Do AMD cpus at least give a smoother desktop experience w/more cores?

[Yuriman]

How is performance per watt equal when one CPU is faster than the other ??

Just use pass 2 results from my graph in order to simplify it.

A8-7600 performance = 8,12fps
A8-7600 energy consumption = 80 Wh

A8-7600 Efficiency = Performance / Energy consumption = 8,12/80 = 0.1015

Core i3 3225 performance = 6,63fps
A8-7600 energy consumption = 80 Wh

Core i3 3225 Efficiency = Performance / Energy consumption = 6,63/80 = 0.0828

A8-7600 Efficiency is higher than the Core i3 3225.

Wh = Energy consumed to finish the job/task.

Wh by it self IS NOT Efficiency.

0.1015 what? 0.0828 what?

Look at your units.

(8.12 frames / 1 second) / (80 watts * 3600 seconds) ->
8.12 frames / (80 watts * 3600 seconds²) ->
0.000028 frames / (watts*seconds²)

Frames / watts*seconds² is not a unit of anything real. You can't compare a rate of work to a unit of total energy used to complete a task. If you want to find out how efficient something is, you need to have your units in work done over energy used.

You COULD do 8.12 frames per second / 80 watts. Watt-hours are not watts. Watts are energy consumption. Watt-hours are energy consumed.

 

[AtenRa]

0.1015 what? 0.0828 what?

Look at your units.

(8.12 frames / 1 second) / (80 watts * 3600 seconds) ->
8.12 frames / (80 watts * 3600 seconds²) ->
0.000028 frames / (watts*seconds²)

Frames / watts*seconds² is not a unit of anything real. You can't compare a rate of work to a unit of total energy used to complete a task. If you want to find out how efficient something is, you need to have your units in work done over energy used.

Wh is not 80watts * 3600 seconds

You take Wh as a unit of energy.

80 Wh in the results on the graph = energy consumed to finish the job/task. The time it took to finish the job is not the same nor the Power in Watts used by the two CPUs.

edit: Efficiency is a dimensionless Quantity, it is just a number. The higher the number the higher the Efficiency.

 

[Yuriman]

Wh is not 80watts * 3600 seconds

I'm sorry, but you're wrong. 80 watt hours is exactly 80 watts * 3600 seconds.

http://www2.enphase.com/myenlighten-help/tip/what-is-the-difference-between-a-watt-and-a-watt-hour/

A watt-hour (Wh) is a unit of energy; it’s a way to measure the amount of work performed or generated.

In a nutshell, watt-hours measure amounts of energy for a specific period of time, and watts measure rates of power at a moment in time.

You can't take the rate of work done, and divide it by work done.

EDIT: At least, you won't get anything that makes sense to anyone.

http://whatis.techtarget.com/definition/watt-hour-Wh

The watt-hour (symbolized Wh) is a unit of energy equivalent to one watt (1 W) of power expended for one hour (1 h) of time.

An energy expenditure of 1 Wh represents 3600 joules (3.600 x 103 J).

A watt is equal to 1 joule per second.

https://en.wikipedia.org/wiki/Kilowatt_hour

The kilowatt hour (symbol kWh, kW·h, or kW h) is a derived unit of energy equal to 3.6 megajoules. If the energy is being transmitted or used at a constant rate (power) over a period of time, the total energy in kilowatt-hours is the product of the power in kilowatts and the time in hours. The kilowatt-hour is commonly used as a billing unit for energy delivered to consumers by electric utilities.

You can't divide the rate at which you do something by the amount you'll be billed by your electric company to do it.

 

[Essence_of_War]

How is performance per watt equal when one CPU is faster than the other ??

Just use pass 2 results from my graph in order to simplify it.

A8-7600 performance = 8,12fps
A8-7600 energy consumption = 80 Wh

A8-7600 Efficiency = Performance / Energy consumption = 8,12/80 = 0.1015

Core i3 3225 performance = 6,63fps
A8-7600 energy consumption = 80 Wh

Core i3 3225 Efficiency = Performance / Energy consumption = 6,63/80 = 0.0828

A8-7600 Efficiency is higher than the Core i3 3225.

Wh = Energy consumed to finish the job/task.

Wh by it self IS NOT Efficiency.

Total energy required to complete a job is not an unreasonable definition of efficiency. It's also not the only definition of efficiency. The reason your car analogy was confusing is because miles per gallon, or kilometers per litre (for the civilized among us) are very common measures of fuel efficiency for motor vehicles and they do not involve the speed of the car at all. In fact, most cars that I've see hit

I think the confusion in this thread arose largely from the fact that you did not very clearly explain what you meant by either "efficiency", or "performance".

