Originally Posted by Am*
I'm not buying into this.
These processors all have different HyperTransport/bus speeds, not to mention the completely different memory modules in each setup. For example, this is in their "test setup":
They're using 4 DDR2 memory modules in their older setups, which would stress the crap out of the IMC on both of the older AMD dual cores, yet they're using the optimal 3 modules for the tri-channel 1366 CPUs and 2 modules for dual-channel CPUs. That renders this article worthless, as well as the fact that there are even more variables to take into account, like those I mentioned above.
Also I don't see the Kuma in this comparison, which was a beast competing with quad cores/high end dual cores at the time in some applications. It beat my 965 BE at a higher clockspeed from what I remember, it should've replaced the 5400+ BE in this comparison.
Can people never be satisfied?
You can only take benchmarks so far. For all intents and purposes, a CPU "clock-for-clock" comparison assumes a "standard" system configuration. By your logic, you can never do a direct comparison between platforms unless it's simplified beyond the point that anyone would ever use it.
IMHO, if a system is configured as an end-user would have (i.e., if it's a dual-channel system and the test system runs in dual channel vs. a tri-channel and the test system runs in tri-channel), and then compared at the same clock speed with the same number of cores, that's about as fair as you can make it. What's the point in doing cross-platform studies and benchmarks with configurations that no normal user would have/use?
If you really need a "cpu-only" benchmark, then just compare superpi 512k times or something like that.