Athlon x2 vs Athlon x4 power consumption?

[Ozymandias]

I have been looking at the latest articles on the Athlon II X4 and the Athlon II X2, and it actually looks like the X4 uses less power at idle??? I don't understand how this is possible. The test rig looks very similar in both reviews.

Am I missing something here?

Also which brand is better when power consumption is a concern? I am looking for a low power multi-core CPU that costs less than $120.

 

[lyssword]

1. Maybe new x4 had tweaked power management
2. x4's are running 2.6 and 2.8 vs 3.0 and 3.1 ghz x2
3. If only 1 core works when idle, x4 uses only .5mb cache per core where x2 uses 1mb per core (assuming amd can switch off l2 cache)

 

[alyarb]

The Q8200 or 605e is your best bet for low power. anand's problem, and most everyone's problem with power consumption tests is that they test the entire system's power consumption on the AC side of the PSU. hopefully sooner or later they will wake up and realize that nobody cares what the total system consumption is because holistically, the system in its entirety doesn't apply to the reader. they are reviewing and testing ONE component and no one is going to have the same exact machine, so why let other components (particularly a large, inefficiently loaded PSU) in the machine corrupt their power data?
edit: jesus, anand's review doesn't even mention what power supply they use for the testing. So they put all this "effort" into a low-power CPU article and don't do any objective consumption testing, they just hook it up to their "watts up?" meter and completely neglect to mention the power supply they're using. this is exactly why we can't compare power consumption from article to article. doing it on the AC side of an unknown PSU is a complete waste of time and they may as well not have even performed the test. depending on how efficient or inefficient it is at a given load that could be a 10-15 watt error, and on that note i would advise you to buy an appropriately-sized power supply for your low-power build.

there is one tech site who at least has a clue as to what they're doing. at xbitlabs, they take an ammeter and put it right on the ATX12V cpu line AND they show the total system consumption so we at least can glimpse what's going on.

http://www.xbitlabs.com/articl...ii-x4-630_9.html#sect0

we see the Q8200 loads at 53 watts and the Athlon II X4 620 at 86.5 watts. I'd imagine the 605e to load somewhere around 40-45 watts. It'll be a touch slower than the Q8200, but who cares.

unfortunately you aren't going to be getting either of these for $120. the Athlon 605e is going to be in high demand when it comes out and a couple sites are taking preorders for $160. Q8200 is $140 at amazon with free shipping. Plan on $55-60 for the motherboard regardless of which CPU you choose.

 

[Lemon law]

I have to disagree with the alyarb point here. If the cpu is going to consume 95+ watts of power, it going to be all dissipated as heat. Since the cpu is fairly small, all that heat must be radiated by the heat sink and fan.

Just color me very soured by 95+ watt processors, been there done that, and even at stock speeds you are behind the 8 ball.

 

[Ozymandias]

When is the 605e coming out? Looks like there might be some low power x2 and x3 products coming out alongside it. What would you recommend if I didn't want 4 cores?

 

[alyarb]

er whoops.


p.s. i'm drunk.

 

[alyarb]

Originally posted by: alyarb

Originally posted by: Lemon law
I have to disagree with the alyarb point here. If the cpu is going to consume 95+ watts of power, it going to be all dissipated as heat. Since the cpu is fairly small, all that heat must be radiated by the heat sink and fan.

Just color me very soured by 95+ watt processors, been there done that, and even at stock speeds you are behind the 8 ball.

i don't think thermal output is what's being discussed here.

Originally posted by: Ozymandias
When is the 605e coming out? Looks like there might be some low power x2 and x3 products coming out alongside it. What would you recommend if I didn't want 4 cores?

there are low power tri-cores coming out also.

 

[Schmide]

Some good just the processor power readings.

Lost Circuits Athlon II x4

Yorkfield (non-extreme)/Phenom II (non-extreme)/Athlon II all generally within 10w load

 

[Idontcare]

Originally posted by: Ozymandias
I have been looking at the latest articles on the Athlon II X4 and the Athlon II X2, and it actually looks like the X4 uses less power at idle??? I don't understand how this is possible. The test rig looks very similar in both reviews.

Am I missing something here?

Also which brand is better when power consumption is a concern? I am looking for a low power multi-core CPU that costs less than $120.

Power consumption will be clockspeed and voltage dependant in addition to chip-to-chip differences in the leakage/capacitance attributes that effect both static and dynamic leakage.