Let me try to be very explicit. When you say "efficiency", you mean the average FPS reading divided by the total amount of energy used to perform the job, is that correct? The resulting units are:

 (Frames / second) / (watt* hr) -> frames / (Joule * hour)

This seems kind of weird to me.

When I think of encoding efficiency I think about something like (in english terms) "on average, how many frames/second of encoding speed does each watt of power consumption get me", which if you express it in terms of units:

(total frames / encode time) / ( (1 / encode time) * integral ( Power(t) dt) )  = total frames / integral( Power(t) dt)

Where the part with the integral is the mean-value of the Watts vs. time curve over the encode time.

and this has dimensions of frames per energy. The time dimensions effectively "cancel" out. What do you think of this measure of efficiency? I feel like I understand what it's saying both mathematically and intuitively, and I really don't have any intuition for the definition you posed, and it seems like this might be a problem for other people too.

 

[AtenRa]

Total energy required to complete a job is not an unreasonable definition of efficiency. It's also not the only definition of efficiency. The reason your car analogy was confusing is because miles per gallon, or kilometers per litre (for the civilized among us) are very common measures of fuel efficiency for motor vehicles and they do not involve the speed of the car at all. In fact, most cars that I've see hit

Energy consumption alone is not a reasonable definition of Efficiency in Electronics. Without Performance, energy consumption alone is useless.

Let me give you an example,

According to my results Core i3 4330 finishes the x264 benchmark in 61,51 minutes and consumes 60Wh of energy.

Now I can take a Pentium 4, downclock it and make it consume the same energy of 60Wh but it will finish the benchmark in 12 hours (hypothetical number).

Although both CPUs consumed the same energy to finish the same job, it is clear that the Pentium 4 doesnt have the same efficiency and it is unsuitable for the task.

I think the confusion in this thread arose largely from the fact that you did not very clearly explain what you meant by either "efficiency", or "performance".

I have specifically said Efficiency = Performance (fps or time) / Watt or Wh more than once.

Let me try to be very explicit. When you say "efficiency", you mean the average FPS reading divided by the total amount of energy used to perform the job, is that correct? The resulting units are:

 (Frames / second) / (watt* hr) -> frames / (Joule * hour)

This seems kind of weird to me.

As i have said Efficiency is just a number. When they measure efficiency in the Green 500 Super computers they measure FLOPs / watt. The outcome is just a number derived from mflops/watt.

http://www.green500.org/news/green500-list-june-2015?q=lists/green201511

 

[ShintaiDK]

So you say efficiency is performance/watt. But you reject your own numbers because it ended up equal?

The relative speed and time is irrelevant in performance/watt.

 

[mrmt]

So you say efficiency is performance/watt. But you reject your own numbers because it ended up equal?

The relative speed and time is irrelevant in performance/watt.

Our resident AMD reseller should be diversifying its business, he should have entered in partnership with PPC, the Greek energy company.

 

[Yuriman]

Energy consumption alone is not a reasonable definition of Efficiency in Electronics. Without Performance, energy consumption alone is useless.

Let me give you an example,

According to my results Core i3 4330 finishes the x264 benchmark in 61,51 minutes and consumes 60Wh of energy.

Now I can take a Pentium 4, downclock it and make it consume the same energy of 60Wh but it will finish the benchmark in 12 hours (hypothetical number).

Although both CPUs consumed the same energy to finish the same job, it is clear that the Pentium 4 doesnt have the same efficiency and it is unsuitable for the task.

I doubt you could finish the benchmark consuming only 60 watt hours with a Pentium 4. If it takes 12 hours it finish, the CPU would have to be consuming 5 watts. (Though I know you made up a number for the sake of argument)

I have specifically said Efficiency = Performance (fps or time) / Watt or Wh more than once.

FPS / watts is a reasonable measure of efficiency, because it's rate of work divided by rate of energy usage, which simplifies to work / energy.

FPS / Wh makes no sense.

Time / Wh makes no sense.

Time / watts makes no sense.

A given task / Wh makes sense, however, because it shows what you did (a given task, e.g. encoded 23,000 frames) over how much energy was used to do it (Wh = rate of energy consumption * time used).

