Beginners: why more Ghz = less performance?

hello, first post.
i've been trying this basic overclock for an old AMD phenom 9750 non-BE CPU (yes i know..). i'm using AMD Overdrive and have pushed the HT ref clock from 200 to 225 (2.4Ghz to 2.7ghz). running in 225 HT ref is stabe. Prime95+Battlefield 3 or just prime95 run for 2 hours and more and maxing at 60C. yes i know it's hot, but as long as it's in the limits i'm going with it.

the thing is, it doesn't matter in the benchmarks. i'm using PerformanceTest and the built in benchmark feature in overdrive. sometime, pushing up the clock give worse results! what gives? i don't mind too much about the health of my hardware - it's old but i'm sure it will last forever. so i'm pushing it.

the only question at this point is why benchmarks are showing worse result with more Ghz compared to the stock HT ref. and if not worse, defentialy not far better.
also, how can i tell if memory timing is what keeping me back?

Thanks in advance
Tom

Its probably unstable at those clocks. It either needs to be tweaked with voltage adjustments, temperatures need to come down, or its just reach its overclocking limit on your cooling and cant handle the speeds.

As far as your memory holding you back, lower the RAM divider to rule it out. Quickest and easiest way.

that was quick.
how can i tell what is the best freq at the moment? 60C max is not instant but takes a while. and the cpu does read 2.7Ghz as opposed to the stock 2.4Ghz so i was surprised to not see performance gain.

what benchmark should i use for CPU benchmarking only?

That's the problem with software OCing, it's hard to really know what the software is doing. A slight increase in clock speed (which is what I would consider a OC from 2.4 to 2.7) can be slower than stock if memory setting are loosened way to much.

As for checking memory timing, the same tools for traditional BIOS OCing is recommended; CPU-Z, Prime 95 and HWMonitor.

so if the system is stable after CPU overclock then the only possible bottleneck is the memory? how do i calculate the memory divider change that i need for allowing the OC to work?
by the way, memory bottleneck shouldn't be with issues and errors other then just not improving the benchmark results? also, can voltage really change benchmarks?

zartom said:

so if the system is stable after CPU overclock then the only possible bottleneck is the memory? how do i calculate the memory divider change that i need for allowing the OC to work?
by the way, memory bottleneck shouldn't be with issues and errors other then just not improving the benchmark results? also, can voltage really change benchmarks?

Click to expand...

I'm thinking you are ready to dive into real overclocking and uninstall that useless Overdrive software. Time to do some homework, read, learn and try:

http://www.overclockers.com/forums/showthread.php?t=596023
http://www.overclockers.com/forums/showthread.php?t=686655

Voltage, in an indirect way affects clock speed, by allowing you to be stable at higher clock speed which affects benchmarks.

To answer your earlier question, a benchmark is a benchmark. Theoretically it should increase with increased system OC. If not then that's an indicator that something is not right. It's like driving, the faster you can go the faster you should get there. If the time took longer then maybe you too a wrong turn somewhere.

if anything read the articles that cigarsmoker put up and start to OC from the bios best recommendation would be to drop the memory clock ratio so the memory clock speed drops till you find you best CPU OC and after your satisfied with the temps and stability go back and work on your memory OC and decide were you want to go with that.

By separating your OC's you can get a good idea of where issues are coming from. and to me it sounds like the heat may be causing your issue it sounds like the CPU is throttling but im not sure if that chip has the ability to throttle itself.