That's correct for the 3DMark overall scores - all the other tests were measured with a 100% GPU limit and are therefore not affected by the CPU performance.

You can't avoid CPU core differences when you are measuring 5-10% differences. The test should be re-done with the cores disabled for better comparison.

Quote from: gc on October 10, 2012, 19:05:26
Interesting.  Some related questions:

- Does a larger cache enable an Ivy Bridge system to use (for office work)...
-- less battery power (because it doesn't have to power memory pins as often, e.g., for screen), or
-- more battery power (because it has to power more cache)?

That's difficult to answer. I suppose, a larger L3 improves battery life, because the performance gain should be higher than the increase in power consumption -> better battery life.

Quote from: gc on October 10, 2012, 19:05:26- Can Ivy Bridge drive the screen completely from cache when the screen is not changing?
-- and does screen resolution (pixel count) make a difference to this answer?

That's one of the new features of the upcoming Haswell generation (GT3 GPU with on-package memory only).

Quote from: gc on October 10, 2012, 19:05:26- When a laptop has mismatched memory capacity on its channels (8GB + 4GB seems popular),
does GPU memory come out of
-- the faster dual channel section (bottom 2/3) or
-- the slower single channel section (top 1/3)?

GPU-Performance with 4 + 8 GByte is almost identical to 4 + 4 / 8 + 8 GByte. That indicates number 1:  "-- the faster dual channel section (bottom 2/3)".

Interesting.  Some related questions:

- Does a larger cache enable an Ivy Bridge system to use (for office work)...
-- less battery power (because it doesn't have to power memory pins as often, e.g., for screen), or
-- more battery power (because it has to power more cache)?

- Can Ivy Bridge drive the screen completely from cache when the screen is not changing?
-- and does screen resolution (pixel count) make a difference to this answer?

- When a laptop has mismatched memory capacity on its channels (8GB + 4GB seems popular),
does GPU memory come out of
-- the faster dual channel section (bottom 2/3) or
-- the slower single channel section (top 1/3)?
(Note: datasheet seems to say that, to get dual channel bandwidth when the memory chips are different size, the larger memory chip has to be on channel A, and smaller memory chip on channel B.)