As i have said Efficiency is just a number. When they measure efficiency in the Green 500 Super computers they measure FLOPs / watt. The outcome is just a number derived from mflops/watt.

http://www.green500.org/news/green500-list-june-2015?q=lists/green201511

Yes, watts, not watt-hours. FLOPs is the rate at which work is done. Watts is the rate at which energy is consumed. Watt-hours is the rate at which energy is consumed, multiplied by the time over which it was consumed, giving you the total energy consumed over a period of time.

You cannot use watts and watt-hr interchangeably. They are not the same thing.

 

[Phynaz]

Somebody failed middle school science.
Assuming they have middle school in Greece

insulting other members is not allowed
Markfw900

 

[Yuriman]

Look at it this way - if your power company bills you based on watt hours (which they do in the US), they don't bill you for simply having a 60 watt light bulb plugged into the wall, they bill you for how long it was turned on, too. Watt hours = watts used * hours used. That's why it's watt-hours, and not just watts.

 

[Essence_of_War]

Energy consumption alone is not a reasonable definition of Efficiency in Electronics. Without Performance, energy consumption alone is useless.

Let me give you an example,

According to my results Core i3 4330 finishes the x264 benchmark in 61,51 minutes and consumes 60Wh of energy.

Now I can take a Pentium 4, downclock it and make it consume the same energy of 60Wh but it will finish the benchmark in 12 hours (hypothetical number).

Although both CPUs consumed the same energy to finish the same job, it is clear that the Pentium 4 doesnt have the same efficiency and it is unsuitable for the task.

I hear what you're saying. You'd like "efficiency" to encode in a single number "how good this CPU is at this task" and I think you're asking too much of a single number and distorting any intuitive definition of "efficiency" in the process.

The more intuitive and meaningful measure of efficiency that I provided (how many FPS does a Watt of power get me) works pretty well here!

From the graphs you've provided and the frames you've mentioned:

totalFrames = 5906*4*2 passes

           t (s)       kJ         fps / W
A10-OC    2.968e3      432        0.109
A10-      3.381e3      360        0.131
A8-76-65  3.613e3      324        0.146
A8-76-55  3.780e3      288        0.164
A8-76-45  4.048e3      288        0.164
A10-58    3.921e3      360        0.131
A8-38     3.842e3      396        0.119
A8-38OC   4.163e3      432        0.109
i3-43     3.691e3      216        0.219
i3-32     4.328e3      288        0.164

The i3-43 stands out. As do the A8s and the i3-32. The OC'ing is particularly bad for efficiency (which is something we know is intuitively true because higher clocks usually require higher voltage, and so power typically isn't linear with clockspeed for most overclocks that people care about).

As i have said Efficiency is just a number. When they measure efficiency in the Green 500 Super computers they measure FLOPs / watt. The outcome is just a number derived from mflops/watt.

http://www.green500.org/news/green500-list-june-2015?q=lists/green201511

"Efficiency" is not just a number. It is a number that has units and units mean things. If you want to make up a suite of statistics for your own purposes, you can pick whatever you want, but if you want to communicate something meaningful to other people, you have to care about what your units and what characteristics your choice of units is trying to express.

I have specifically said Efficiency = Performance (fps or time) / Watt or Wh more than once.

So which is it?

fps doesn't have the same units as time and watts don't have the same units as watt*hours.

Do you mean it is either (fps / watt), or time / watt*hour? Because that's even more confusing. The first of those is what I'm claiming is a good measure of efficiency, and what I think most people think of when they discuss efficiency of an encode job. The 2nd is, as far as I can tell, just inverse Watts, which is not really meaningful here.

 

[AtenRa]

The more intuitive and meaningful measure of efficiency that I provided (how many FPS does a Watt of power get me) works pretty well here!

From the graphs you've provided and the frames you've mentioned:

totalFrames = 5906*4*2 passes

           t (s)       kJ         fps / W
A10-OC    2.968e3      432        0.109
A10-      3.381e3      360        0.131
A8-76-65  3.613e3      324        0.146
A8-76-55  3.780e3      288        0.164
A8-76-45  4.048e3      288        0.164
A10-58    3.921e3      360        0.131
A8-38     3.842e3      396        0.119
A8-38OC   4.163e3      432        0.109
i3-43     3.691e3      216        0.219
i3-32     4.328e3      288        0.164

The i3-43 stands out. As do the A8s and the i3-32. The OC'ing is particularly bad for efficiency (which is something we know is intuitively true because higher clocks usually require higher voltage, and so power typically isn't linear with clockspeed for most overclocks that people care about).

