Monday, January 14th 2013
BClk-based Overclocking Returns with Haswell?
With Intel's 2011-launched Core "Sandy Bridge" processors, Intel CPU overclocking as we know it changed. No longer could you overclock the CPU by stepping up BClk (base clock), a frequency that processors use to time various components, including the effective clock speed, and in some cases, memory, and uncore. Sandy Bridge left consumers with only one effective way of overclocking, stepping up an unlocked BClk multiplier, a feature only available with a handful expensive models.
According to a Hardcoreware report, when Intel took up the "one BClk to rule them all" approach with Sandy Bridge, it may have overlooked the possibility of the integrated GPU waking other components up from lower power states to use the L3 cache, affecting the chip's overall energy efficiency, which carried on to successive Core "Ivy Bridge" silicon. "Haswell" may present Intel with an opportunity to split core and uncore from sharing the same base clock, and as such it could be possible to crank up CPU clock speeds using BClk, without destabilizing the uncore. The author admits this is speculation on his part, but quite likely.
Source:
Hardcoreware
According to a Hardcoreware report, when Intel took up the "one BClk to rule them all" approach with Sandy Bridge, it may have overlooked the possibility of the integrated GPU waking other components up from lower power states to use the L3 cache, affecting the chip's overall energy efficiency, which carried on to successive Core "Ivy Bridge" silicon. "Haswell" may present Intel with an opportunity to split core and uncore from sharing the same base clock, and as such it could be possible to crank up CPU clock speeds using BClk, without destabilizing the uncore. The author admits this is speculation on his part, but quite likely.
36 Comments on BClk-based Overclocking Returns with Haswell?
The OP also neglects to mention that the primary reason for integrating the clock generator was to save costs and PCB space. Thus, I wouldn't be hopeful that even if there were two clock generators that they would be able to operate much out of specification (i.e. overclock).
I personally find an iGPU to be a great value add-in because I have had many times where a discrete GPU failed and that made the system basically useless, but with an iGPU I can keep the system usable until I get a new GPU. I don't need to keep a video card in my closet just for that scenario and I can cash in on any spare graphics cards I may have.
Intel = more $$$
I guess it does make sense.
Also, have I never noticed this before or is it unique to this post that a lot of people do not bother to read the entire thread before posting things that have already been said? I thought that was what the "Thanks" was for - to show you agreed with another post.
Only if they create a super mega overclocking SKU for LN2 OCing with a huge price premium...
If they had a separate CPU with no GPU you would moan about it being more expensive.
Having to produce one die for all is the cheapest way to go.
So I don't see why would you want to pay more for a different chip that would have identical performance.
Anytime you see an Intel chip with "X" in the name(x6800, qx9650, 980x, 3960x, etc.), you know it's hugely overpriced and is meant for that guy who doesn't want to buy a bunch of regular chips and bin them himself, and will most likely be run sub-ambient at least once.