Best performance per watt

[ams23]

Example (numbers are not real) :

Lets say that Quad Core ATOM Baytrail takes 20 hours to encode a video using 2Watts per hour. That means it will use 40Watts to finish the job.

Core i7 4770 Haswell needs 2 Hours to finish the same encode, that means it will consume 2x 87Watts = 174Watts.

Now, if you calculate the Performance per Watt you will find that ATOM has higher performance per Watt than Haswell Core i7. That is true but, ATOM Baytrail is also 10x slower than Haswell.

This is completely wrong. If the i7 processor really was 10x faster than an Atom Bay Trail processor (which, of course, is way off in the first place), and it used only 4.35x more power to complete the exact same task, then it would have higher perf. per watt compared to Atom! So what you are saying makes no sense.

In an environment that is not constrained by power, higher perf. per watt may or may not translate to higher perf. In an environment that is constrained by power, higher perf. per watt will directly translate to higher perf.

 

[AtenRa]

This is completely wrong. If the i7 processor really was 10x faster than an Atom Bay Trail processor (which, of course, is way off in the first place), and it used only 4.35x more power to complete the exact same task, then it would have higher perf. per watt compared to Atom! So what you are saying makes no sense.

In an environment that is not constrained by power, higher perf. per watt may or may not translate to higher perf. In an environment that is constrained by power, higher perf. per watt will directly translate to higher perf.

Yes my mistake, I took Hours as performance. The example is valid if ATOM takes less than 4x the time of Core i7(approximately).

 

[fuzzymath10]

It's just semantics. In Canada, we measure mileage (kilometreage?) in litres per 100km. In the US, it's miles per gallon. Either way, you're comparing the same things except the plug is different. They follow a constant inverse relationship.

Similarly, double the performance per watt is equal to saying half the watts per unit performance. They're essentially measuring the same thing.

 

[ElFenix]

You can measure 2 things. Performance/Watt for the CPU or Performance/Watt for the entire system. The performance would be a workload case. For example encoding a movie or running a game. When task is done, you look at energy used over the timeframe.

that's actually watt/performance

 

[AtenRa]

It's just semantics. In Canada, we measure mileage (kilometreage?) in litres per 100km. In the US, it's miles per gallon. Either way, you're comparing the same things except the plug is different. They follow a constant inverse relationship.

Similarly, double the performance per watt is equal to saying half the watts per unit performance. They're essentially measuring the same thing.

I mean I only used Hours to show the performance per watt in my example.

So lets see the example again (Correct this time).

Core i7 4770 encodes with 60fps per hour using 87Watt per hour and it takes 2 hours to finish the job using 2x 87Watt = 174Watt

Performance per Watt = 60fps / 87Watt = ~0.69fps per Watt

Atom encodes with 6fps per hour using 2Watt per hour and it takes 20 Hours to finish the same job using 20x 2Watt = 40Watt.

Performance per Watt = 6fps / 2Watt = 3fps per Watt

ATOM has ~3x times more performance per watt than the Core i7 but it is 10x slower than Core i7.

This is only an example to show us that higher performance isnt more efficient measuring performance per watt. It also show that if Time is not important and we only want to encode a single Movie, encoding with ATOM is more efficient.

But this is only measuring Response time, how much time it takes for a CPU to finish a single job.

What happens with Throughput ??
Now lets assume that Core i7 can encode eight(8) tasks at the same time. It can encode 8 different movies simultaneously with 20fps using 87Watt. That means it can finish 8 movies per 6 Hours. Per day (24 hours) it will finish 32 movies using 24x 87Watt = 2088Watt or 2.08Kw

Throughput Performance per Watt = 32movies / 2.08Kw = 15,38 Movies per Kw or 0,015 Movies per Watt.

ATOM being a Quad Core(No HT) with less Cache will be severely bottlenecked in Multitasking with 8 encodes simultaneously. Lets say it can encode eight(8) tasks with 0.5fps. Per day it will finish 0.25 Movies using 2Watt x 24 = 48Watt.

Throughput Performance per Watt = 0.25movies / 48w = 0,0052 Movies per Watt.

So now Core i7 has more Performance per Watt than ATOM.

 

[Torn Mind]

Core i7 4770 encodes with 60fps per hour using 87Watt per hour and it takes 2 hours to finish the job using 2x 87Watt = 174Watt

It's not a matter of converting W/h to watts. It's a matter of converting Watts(energy/second) to a measure of energy, such as Joules or Kilowatt-hours. Finding energy involves multiplying the watts by time. The Core i7 uses 174 watt-hours.

Watt per hour measures the change in power per unit of time, which has no practical purpose in measuring energy consumption of residential products.

 

[Torn Mind]

thats because i believe it becomes multiplicative due to the higher loss as HEAT at the faster speed...
Also the IC's are more efficient and faster at lower temperatures.

You guys forget... the faster you push your CPU, what happens?? it becomes hotter.
Heat is a waste product not a performance indicator.

Heat energy is measure in Joules and since CPUs can be treated as perfect resistors, all the electrical energy coursing through the CPU can be assumed to be converted to heat energy. It's hotter because higher clockspeeds and voltages need more require more energy.

 

[sm625]

It's not a matter of converting W/h to watts. It's a matter of converting Watts(energy/second) to a measure of energy, such as Joules or Kilowatt-hours. Finding energy involves multiplying the watts by time. The Core i7 uses 174 watt-hours.

Watt per hour measures the change in power per unit of time, which has no practical purpose in measuring energy consumption of residential products.

lol now that's semantical overload. No one uses watts per hour to describe the change in power per unit of time, it is almost always used by mistake to mean watt-hours. If my computer was using 60 watts per hour at idle then it would be using 60 watt-hours per hour, and I would be hard-pressed to see how any part of that statement could be semantically incorrect, even if I said "my computer used 60 watts per hour". I think we all know what it means even if its technically not a correct way of saying it.

 

[TuxDave]

lol now that's semantical overload. No one uses watts per hour to describe the change in power per unit of time, it is almost always used by mistake to mean watt-hours. If my computer was using 60 watts per hour at idle then it would be using 60 watt-hours per hour, and I would be hard-pressed to see how any part of that statement could be semantically incorrect, even if I said "my computer used 60 watts per hour". I think we all know what it means even if its technically not a correct way of saying it.

As an engineer, "watts per hour" is a really odd metric and consuming "60 watts" over an hour is also a weird thing to say. It's like hearing.

Q: "How far is your house from here?"
A: "60 mph"