Successor to Pentium 4?

[Dresdenboy]

The most important things, which might help GIMPS, are quick L1-to-L1-cache-transfers (if this is true) and that these CPUs will have a further developed micro-op fusion, which doesn't just fuse micro-ops, which belong to the same original x86 instruction (this is, what I think to be the case in Pentium M), but allows to fuse multiple x86 instructions to single macro-ops. The use of an internal FMAC instruction could be possible:

http://the-inquirer.net/?article=25623

There is also some info on Yonah regarding micro-op fusion for SSE2 instructions etc:
http://the-inquirer.net/?article=25643

[jinydu]

So the main question is: Are the new chips going to crunch for GIMPS faster than the Pentium 4?

[jinydu]

It looks like Intel has made the presentations available to the public:

http://www.intel.com/idf/us/fall2005/webcast.htm#

(Otellini's Opening Keynote introduces the new CPUs)

And they've also put up a mini-website on the new CPUs:

http://www.intel.com/technology/comp...ngma/index.htm

[jinydu]

Originally Posted by jinydu

http://www.intel.com/idf/us/fall2005/webcast.htm#

Before he introduces the Merom, Conroe and Woodcrest, Otellini says that Intel is changing from a "More GHz" approach to a "Performance per Watt" approach. The part about the new processors starts at about 13:30. He then goes on to talk about 65nm in general, dual cores in general, then low power chips. He promises to release "handtops" which use 10 times less power (0.5W) than today's laptops and desktops which are ten times faster than today's desktops; but this will likely take a very long time ("end of decade").

(It's the summer holidays. I have too much free time...)