For an example that showcases how chip-to-chip variability really makes an impact just look at the idle power consumption in Anand's Athlon II X4 review and notice that their 2.8GHz 630 system actually idles some 10W (9%!) lower in power-consumption than the lower clocked 2.6GHz 620 system.

What we would expect is that for any two random CPU samples from both 620 and 630 SKU's the 620's would on average use less power than the 630's but there will be those atypical cases where the higher clocked chip actually consumes less power.

As to why the X4's idle at less power-consumption than the X2's my bet is that the X2's idle at a higher voltage and higher GHz (CnQ involved here) than the X4's.

 

[alyarb]

dude, they're measuring AC power, and the efficiency of the power supply increases as the load on the unit increases.


you can't just throw together a 200W machine full of low power components, hook it up to a 500W power supply and test it on the AC side and expect any of your data to be worth a damn because the 500 W unit will reach its peak efficiency when it's loaded around 400W, and the efficiency steadily increases as you approach this number.

the AC-DC conversion may very well have been more efficient with the higher-clocked CPU, but since they don't even mention what unit they're doing the conversion on i can't accept any of the data because there's nothing consistent to compare it to.

if you test only the DC going into the CPUs you will find that the 2.6 GHz CPU idles with less power than the 2.8 GHz CPU, and that would be considered useful data.

 

[Soulkeeper]

Anyone done any numbers on the difference between wall drawl from going between say a 400Watt psu and a 600Watt psu on the same hardware, my guess is for most intents it's nearly negligible.
i'm too lazy to unplug/replug everything myself and do the test

 

[Idontcare]

Originally posted by: alyarb
dude, they're measuring AC power, and the efficiency of the power supply increases as the load on the unit increases.


you can't just throw together a 200W machine full of low power components, hook it up to a 500W power supply and test it on the AC side and expect any of your data to be worth a damn because the 500 W unit will reach its peak efficiency when it's loaded around 400W, and the efficiency steadily increases as you approach this number.

the AC-DC conversion may very well have been more efficient with the higher-clocked CPU, but since they don't even mention what unit they're doing the conversion on i can't accept any of the data because there's nothing consistent to compare it to.

if you test only the DC going into the CPUs you will find that the 2.6 GHz CPU idles with less power than the 2.8 GHz CPU, and that would be considered useful data.

The efficiency reality is such that from a CPU perspective it overestimates just how low the power consumption could be with a higher efficiency PSU.

So yes to be sure that 100W idle number would be all the more lower had the reviewer used a higher efficiency PSU at that point on the power-consumption curve. And to be sure the 110W number would be lower as well.

But no amount of optimal PSU efficiency matching is going to reverse the actual rankings, the system consuming 110W is not suddenly going to consume even less power than the system consuming 100W if both systems are outfitted with a higher efficiency PSU. (this is where your post gets wonky because you really have to start constructing boundary conditions to make your case valid - i.e. "well maybe they used different PSU", etc, those are assumptions you are now making so as to make your conclusions valid...you can do it but it does little to convince me)

If you are really interested in isolating just the CPU's power-consumption at the socket (which is not its footprint on your electricity bill) I just go to lostcircuits as they all the work necessary to isolate it.

Windows Power Consumption (Idle), CnQ Enabled

But I disagree with you that measuring power-consumption at the wall is useless or without value. I measure my system's power-consumption at the wall (kill-a-watt) and it tells me information I have routinely found to be helpful and useful. Yes the PSU's subtly changing efficiency over the power-range of a rigs idle to load parameters does result in an overestimation of the power-consumption values of the components downstream of the PSU but that is what error-bars and boundary conditions are for.

The thing with Anand's data, any reviewers who tally their power-consumption like that, is the data lends itself readily to analysis that need precision versus accuracy.

http://en.wikipedia.org/wiki/Accuracy_and_precision

Provided the same PSU is used (or just one with a similar efficiency power curve) the resultant data will have high precision, but yes as you say it will lack accuracy (that is the efficiency of the PSU)...precision is great for taking delta's. If you wanted to know how much more power-consumption a i7 975 takes over an i7 920 then you'd want higher precision, not accuracy, in your data so that your delta value has lower error-bars.

(BTW you'll note you can't measure the power-consumption of Nehalem-derived CPUs such as Lynnfield or Bloomfield at the socket...you have to go the "at the wall" route)