Your fps / W numbers are wrong

If we take pass 2 fps results and Wh and time to finish the benchmark we have

Power (W) = Wh / h

Efficiency (n) = Performance (fps) / Watt (power)

So from those we get,

A8-7600 55 TDP
Time to finish the benchmark = 63 Minutes = 1,05 hours
Power (W) = 80Wh / 1,05 = 76,19 Watts
Performance (pass 2) = 8,12 fps

Efficiency = fps / w = 8,12 / 76,19 = 0.106

Core i3 3225
Time to finish the benchmark = 72,13 Minutes = 1,2 hours
Power (W) = 80Wh / 1,2 = 66,66 Watts
Performance (pass 2) = 6,63 fps

Efficiency = fps / w = 6,63 / 66,66 = 0.099


Core i3 4330
Time to finish the benchmark = 61,51 Minutes = 1,02 hours
Power (W) = 60Wh / 1,02 = 58,82 Watts
Performance (pass 2) = 7,83 fps

Efficiency = fps / w = 7,83 / 58,82 = 0.133

Higher number = Higher Efficiency

Same applies for the rest of the CPUs.

As i have said before, doing the same work faster (Less time) using the same energy makes you more efficient.
Here A8-7600 consume the same Energy to do more the same work faster (more calculations per time = less time = faster) by consuming the same Energy. That makes it more efficient than the Core i3 3225.

 

[SAAA]

A8-7600 55 TDP
Time to finish the benchmark = 63 Minutes = 1,05 hours
Power (W) = 80Wh / 1,05 = 76,19 Watts
Performance (pass 2) = 8,12 fps

Efficiency = fps / w = 8,12 / 76,19 = 0.106

Core i3 3225
Time to finish the benchmark = 72,13 Minutes = 1,2 hours
Power (W) = 80Wh / 1,2 = 66,66 Watts
Performance (pass 2) = 6,63 fps

Efficiency = fps / w = 6,63 / 66,66 = 0.099

...Here A8-7600 consume the same Energy to do more work (more calculations per time = less time = faster) by consuming the same Energy. That makes it more efficient than the Core i3 3225.

Wooow we have finally discovered that a quad core is more efficient than a dual core at doing highly parallel tasks!

Beside the irony I'd expect that to happen when one architecture is designed for high IPC and frequency with larger cores while the other excels at parallel workloads, regardless of the node they are built on.

Btw the newer i3 has better efficiency while being on the same 22nm process... designs matter as much as new nodes nowadays.

 

[TheELF]

Wooow we have finally discovered that a quad core is more efficient than a dual core at doing highly parallel tasks!

Yeah by .007 of whatever unit this is supposed to be...whoop de doo.

Also x264 is one of the most inefficient codes out there and unless you are going to use x264 command line you will probably never encounter it.

But, easily run benchmark I guess.

 

[frozentundra123456]

Energy consumption alone is not a reasonable definition of Efficiency in Electronics. Without Performance, energy consumption alone is useless.

Let me give you an example,

According to my results Core i3 4330 finishes the x264 benchmark in 61,51 minutes and consumes 60Wh of energy.

Now I can take a Pentium 4, downclock it and make it consume the same energy of 60Wh but it will finish the benchmark in 12 hours (hypothetical number).

Although both CPUs consumed the same energy to finish the same job, it is clear that the Pentium 4 doesnt have the same efficiency and it is unsuitable for the task.


I have specifically said Efficiency = Performance (fps or time) / Watt or Wh more than once.

As i have said Efficiency is just a number. When they measure efficiency in the Green 500 Super computers they measure FLOPs / watt. The outcome is just a number derived from mflops/watt.

http://www.green500.org/news/green500-list-june-2015?q=lists/green201511

Give us your formula for calculating "efficiency" and the units of the result. For instance, according to your definition, whatever it is, how do you determine which is more efficient when one processor uses less energy but takes longer to complete the task?

