550 to higher idle and load power consumption.

I imagine it has something to do with the 250 having a new core design, and possibly making some tradeoffs in power usage for a smaller die/higher yields.

 

[hans007]

i do not think the 65 vs 45 W tdp should matter at idle.

i mean a 65 W brisbane say 6000+ g2 core still idles lower than the 250.

The full load power use make sense, though only barely. I mean intel puts out e7200s which are "65W" and they use barely 30W at orthos load.

and at idle when cool and quiet is on the athlon II x2 250 is running at lower clocks than the 5050e. I dont think tdp rating has anything to do with idle wattage on a lower clockspeed.

 

[cusideabelincoln]

It could be the fact the older Athlon X2 architecture has been tweaked so much and the 65nm process is much more mature.

But in AT's review the 45nm Athlon X2s are on par with Intel's 45nm chips (although I'm not sure how much platform, i.e. motherboard, has to do with those numbers).

 

[imported_Scoop]

Originally posted by: cusideabelincoln
It could be the fact the older Athlon X2 architecture has been tweaked so much and the 65nm process is much more mature.

But in AT's review the 45nm Athlon X2s are on par with Intel's 45nm chips (although I'm not sure how much platform, i.e. motherboard, has to do with those numbers).

Don't read AT reviews about power consumption, they get the weirdest results every time.

AMD has nothing on the Intel chips when it comes to power consumption and energy efficiency.

http://www.silentpcreview.com/article949-page3.html

http://www.silentpcreview.com/article949-page4.html

 

[Gikaseixas]

the X2 5050e was designed to perform well while saving at the same time while the X2 250 is a different kind of animal.

 

[hans007]

Originally posted by: Gikaseixas
the X2 5050e was designed to perform well while saving at the same time while the X2 250 is a different kind of animal.

Well thats not entirely true.


At load sure. but any brisbane g2 core will be as low or nearly as low as a 5050e at idle. I had a 6000+ 3.1ghz and it was maybe 1 watt more than a 5050e .

I think the idle on the x2 250 is just not as good

 

[Rhoxed]

does cashe have anything to do with tdp?

as the 250 has twice as much?

 

[hans007]

Originally posted by: Rhoxed
does cashe have anything to do with tdp?

as the 250 has twice as much?

it does at full load. obviously the cache has to be turned on.

I suppose it would also be on at idle, but I dont think its 11 watts worth at 800mhz. Its really weird honestly.

I mean I still am going to keep the 250 over the 5050e, since its significantly faster, but the idle readings are pretty odd on it.

Even in a lot of independant reviews, it seems like the 250 idles higher than some quad core phenoms which makes no sense seeing as it has a muchs maller die.

 

[deputc26]

As mentioned before, undervolt both procs to their min stable vid and then compare. I think that will make the 250 look a LOT better. If you're looking for a min power system you should be undervolting everything anyway.

 

[Gikaseixas]

Originally posted by: hans007

Originally posted by: Gikaseixas
the X2 5050e was designed to perform well while saving at the same time while the X2 250 is a different kind of animal.

Well thats not entirely true.


At load sure. but any brisbane g2 core will be as low or nearly as low as a 5050e at idle. I had a 6000+ 3.1ghz and it was maybe 1 watt more than a 5050e .

I think the idle on the x2 250 is just not as good

ofcourse any G2 will be as low because the 5050e itself is a G2 revision but the similarities end right there.

any processor that ends with a "e" was design with power savings in mind.

example 1

X2 5050e 65nm, 45 watts, 2.6ghz and costs 62.99 @ newegg

X2 5200 65nm, 65 watts, 2.7ghz and costs 59.99 @ newegg

example 2

X3 710 45nm, 95 watts, 2.6ghz and costs 99 @ newegg

X3 705e 45nm, 65 watts, 2.5ghz and costs 129 @ newegg


My point is that we cannot expect the 250 with a larger cache to be better than the 5050e at power savings even at idle because they have different features/design.

You also pay a premium for those cpus, a good example is the Q9550S

 

[jrichrds]

Originally posted by: Gikaseixasofcourse any G2 will be as low because the 5050e itself is a G2 revision but the similarities end right there.

any processor that ends with a "e" was design with power savings in mind.

The BE-2XX0 and the "e" revisions are just cherry-picked Brisbanes set to run at lower stock voltage. Nothing else special about them. The advantage vs. regular Brisbanes is that they typically have more undervolt/overclock potential.

 

[hans007]

Originally posted by: deputc26
As mentioned before, undervolt both procs to their min stable vid and then compare. I think that will make the 250 look a LOT better. If you're looking for a min power system you should be undervolting everything anyway.

I am going to do that soon.

I am really not too concerned with the load voltage, sinec my system is @ the idle state 98% of the time.

Just I am not really sure how much lower I can go than 1.016 volts @ 800mhz. It seems that silent PC review was able to go down to 1.175 v from 1.325 volts for vcore (whcih makes some corresponding drop in idle voltage) apparently @ 3ghz. I'm hoping to do better than that @ 2.4 ghz or so.

At first I didn't think I could do this on an ITX board, but this one seems to have full voltage control of every single voltage and setting on the motherboard.

 

[deputc26]

At first I didn't think I could do this on an ITX board, but this one seems to have full voltage control of every single voltage and setting on the motherboard.

Awesome! At low clocks voltage can go extremely low, in undervolting lappies I've found most procs can hold the lowest available vid to 2Ghz which is .95 on my T8100 and .875 on my friends P8600.

I would love it if you take the proc to min Vid at 800mhz and let us know what it is so we can get an accurate comparison of Intel v. AMD at very low clockspeeds.

 

[richierich1212]

I can lower my 550 to .928v @ 800MHz using PhenomMsrTweakr. You should be able to lower your 250 as well.

 

[deputc26]

Originally posted by: richierich1212
I can lower my 550 to .928v @ 800MHz using PhenomMsrTweakr. You should be able to lower your 250 as well.

I would have expected lower for such a low freq. Looks like my theory that SOI is more efficient at low clocks and HKMG is more efficient at high clocks doesn't hold water. I initially was surprised by how good AMD's 45nm process was but apparently it's still not as good as Intel's (but close!)

 

[richierich1212]

Well how low of a voltage can an Intel quad-core processor go? The 550 is quad-core based.

edit:

Looks like the Q9550s vid range is 0.85V ? 1.3625V. So yeah I guess the Intel's process is still a lot better.