Intel Ivy Brdige Third Gen 3000 Series Reviews

[Goku]

Today is the launch day for Ivy Bridge. Seems the rumors are correct in that it runs hot and overclocks not as well as SB. Still good if you are looking to upgrade from anything else than SB though. Its expected you can start buying these next Monday the 29th. Some B&M like Micro Center and Frys could potentially start selling earlier.

Hard OCP
Anandtech
Guru3D
Hardware Canucks
Xbit Labs

[Rbstr]

I don't know how optimistic you should really be. We're sort of running out of space. Though, I do have some faith in Intel recently. It's an engineer's kind of company, even in the face of really poor competition in the desktop space they aren't slowing down a hellovalot (I guess they learned their lesson several years ago).
The 16nm node is butting into a scale where lithography is quite fuzzy and quantum tunneling could start screwing things up. The next node, 11nm, seems like it will have to go to something besides traditional microelectronics manufacturing, that could, of course, bring whole new levels to overclocking or it could leave us with very sensitive chips.
It's also really important to remember clock speeds are not linear with power consumption or actual performance.

I'll be waiting for the 22nm Haswell parts to do another cpu upgrade...I like to skip a whole generation.
I'm also quite interested in if Intel is going to go to a new microachitecture, instead of kinda sticking with the Pentium III basis as they have been. I guess, once you're off silicon, it's a great opportunity?

[Goku]

Wasn't the tri-gate supposed to help with all what you are mentioning Rbstr? Maybe that was only a temp fix to 11nm or something like that.

[Rbstr]

Yeah, they're certainly part of a way to enable smaller transistors.

But the issues are multiple: If you're familiar with how chips are made they use a technique called lithography to basically stencil chips. Light's wavelength becomes an issue that's insurmountable, so you've got to move to e-beam, but at 16nm that can be hard. Plus getting the materials to be pure enough and deposit well at smaller and smaller scales.
Then you simply run into the scale being such that quantum mechanical effects like tunneling become legitimate issues. Electrons can tunnel 12nm in SiO2 in some cases...that could be a huge issue.

Atoms have a finite size and in really small amounts they don't really act the same as a whole bunch of them. They act more like single atoms. If that makes sense?
Crystalline silicon's unit cell is ~.5nm across. SiO2 is bigger.
So at 10nm we're talking a number of atoms you can count on your fingers and toes.

[Goku]

Do you think we will actually see Intel moving towards a new type of material to use transistors with the next decade? I don't see how else Intel will continue to boost performance otherwise.

[termhn]

There's gonna have to be something big and pretty soon because as Rbstr has pointed out, there are serious problems with just getting smaller. What that will be, I have no idea I've decided to wait for them cause my dad said he'll kick in a few bucks so I can get a Z77 and Ivy setup, partly because it'll eventually be a part-time hackintosh that he uses as well, and some of the new Z77 boards support hackintosh-ing better

[Rbstr]

Originally Posted by Goku

Do you think we will actually see Intel moving towards a new type of material to use transistors with the next decade? I don't see how else Intel will continue to boost performance otherwise.

They might last the decade or two with Silicon/traditional semiconductors and photolithography with stuff like http://en.wikipedia.org/wiki/Multiple_patterning#Intel
Or Electron beam lithography...but e-beam creates big issues in how long it takes to create patterns.
Photolithography is pretty famous for being said to be on the way out, and then it just keeps on getting better...so who really knows.

But soonish, new materials are needed and I can throw any number of buzzwordy things at you, nanotubes, graphene and on.
or a new way of creating the transistor with what we've got that makes use of the "problems" small scales create.
Hell things may not get smaller but simply better.
I'm not an EE so I don't really know what they're looking at or what the state of the art is. I just work in related stuff people in my lab use really simple photolith, electron microscopes, thin film stuff.
(To be kinda familiar with what's used now I recommend the wiki article, as always, http://en.wikipedia.org/wiki/Photolithography)

Moore's law has been surprisingly accurate, well beyond expectations. I, and most others IMO, don't think it will go on forever. Kurzweil will tell you otherwise...but he's not an EE or Physicist, he's a programmer.

[The Desert Fox]

I will probably upgrade, only because as a Best Buy employee Intel gives us an amazing accommodation one month a year that gives us a 2nd/3rd Gen i7 plus Motherboard for around 200$