 

[Essence_of_War]

Your fps / W numbers are wrong

If we take pass 2 fps results and Wh and time to finish the benchmark we have

Power (W) = Wh / h

Efficiency (n) = Performance (fps) / Watt (power)

So from those we get,

A8-7600 55 TDP
Time to finish the benchmark = 63 Minutes = 1,05 hours
Power (W) = 80Wh / 1,05 = 76,19 Watts
Performance (pass 2) = 8,12 fps

Efficiency = fps / w = 8,12 / 76,19 = 0.106

Core i3 3225
Time to finish the benchmark = 72,13 Minutes = 1,2 hours
Power (W) = 80Wh / 1,2 = 66,66 Watts
Performance (pass 2) = 6,63 fps

Efficiency = fps / w = 6,63 / 66,66 = 0.099


Core i3 4330
Time to finish the benchmark = 61,51 Minutes = 1,02 hours
Power (W) = 60Wh / 1,02 = 58,82 Watts
Performance (pass 2) = 7,83 fps

Efficiency = fps / w = 7,83 / 58,82 = 0.133

Higher number = Higher Efficiency

Same applies for the rest of the CPUs.

As i have said before, doing the same work faster (Less time) using the same energy makes you more efficient.
Here A8-7600 consume the same Energy to do more work (more calculations per time = less time = faster) by consuming the same Energy. That makes it more efficient than the Core i3 3225.

I used the total number of frames including pass 1 and pass 2 because you only provided energy-consumption for the entire benchmark, not broken down by pass-1 and pass-2. It is not correct to pick fps-during-pass-2 as your performance metric but then take power measurements from the full benchmark that includes both passes to compute the "efficiency". It's probably doesn't break the comparison too much as long as you do this uniformly for each CPU, but I think the correct measure should be:

(avg FPS for pass N) / (avg power for pass N)

OR

(avg FPS overall) / (avg power overall)

I double checked the numbers that I entered above, and they are correct for the values you provided with the formula for fps / W that we now apparently agree is the correct measure of efficiency:

(totalFrames / time (s) ) / ( total Energy (J) / time(s) )

And like I said, the i3-43 is the stand out, with the i3-32 and the lower TDP A8's a 3-way tie for 2nd. Have I misunderstood the values you've provided for total energy? It says on the graph that they are for "completing the benchmark", so that should be both pass-1 and pass-2, correct?

 

[Yuriman]

Your fps / W numbers are wrong

If we take pass 2 fps results and Wh and time to finish the benchmark we have

Power (W) = Wh / h

Efficiency (n) = Performance (fps) / Watt (power)

So from those we get,

A8-7600 55 TDP
Time to finish the benchmark = 63 Minutes = 1,05 hours
Power (W) = 80Wh / 1,05 = 76,19 Watts
Performance (pass 2) = 8,12 fps

Efficiency = fps / w = 8,12 / 76,19 = 0.106

Core i3 3225
Time to finish the benchmark = 72,13 Minutes = 1,2 hours
Power (W) = 80Wh / 1,2 = 66,66 Watts
Performance (pass 2) = 6,63 fps

Efficiency = fps / w = 6,63 / 66,66 = 0.099


Core i3 4330
Time to finish the benchmark = 61,51 Minutes = 1,02 hours
Power (W) = 60Wh / 1,02 = 58,82 Watts
Performance (pass 2) = 7,83 fps

Efficiency = fps / w = 7,83 / 58,82 = 0.133

Higher number = Higher Efficiency

Same applies for the rest of the CPUs.

As i have said before, doing the same work faster (Less time) using the same energy makes you more efficient.
Here A8-7600 consume the same Energy to do more work (more calculations per time = less time = faster) by consuming the same Energy. That makes it more efficient than the Core i3 3225.

Here's your fallacy:

The A8 does not do more work than the i3 - both of them have completed the same task, and thus have done the same work.

Quoted again for emphasis:

Here A8-7600 consume the same Energy to do more work (more calculations per time = less time = faster) by consuming the same Energy.

The A8 did the work faster, but it did it while consuming more power per unit time. It did not do the work faster while drawing power at the same rate, as you suggest. According to your own numbers, the A8 drew 76.19 watts, while the i3 3225 drew 66.66 watts during the benchmark.

The A8 and the i3 both did the same task. The A8 did the task faster, by drawing more power. The i3 did the task slower, but drew less power. Both finished at different times. At the end of the task, both systems used the same amount of power to do the same task. What we can draw from this is that the A8 is faster, but that they both need the same amount of energy to do a task. Being faster is a positive characteristic, nobody is trying to deny that, but "faster" is not synonymous with "more efficient". The A8 is faster and equally efficient - which is good.

 

[MiddleOfTheRoad]

Wooow we have finally discovered that a quad core is more efficient than a dual core at doing highly parallel tasks!

Beside the irony I'd expect that to happen when one architecture is designed for high IPC and frequency with larger cores while the other excels at parallel workloads, regardless of the node they are built on.

