Any point to speedstep when overclocking?

[Comdrpopnfresh]

to the op: some boards disable automatic voltage adjustments when manually setting cpu voltage manually. I'm only familiar with core2 chips when it comes to this stuff.

What brand board do you own? Although many people hate manufacturer software, I've found using ASUS's epu software to be effective at reducing power consumption at load and during various performance states (multiplier greater than 6 and lower than max for a chip). You cannot, however, manually set the voltages in many such programs
If you don't want to settle for the AUTO voltage setting in bios (because it is higher than necessary), I'd suggest you check out crystalcpuid. It'll allow you to create three processor states of multiplier and voltage settings. Voltage adjustment and allowable voltage range vary with processors and motherboards; but, the program allows you to set aggressive voltages for the three multiplier settings that beats the conservative voltages of speedstep. The program does work with AMD cpus better due to the difference of where the performance states come from for AMD vs Intel chips. It'll let you set load percentages to increase and decrease the multipliers too. Check out the software documentation to modify a shortcut to set multiplier/voltage adjustment to auto, set the program to resident, and minimize upon start- then place the shortcut in the startup folder. Here's the url:
http://crystalmark.info/software/CrystalCPUID/index-e.html
Only thing is- setting a low and med multiplier with a high load required to increase multipliers may cause more power to be used through more clock cycles needed to complete a cycle.

Before settling on a method of reducing power consumption, I'd check out hardware monitor [ http://www.cpuid.com/hwmonitor.php ]. It's made by the same people that make cpu-z, and on many boards is able to provide cpu power consumption. Also, consider buying a something like the "killawatt:" a load and power metering device which reports total power consumption for a computer at the outlet. Using it, I've realized that unless heavily loaded, cpu power consumption plays a smaller role in overall power consumption than one may instinctively think, and that at idle it is very hard to attain lower power consumption with any one method over another (probably due to halting of clocks). It's also a good way to turn away from the habit of having an awesome overclock that provides little performance gain while shrinking your wallet due to electric bills. Beside reducing the voltage and frequency of the cpu, I've reduced the power consumption of my system significantly by bios modding my graphics card to lower, yet stable, voltages and through buying a motherboard with bridge chips made on a smaller process. Before these changes, my system typically consumed about 230W and up to 350W under load.

I've found the same measures to make a system energy efficient are often well aligned with getting good, stable, overclocks. Using a line-conditioning UPS, an efficient PSU sized to the components in a system, and good power management settings all work toward these goals. The end result is stable voltages with little ripple and line noise, less heat produced, and often better overclocks in terms of cpu frequency and voltage needed to get said frequency. Having the latest OS (whatever your choice is) helps to reduce consumption and provide better responsiveness from hardware too.

 

[lopri]

On topic: I'd definitely use SpeedStep as well as S3 for a 24/7 system.

 

[alyarb]

So Since when did AMD get intel's patents on the Hi-K gates?

http://download.intel.com/technology/silicon/HighK-MetalGate-PressFoils-final.pdf

Superior? Dont make me laugh.... seriously..

GF can use IBM's high-k process if they aren't developing their own. The big question isn't whose process it is, but why did they wait so long to implement it.

 

[Munky]

GF can use IBM's high-k process if they aren't developing their own. The big question isn't whose process it is, but why did they wait so long to implement it.

I don't think it's such a simple decision. I've read that Intel's 45nm process came before immersion lithography was mature, so they had to come up with other ways to push the envelope. AMD's 45nm process came later, and the use of immersion lithography made the need for hi-k gates less important, or that's how I understand it.

 

[Comdrpopnfresh]

I don't think it's such a simple decision. I've read that Intel's 45nm process came before immersion lithography was mature, so they had to come up with other ways to push the envelope. AMD's 45nm process came later, and the use of immersion lithography made the need for hi-k gates less important, or that's how I understand it.

Is it the same patented process anyhow? There's got to be differing ways to implement high-k gates. To your point- I thought the reason high-k was implemented at the 45nm node is because the gate density naturally increases leakage; though, I don't know much about immersion lithography vs what was used prior besides assuming it increases the resolution of etchings.

 

[Munky]

Is it the same patented process anyhow? There's got to be differing ways to implement high-k gates. To your point- I thought the reason high-k was implemented at the 45nm node is because the gate density naturally increases leakage; though, I don't know much about immersion lithography vs what was used prior besides assuming it increases the resolution of etchings.

Yes, gate density does increase leakage, but from my understanding, immersion lithography allowed AMD to continue into 45nm without hi-k gates, while Intel needed a new approach. I'm no expert in this field, though, maybe someone else will explain in more detail.