Have you tested the new memory by itself - as in, replace the existing RAM with the new RAM (in the same DIMM sockets as you original RAM) ?
That aside, I'm not sure how well those programs are written with regard to their immunity to the effects of the CPU caches. So if you feel like tinkering, you could test the programs by tinkering with the RAM setups.
For example, the motherboard uses the P35 chipset, which supports dual channel operation, when DIMM is installed in the approriate sockets (see the manual, page 16). So in order to test the testing program's immunity to caching effects, you need to test with varing RAM capacities.
As you only have 1GB modules, then that means tesing with 1GB, 2GB, 3GB, and 4GB. The problem though, is the dual channel business. The motherboard will configure itself for dual channel operation, when it can. So testing each of those capacity points is not going to be testing just the programs immunity to cache effects. But it should give us an idea testing 1GB and 2GB single channel, plus 2GB and 4GB dual channel. You already have the dual channel numbers, so all you need to test now are the single channel configurations.
What I expect to see is that the single channel performance should be slower, but not twice as slow. Moreover, there will be a similar discrepancy in the throughput rates between the 1GB and 2GB single channel figures, as there are between the 2GB and 4GB dual channel figures.
As for the the 3GB capacity point, it would be interesting to see both what the motherboard does with regard to dual and single channel operation. But more importantly, what the testing programs report.
EDIT: Oh, and if those programs don't already tell you this, you should confirm the memory modules are compatible with each other. ie get their SPD details, etc. I usually use [CPUZ] for that.