No - the point I am making is that a 12 - 29% increase in synthetic benchmark performance is great --- if you run benchmarks. The gains were made in specific areas of memory handling which make up just a tiny slice of a machine's actual memory performance. It's like saying that a car accelerates from 35 MPH to 38 MPH faster than another brand - the impact on overall driving is slim to none if all other areas are comparable.yenko said:The above is memory thread related.
The above is not.
When it came down to actual application testing, there was no net performance gain from 128-bit access.
XBench is simply a terrible predictor of actual performance.
In my experience, there is certainly more performance improvement from 1.5 Gb of unmatched RAM than 1 Gb of matched RAM.
When you choose the matching 512 Mb option, your opportunity cost is having to get rid of one or both 512's to upgrade later.