Does higher FSB make CPU run hotter even though at the same Ghz?

[GundamF91]

I'm running stability test, partially to see if there's difference when I use different FSB and multiplier combo. My E4500 runs very well at 1.32v and 3.0Ghz.

With 335x9, I get 3015Ghz, and Orthos after 20min brought load temp to 50c.

With 275x11, I get 3025Ghz, and Orthos after 20min brought load temp to 46c.

So I'm wondering what's the discrepancy on this. Obviously, from a pure speed standpoint, faster FSB makes the whole system faster, including memory. So 333FSB would be preferred, but the 4-5c difference isn't a small deal. There must be some explaination. I'm curious to see if anyone else has seen this with their different FSB x Multiplier combinations.

 

[MarcVenice]

9x335 is going to be 1-2% faster in real world applications, IF you get that lucky, most of the time it won't matter at all. Just go with the highest multiplier. Btw, did you take into consideration that your case was heating up/ambient temps rose, after first running 11x275 ?

 

[GundamF91]

Yeah, I tried to take the "build up temp" out of equation, so I ran it yesterday, and ran it today with different FSB. The motherboard temp for both runs around 35c so I can be pretty sure the case temp is equivalent for both tests.

I'd like to use the full multiplier for 2 reasons, one I can use Speedstep so I don't always run the CPU at 3Ghz when it doesn't have anything to do. Also if I run 275FSB, I can use 1:1 divider with CPU:Memory, this way my DDR2-800 memory would be running at 550mhz, so I get get 4-4-4-12 timing instead of 5-5-5-15 that I run when it's at 670mhz.

 

[sutahz]

And running your ram at 4-4-4 vice 5-5-5 also gains you nothing in real world situations/benchmarks. But if your favorite games are SuperPi and PCMark, then disreguard this post.
As it sounds, your not going past DDR2-667 so if 4-4-4 works, may as well use it. Does speedstep drop the mutli down to 6x using either multi?

 

[GundamF91]

I can push the CPU:Mem divider to 1:1.50. But does it really gain any computing gain when the memory is outrunning the CPU? Without instruction, wouldnt memory just sit there?

As for Speedstep, I use RightMark CPU tool to control it. So I can set it as 6x or 11x or in between as I wish. The problem with having Speedstep is that on boot into Windows, it defaults to 11x overwriting the multiplier set at BIOS. So going into Windows, it's 11x at whatever FSB you have in BIOS. THus if I had set 333x9 in BIOS, it'll be 333x11 going into Windows. This doesn't work. So the only way is to use a lower FSB, unless you turn off Speedstep all together.