SPCR

True or false:

Given the same TDP and clock frequency ratings, for a single thread application, a dual-core CPU consumes less power under full load (of the application) compared to a single-core counterpart.

In other words, the application will drive the CPU to 100% in a single core, thus hitting the full TDP, while a dual-core will only drive the total CPU usage to 50%, thus consuming far less than the rated TDP (for argument's sake, 50% of the rated TDP).

Does anyone see holes in this argument?

What you say makes some sense, but unfortunately TDP isn't normally the power dissipation that would occur when the cpu is at 100%. AMD processors generally would never reach the advertised tdp unless you overclock and/or overvolt. Intel processors might exceed their TDP before 100% load. I think the idea is that a system designer should provide a heat sink and fan that can carry away the TDP, then it should work OK. Intel may be relying on the cpu to shut itself down if it overheats to avoid damage in the case it exceeds that power.

If both processors are the same process (for example both 90 nm) the dual core cpu's active cpu would be doing the same work, and therefore probably dissipating about the same power, as the single core. If you compare a 65 nm dual core to a 90 nm single core, the 65 nm dual core would dissipate less heat.

frank2003 wrote:True or false:

In other words, the application will drive the CPU to 100% in a single core, thus hitting the full TDP, while a dual-core will only drive the total CPU usage to 50%, thus consuming far less than the rated TDP (for argument's sake, 50% of the rated TDP).

Makes sense, but it would be 50% + change, because the second core has to idle.

Thanks for your inputs.

My motivation for posing this question was this: If I can find a single-core processor that is sufficiently powerful to do a job such as decoding 1080i HDTV, then I can get a comparable dual-core processor and run the OS in single processor mode and thus can realize substantial power savings, all without undervolting.

The reasons for running the OS in single processor mode are two fold:
1. The maximum power consumption (in terms of TDP) will never reach the maxium, and
2. A single multithreaded run-away process, or multiple run-away processes will only drive one core to the maxium.

So in theory, if I get a 45W dual-core BE-2350 and run it in single processor mode, I will be seeing significantly less than 45W power consumption at full load.

you seem to be going an awfully long way to avoid undervolting. which OS are you planning to use? RMClock and CCPUID work in windows to dynamically undervolt.

I would consider undervolting if the BIOS supports it, without using any external tools that run in the OS.

So do these tools work without BIOS support for voltage control? If so, I'll take a look. Thanks.