Are there power savings when running 2C/4T vs 4C/4T CPUs?

[Borealis7]

i thought about something but i cant really check it...
are there power savings or even lower temps when running a 2C/4T CPU at the same speed as 4C/4T CPU from the same architecture?
i imagine there would be since each core means more transistors which all do more "work" and take some voltage to run and subsequently produce heat, although you do get less cache overall on the CPU.

if there is some saving and performance is virtually equal, then logic dictates that 2C/4T CPUs are superior, and could probably O/C higher. am i right?

 

[3DVagabond]

I don't think the performance would be roughly, or anywhere near, equal. Hyper-threading doesn't come close to doubling performance, AFAIK. Twice as many whole cores does though.

 

[JFAMD]

Since this is a hypothetical situation it is hard to give you an answer, however, the real power difference is not at full load, but at half load.

The 4C/4T processor, with only 2 threads active, should draw less power because 2 cores can essentially shut down or power gate the remaining 2 cores.

The 2C/4T processor still has 100% of its cores active. The only way that the processor could have half active (and power gating) is if it is running the 2 threads on a single core, in which case you are taking a performance hit.

 

[LoneNinja]

Technically not the exact same processor, but look at the I5 dual cores vs the I5 quad cores, huge performance difference comparing 2C/4T to 4C/4T. Difference in load power consumption too, look at any of the I3/I5 reviews and you'll see that. Idle power is very similar between most processors these days, so I wouldn't even worry about that.

 

[khon]

2C/4T is not at all the same as 4C/4T.

For example looks at this comparison between the i5-650 and the i5-750: http://www.anandtech.com/bench/Product/109?vs=144

Even though the i5-750 runs at a much lower clock rate (2.66GHz vs 3.2GHz) it still beats the i5-650 at almost everything, and by a fairly large margin too.

As for power usage its hard to compare the two since one is 45nm and the other is 32nm, but yeah you probably would see some power savings with the 2C/4T vs 4C/4T.

 

[bryanW1995]

The theoretical max capability of ht is ~ 50%. In the real world it is typically quite a bit less than that. A 2c/4t cpu is probably most comparable to an amd 3 core.

 

[khon]

The theoretical max capability of ht is ~ 50%. In the real world it is typically quite a bit less than that. A 2c/4t cpu is probably most comparable to an amd 3 core.

Even that is probably overstating it. HT seems more like 20-30% improvement in real applications, so essentially 2C/4T would be about the same as 2.5C.

It's a shame the i5-760 isn't in the anandtech bench tool so we could do an exact comparison with i7-860, but this comparison between i5-750 and i7-920 gives some idea of what to expect:

http://www.anandtech.com/bench/Product/109?vs=47

 

[cusideabelincoln]

You have to factor in what type of workload is being done. Is it time sensitive? If so then you have to look at performance per watt over a specified time. 4C/4T will use more power, but it will probably finish the task quicker. Over all power consumption can be less if 2C/4T takes significantly longer.

Even that is probably overstating it. HT seems more like 20-30% improvement in real applications, so essentially 2C/4T would be about the same as 2.5C.

It's a shame the i5-760 isn't in the anandtech bench tool so we could do an exact comparison with i7-860, but this comparison between i5-750 and i7-920 gives some idea of what to expect:

http://www.anandtech.com/bench/Product/109?vs=47

There's platform differences in power consumption too.