Makes sense I suppose, kinda like how ATi used to make pipe cleaner chips to get their feet wet with a new process before releasing the top end stuff.
Ivy sounds like it will be a good choice for CULV parts if what is quoted about performance at low voltage is true though?
No, not really, 37% performance doesn't mean anything for chips, its the process.
If Intel make a smaller Ivybridge chip thats 2Ghz and the same basic architecture as Sandybridge, it will perform the same as a 2Ghz 32nm Sandy.
If they decide to bump the clocks up by 37%, it would scale quite well, thats very unlikely. The lowest voltage where the performance increase is the highest, and in terms of process, performance increase means switching speed, the Ghz its capable of essentially, you're talking about 0.5-0.6v, its closer to 10-15% at higher speed.
This is basically always the case, for every process. Most processes simply use new tricks to get transistors closer together without the problems from moving them closer together stopping it working. These will increase switching speed, drop leakage, drop power usage, etc, etc.
I would dare say that 32nm had a fairly similar performance advantage over 45nm 10-20% more "process performance" at the high end of the voltage scale.
Don't forget Intel is planning to move Atom to 22nm, which will also be a low power part, and is quite probably the chip to gain the most from the "performance increase" at low voltage.
If Intel decide to put out a 3.6Ghz chip, it won't be 37% faster, if they decide to release a 5Ghz chip, it should be 37% faster. They won't though, I'd look for, better competition with llano/trinity next year in terms of igp performance, a bump in clock speeds, a higher turbo clock, and a small bump in clock speed of the highest models, a very decent bump in benchmarks involving the IGP, and very little increase IF ANY in cpu only benchmarks clock for clock. I'd also expect the top chip to be a bit faster than the top Sandy in clock speed 100-200Mhz.
Whatever is after Ivy looks very interesting, the rumours I heard a while back would be along the lines of octo cores replacing quads, which frankly, will be I would think, 70-80% faster but will depend on platform, bandwidth(will they stay dual channel and cheap in the mainstream?) and if they cut the GPU from say the 8-6 core parts or not. I wouldn't be surprised if we got octo cores without gpu's, and quad cores with the IGP.
