Bad benchmarks with Cool n Quiet enabled...

[mvbighead]

Recently decided to delve into OCing my CPU, and found it really easy to overclock my AMD Phenom 2 X3 Black Heka 740 to 3.5GHz with all 4 cores enabled. Board is an ASUS M4A88TD-V EVO/USB3, and cooler is a Hyper 212+.

However, I performed two tests, one with Cool n Quiet enabled, and the other with it disabled. The test with it disabled showed double the performance of the one with Cool n Quiet enabled. (24000ish vs 11000ish.)

My understanding is that Cool N Quiet should throttle down when the CPU isn't needed, but should jump up as the processor is needed. It seems to be doing that some, but only bumps the multiplier up to 12 or so instead of the 17.5 I have it set to for the OC.

Should I just disable Cool N Quiet altogether? Or is there perhaps a power setting in Windows that would allow it to work, but actually reach the max OC when the processor is needed? I would prefer to use it whenever possible to reduce the heat build as I don't game all that often, but if I must disable it altogether, I suppose that's what I'll do.

 

[podspi]

First of all, what benchmark are you referring to? Windows 7 does have the ability to change the processor speed, but you shouldn't be seeing such a huge difference just from CnQ. Even back in the Phenom 1 days I don't think CnQ incurred that large of a performance hit...

 

[mvbighead]

First of all, what benchmark are you referring to? Windows 7 does have the ability to change the processor speed, but you shouldn't be seeing such a huge difference just from CnQ. Even back in the Phenom 1 days I don't think CnQ incurred that large of a performance hit...

Benchmark was 3DMark2005. Just wanted something simple to test with.

As for changing processor speed, the power settings typically affect how CnQ is used. IE - Max processor state/Min Processor State, etc. in the power settings. If a system is set to max performance, a feature like CnQ (or Intel's speedstep) is never used and the processor runs at max the whole time.

My issue is that I'd like to run CnQ, but have it allow the high performance when in use. What I am seeing is that CnQ seems to work, but doesn't not let the processor go up to the max when a core is used. It simply goes up from something like 800mhz to 2.1GHz.

Problem being, CnQ works, but not optimally from what I can tell.

 

[dorion]

Use a benchmark that puts stress on all cores. Or set the affinity on your benchmark to one core.

I say this because it sounds like you are only stressing one core, don't know how well 3dMark05 is multithreaded.

 

[mvbighead]

Use a benchmark that puts stress on all cores. Or set the affinity on your benchmark to one core.

I say this because it sounds like you are only stressing one core, don't know how well 3dMark05 is multithreaded.

I guess the point, however, is that the intended design of CnQ is to increase speed when necessary. Any benchmark that pushes the CPU to 30% or so, IMO, would constitute a load worthy of pushing the CPU up.

While I understand what you are getting at, my issue is with running a game such as Mafia 3 or anything else, and suffering with low frame rates because the CPU isn't adjusting properly. If my CPU is 99% idle, I can see leaving it throttled to 800Mhz. But when it is under some level of load, it should jump up... no?

With those things in mind, I am just curious what others system's do when CnQ is enabled (or speedstep for that matter). I was using CPUz and Taskmgr to monitor the system while 3DMark05 was running, and could see where the CPU was throttled while the tests were being run. The system starts just fan and is running at 3500MHz when CPUz is first loaded, but drops down and never resumes the full OC when the benchmark runs. It seems to stick around 2 GHz, and the FPS on one test was at about 30fps vs the 80 or so it was getting with CnQ off.

Thanks all for the responses thus far.

 

[podspi]

That's very strange, you might want to make sure that it isn't something else, or maybe a bad test run?


I've done similar tests on my Phenom II and I've found CnQ does not have a large effect on performance, and, surprisingly, TurboCORE really does work (even though I've never seen it kick in, benchmarks go up by the appropriate &#37. Try running each test a few times, could have just been a bad run.

Also, have you tried updating your BIOS? It could be that the OC is doing something to mess up CnQ.

 

[mvbighead]

Yep, updated BIOS prior to OCing as I hadn't done it in a while. Sitting here watching CPUz, and it just doesn't seem to be kicking up as it should. Hmmms.

 

[(sic)Klown12]

Turn off Cool 'n Queit in the BIOS and download PhenmoMsrTweaker. It gives you full software control of the power states and lets you customize it to fit your personal situation. I now use it on every AMD system I build. With the proper settings, you can make it so you loose no performance when needed yet still save a lot of power(and heat) when idle.
http://phenommsrtweake.sourceforge.net/

 

[podspi]

Turn off Cool 'n Queit in the BIOS and download PhenmoMsrTweaker. It gives you full software control of the power states and lets you customize it to fit your personal situation. I now use it on every AMD system I build. With the proper settings, you can make it so you loose no performance when needed yet still save a lot of power(and heat) when idle.
http://phenommsrtweake.sourceforge.net/

I see you have an X6, I've always wondered how phenommsrtweaker works on Thuban. Does it let you configure Turbo states as well?

The newer AMD CPUs have pretty good power management, so I've been hesitant to mess with them manually as I used to (anybody remember rmclock?!)

 

[(sic)Klown12]

Yes, the newest version allows you control of Turbo. I don't use it myself, but when I first started tweaking I played around with it and it worked perfectly.

I've been using PhenomMsrTweaker since the PII 940 was released and I haven't had any trouble with it. It is a really great piece of software.

 

[mvbighead]

Thanks all for the help. I may have to give that software a try.

I did end up resolving the issue without the use of that software by changing the power profile from power saving to something less power saving. (Not max performance, but normal or something to that effect.) Anyhoo, it now clocks down normally and hits 3.5 in benchmarks while running under some level of load.

Thanks all for the help.