Intel CPUs performance increasing.

[AAA]

http://www.engadget.com/2007/09/28/intel...h-details/

What was the deal/concern once the 'old' pentiums reached high speeds? It was some sort of radio "Active" phenomenon, but I can't recall the term.

will it become a consideration once again as these cores exceed 3 or 4 Ghz?

[M A V I C]

pulling of 45nm is a feat. the main issue with the P4 line was they got really hot at higher clock speeds. the Core processors already run much cooler at lower speeds than the P4's did.

The Core processors have made every AMD fan I know, take another look at Intel. I know a few people who have never owned anything but AMD, that have switched to Intel.

[sekker]

If the Intel map is right, we'll be seeing 8 core CPUs soon.

Amazing, given how recently Intel was wedded to the P4 (single core) business model.

[sam ward]

Don't multi-cores' usefulness depend on software augmentation/adaptation?

[Filliam H. Muffman]

[quote AAA]What was the deal/concern once the 'old' pentiums reached high speeds? It was some sort of radio "Active" phenomenon, but I can't recall the term.
There are actually several things going on. The main ones are leakage and transistor switching losses. Those two account for something like 40% of the power used by the processor before it actually does anything. This graphic illustrates some of the improvements over the previous generation(s).

http://www.hkepc.com/hwdb/intel-penryn-preview/hmtt.png <- if the hotlink does not work.




This short PDF has some fairly theoretical parts but should illustrate advances over the last 10 years.
http://www.intel.com/technology/magazine...n-0306.pdf energy-per-instruction-0306.pdf



[quote AAA]will it become a consideration once again as these cores exceed 3 or 4 Ghz?
Yes, otherwise they would already be at those speeds.


[quote M A V I C]the Core processors already run much cooler at lower speeds than the P4's did.
Running at a lower speeds is one of the main reasons that they ran cooler.


[quote sekker]If the Intel map is right, we'll be seeing 8 core CPUs soon.
Yes, but they need some significant architecture advances to take advantage of all those cores. For many current applications jumping from two cores to four cores only adds 50% to 60% performance. Doubling the number of cores again right now would give even lower gains.

I am expecting something like four channel 240 pin or dual channel 500 pin RAM, and 1300+ "pin" CPU sockets.

[M A V I C]

[quote Filliam H. Muffman]
Running at a lower speeds is one of the main reasons that they ran cooler.
Yes/no. A C2D at 2.4 runs MUCH cooler than some of the P4's at 2.4 ran. MUCH cooler.

[Filliam H. Muffman]

[quote M A V I C][quote Filliam H. Muffman]
Running at a lower speeds is one of the main reasons that they ran cooler.
Yes/no. A C2D at 2.4 runs MUCH cooler than some of the P4's at 2.4 ran. MUCH cooler.
Yes/yes. The current C2D runs at a lower core voltage and has the benefit of smaller traces and two generations of process refinement. If they were fabbed with the same size traces, process technology, a comparable amount of cache, number of transistors, and run at the same voltage, the P4 would be very close when run at the same frequency. The P4 architecture is less efficient per clock cycle but does allow for much higher clock speeds.

Edit: For a bit of history, the first P4 models that reached 2.4 GHz were made with a 130 nm process. If you know of a review comparing a Cedar Mill P4, the last generation made, running at the same voltage and clock speed as a C2D, I would be very interested

[Filliam H. Muffman]

(sigh) cut off by the editing limit.

Looking at this chart http://images.tomshardware.com/2007/07/1...el_big.png , the slowest P4 D 915 runs at 2.8 GHz using 1.31 V and has 376 million transistors with a 95 W power rating. The E6600 runs at 2.4 GHz using 1.35 V and has 296 million transistors with a 65 W power rating. I think they would have very similar power use when running the P4 915 at a 15% lower clock speed.

[M A V I C]

That just goes to show you that running at a lower clock speed isn't why they run cooler It's because of the processes used that they run cooler. (130nm vs...

[Filliam H. Muffman]

If you underclock a processor, it runs cooler. If you overclock a processor, it runs hotter.