OK I have established why you and many others confuse things, its because of TIME.
LESS TIME = Higher Performance
So, Ferrari running the same distance at 2 Secs FASTER than the TOYOTA by consuming the same 1 Gallon of Petrol is the more Efficient of the two CARs.
The more efficient is the one that does MORE work (More fps OR less time) by using the same amount of energy.
In the case of the CARs, the Ferrari runs the same distance FASTER (2 secs LESS TIME) by consuming the same amount of petrol.
In the case of the x264 benchmark, the A8-7600 is doing more work (LESS TIME to finish the same job) by consuming the same amount of energy.
DON’T confuse the same Job/task (x264 benchmark) with the performance of each CPU. And Watthours IS NOT efficiency.
Efficiency = Performance (Time – Less time = higher performance) / Watt (or energy Wh)
Again, you seem confused.
So by your calculation, the Ferrari is what, 17% more efficient, because its rate of acceleration is 1-(10/12)*100 better, while covering the same distance and using the same fuel. What happens if the fuel consumption goes up by 20%? The Ferrari is now faster, but uses more fuel to cover the same distance, while covering the same distance. Does that make it equally efficient, despite using more fuel to cover the same distance, just because it's faster? (Hint: Your problem here is that you are using three variables)
Watts and watt hours are not interchangeable.
Let's go back to this chart again, and analyze your numbers:
I'm going to compare the A8 7600 and the i3 3225 using x264 second pass, in which the A8 has higher throughput/FPS.
A8 7600: 7.52 Frames / second
Total energy used: 80 watts * hours
i3 3225: 6.63 Frames / second
Total energy used: 80 watts * hours
Let's break down the A8's power consumption:
7.52 frames / 1 second = 80 watts * 1 hour / 3600 seconds (converting from watt hours into watt seconds)
This simplifies into:
7.52 frames / 1 second = 0.022 watts / 1 second
Let's rearrange this to get joules, a unit of energy, and not of energy rate:
7.52 frames = 0.022 watts *
1 second / 1 second
And then,
7.52 frames = 0.0222 joules / 1 second
1 frame = 0.0030 joules / 1 second
But wait, that doesn't make any sense! You can't have a single frame equaling a rate of energy. What went wrong?
The key piece of information here is that in order to compare watt-hours to frames, you need to have a fixed number of frames generated, not a frame rate. You can't compare a rate (frames per second) to a non-time dependent unit of energy (watt-hours).
So, let's say for the sake of argument that both systems needed to render 2,000,000 frames - I assume you encoded the same clip with both CPUs. :sneaky:
The A8 used 80 watts*hours to render our hypothetical 2,000,000 frames. 80 watt hours is 288,000 joules, so some basic math shows us that the A8 would have used ~7 joules (2,000,000 / 288,000) to render each frame. It rendered these frames at 7.52 frames per second. 7 joules * 7.52 frames per second = 26.32 joules / 1 second, or 52.64 watts. (I chose 2 million as our frame number because that gives us a realistic wattage).
The i3 also used 80 watts*hours to render our hypothetical 2,000,000 frames, meaning it also used 7 joules to render each frame. However, it only rendered at 6.63 frames per second, and 6.63fps * 7 joules = 46.31 watts.
~
Does this mean the i3 is using less power? Yes, per unit time, but it took longer to do the same task, so it used the same amount of energy.
So, I'll ask again, what exactly are you comparing here? Are you asserting that the A8, rendering at a higher rate but using the same amount of energy to render each frame, is more efficient, because it finished first? Despite pulling more watts?
I'd say the convention here is to say that both CPUs are equally efficient (the A8 and i3 both need the same amount of energy to render each frame), but the A8 is
faster, which has absolutely nothing to do with how efficiently it converts electrical energy into useful work.
Facebook
Twitter