Looking at them from an enthusiast's perspective
Our benchmarks have concentrated on memory performance with a 3.20GHz Core i7 CPU with a standard 133MHz BCLK. Perhaps we're looking at enthusiast memory the wrong way? To this end, the CPU was setup with a 3.6GHz core speed, 200MHz BCLK (18x multiplier), and a x8 memory-multiplier, giving a memory frequency of 1,600MHz.
Using the same basic setup, a DDR3-1,333 CL9 6GB pack was installed and the memory-multiplier dropped down to the nearest (lower) frequency, using a x6 to hit 1,200MHz. The point here is was see how much of a difference the memory frequency makes when the CPU and BCLK speed are at levels that enthusiasts would comfortably push their kit to.
| Sisoft Sandra 2009 SP2 (15.72) - float buffered (200BCLK) | |
|---|---|
| Crucial Ballistx DDR3-1,600 C8 | Crucial DDR3-1,200 C9 |
| 28.65 | 21.3 |
| HEXUS.PiFast - calculation to 10m places (200BCLK) | |
|---|---|
| Crucial Ballistx DDR3-1,600 C8 | Crucial DDR3-1,200 C9 |
| 24.35 | 24.53 |
| DivX 6.8.3 + LAME encoding (200BCLK) | |
|---|---|
| Crucial Ballistx DDR3-1,600 C8 | Crucial DDR3-1,200 |
| 166.3 | 166.49 |
| RAM: ET: QW - 1,024x768 0xAA 0xAF low - (200BCLK) | |
|---|---|
| Crucial Ballistx DDR3-1,600 C8 | Crucial DDR3-1,200 C9 |
| 230.6 | 219.9 |
With a 400MHz frequency deficit and a high BCLK and CPU frequency to make the most out of the RAM, we, again, see little difference between results, suggesting that it would be prudent to buy a cheap-ish DDR3 kit and spend the money elsewhere. Indeed, unless going for an outrageous overclock where every last bit of speed matters, DDR3 scaling doesn't appear to benefit Core i7 in any meaningful way. Let's not forget that we're comparing a £300 kit with a £91 pack - run at lower-than-default speeds - in the above graphs.