FSB/Memory Ratio Effect on Performance?

[trainee]

I may be dating myself here, but in the past there has been a performance penalty for FSB/Memory clock ratios other than 1:1. Back in the day (1ghz athlons?) this was a major issue and running the memory faster than your FSB could actually reduce performance. Ive found very few modern references, but from memory reviews it appears that faster almost always == better for modern systems. Can anyone confirm this for me? I did a cursory search but did not find much on memory ratios affecting performance. Information regarding CAS/clock performance tradeoffs would be of interest to me. Thanks in advance.

[trainee]

In light of the newer Intel Chipsets having more restrictions on bus:memory clock ratios I thought I would revive this topic. I did find some benchmarks in my search for an answer to the original question in this thread, and most pointed to very little importance in 1:1 ratio. I will have to dig it up but an anandtech comparison showed that a cpu with 800mhz fsb performed better with memory at ddr-900, since the gains outweighed the loss of clock synchronization. However it was a very small margin, and most people seem to experience only minor 1-2% less performance running a 1066 core2 with 800mhz memory. So why is Intel intent on locking the ratios?

[adamsleath]

:para:
:confused:
i still use 1:1:yawn:

[BenchZowner]

The most important and performance varying factor with the Intel chipsets is that when you change the FSB/DRAM ratio the "Performance Level" ( internal chipset latencies, etc ) are being changed, which can cause performance enhancement or degradation.

Nice revival :D

[JVguest]

I'm running an 965 gigabyte ds3 and I don't notice any performance improvement from running 1:1 . It's always better to run the memory higher than the fsb since the CPU is be able to get lower latency to RAM the higher the memory speed is, with an increase in performance. Setting perf level in memset has the biggest effect on performance.