Small note: Intel
revised their performance improvements to 6-8%. the 20% increase in CPU performance was either misinterpreted or put in my some other previewer during IDF; quite possibly the latter, since what was said at IDF was "When Intel introduced its 22nm tri-gate transistors Intel claimed that it could see an 18% increase in performance at 1V compared to its 32nm process."
It is likely Intel could just lower operating voltage while keeping performance mostly the same. Ivy Bridge is a 'tick' release and we should not be expecting very much performance improvement over Sandy Bridge; just getting 6-8% more overall performance at lower operating voltage and power is pretty nice though.
What excites me more about Ivy Bridge is the
promise of 35W quad-core i7s; we might potentially see the dual-core parts drop to 25W TDP or thereabouts without having to go to lower voltage ranges. By the time Haswell rolls round, we might be seeing the quad-core desktop parts at 45W TDP, with "lower-power" parts (corresponding to the S- and T-models for Sandy Bridge). going to 20W TDP or so. That's definitely eating into the low-power market, which means Brazos won't be unchallenged for long—Intel
pulling the Atom product line was a step in this direction. [my bad; Intel isn't pulling the Atom line, just the independent dev team.
Core and Atom will be developed under a unified design archi, which means we'll hopefully see more Core-like performance on Atoms.]
As for graphics,
As a result Intel is expecting a 60% increase in 3DMark Vantage scores (Performance Preset) [from GT2] and a 30% increase in 3DMark '06 scores. IVB GT1 on the other hand will only see performance increase by 10 - 20%.
That 60% estimate applies only to the 'higher-end' Intel IB IGP.
I'm afraid AMD has already stumbled on the first column of that block; Performance-per-Watt (PPW because I'm lazy to type) is hardly up compared to Thuban. When Intel released the P4 with eerily similar prescriptions (software has to be tuned for this architecture for us to see improvements [...] this is a 'forward-looking' architecture), they were not so quick to acknowledge their mistake; it was an 8-year mistake which eventually led to nice things (Conroe). I sure hope AMD isn't going to make us wait that long ... if they can even afford to wait that long with the slim profit margins on their products. I don't expect them to match Intel anymore, but they're going to have to show us that 10% PPW improvement if we are to trust that chart. In the meantime, lay off the mushrooms and the delusions, AMD, and either get a proper marketing team, or stop mentioning Intel in your PR.