Btw the newer i3 has better efficiency while being on the same 22nm process... designs matter as much as new nodes nowadays.

AMD has never produced chips on a 22nm node. The smallest node thus far on a chip available to a consumer from AMD has been the older 28nm. Kaveri's efficiency is pretty damn amazing considering it can rival Intel's efficiency despite a node disadvantage.

 

[Sweepr]

Comparing AMD's current APUs to 2012-era Ivy Bridge in a specific benchmark is nonsense/irrelevant. Core i3-6100 is considerably faster and more efficient than Kaveri/Godavari.

And 22nm FinFET vs 28nm planar product comparisons should include Haswell instead.

 

[Essence_of_War]

What we can draw from this is that the A8 is faster, but that they both need the same amount of energy to do a task. Being faster is a positive characteristic, nobody is trying to deny that, but "faster" is not synonymous with "more efficient". The A8 is faster and equally efficient - which is good.

I want to underline this like twenty times.

I tried to express this sentiment in one of my posts above, but this is a very good summary of much of the confusion in this thread. AtenRa is trying to combine the usual definitions of "efficiency" and "speed" into one number that means neither, and this has been confusing to everyone else.

 

[SlowSpyder]

Here's your fallacy:

The A8 does not do more work than the i3 - both of them have completed the same task, and thus have done the same work.

This is why I said earlier that gaming is different than jobs with a beginning and end point. But, obviously in this case the A8 can be more efficient as far as production as you can get more done in a unit of time, more done in a day's work (though the power efficiency would be the same between the two). Though obviously anyone that needs that kind of rendering power is probably not looking at i3's or A8's, but it is still an interesting data point, and I don't consider AtenRa 'wrong', just seems you guys were talking about different measures of efficiency.

 

[Abwx]

Comparing AMD's current APUs to 2012-era Ivy Bridge in a specific benchmark is nonsense/irrelevant. Core i3-6100 is considerably faster and more efficient than Kaveri/Godavari.

And 22nm FinFET vs 28nm planar product comparisons should include Haswell instead.

Lol...

Did you actualy look at this "graph"..??..

 

[Timur Born]

Are we still discussing "smoother desktop performance" here?

 

[AtenRa]

I used the total number of frames including pass 1 and pass 2 because you only provided energy-consumption for the entire benchmark, not broken down by pass-1 and pass-2. It is not correct to pick fps-during-pass-2 as your performance metric but then take power measurements from the full benchmark that includes both passes to compute the "efficiency". It's probably doesn't break the comparison too much as long as you do this uniformly for each CPU, but I think the correct measure should be:

(avg FPS for pass N) / (avg power for pass N)

OR

(avg FPS overall) / (avg power overall)

I double checked the numbers that I entered above, and they are correct for the values you provided with the formula for fps / W that we now apparently agree is the correct measure of efficiency:

(totalFrames / time (s) ) / ( total Energy (J) / time(s) )

And like I said, the i3-43 is the stand out, with the i3-32 and the lower TDP A8's a 3-way tie for 2nd. Have I misunderstood the values you've provided for total energy? It says on the graph that they are for "completing the benchmark", so that should be both pass-1 and pass-2, correct?

The energy numbers are for the entire benchmark, i just used pass 2 to make it easier to understand and because pass 2 is doing the encoding.

 

[AtenRa]

Here's your fallacy:

The A8 does not do more work than the i3 - both of them have completed the same task, and thus have done the same work.

Correct my bad, i fixed that.

The A8 did the work faster, but it did it while consuming more power per unit time. It did not do the work faster while drawing power at the same rate, as you suggest. According to your own numbers, the A8 drew 76.19 watts, while the i3 3225 drew 66.66 watts during the benchmark.

The A8 and the i3 both did the same task. The A8 did the task faster, by drawing more power. The i3 did the task slower, but drew less power. Both finished at different times. At the end of the task, both systems used the same amount of power to do the same task. What we can draw from this is that the A8 is faster, but that they both need the same amount of energy to do a task. Being faster is a positive characteristic, nobody is trying to deny that, but "faster" is not synonymous with "more efficient". The A8 is faster and equally efficient - which is good.

Being faster means Higher Compute = more calculations per unit of time

Example, IPC (Instructions Per Cycle). For every cycle more instructions = higher compute = more calculations per cycle = faster.

If you finish the same work (benchmark) faster (higher IPC = Higher Compute = less time) and you consume the same energy (Wh) you are more efficient.