Quote:
Originally Posted by Tsumi 
You are incorrectly understanding memory speed. Memory speed is based off of a multiplier from the BCLK. In the case of X58, the multipliers are in steps I think from 3 to 9. At the stock BCLK, which is 133 mhz, that would correspond to 800 mhz to 2400. At a BCLK of 200 mhz, that will give you memory speeds of 1200 mhz to 3600 mhz. Since the multiplier of 4 corresponds to 1600 mhz, you were perfectly capable of achieving 4 ghz with 1600 mhz RAM. In fact, that's what I did with my 920 system. In no way does lowering the RAM multiplier "bottleneck" the system. A multiplier of 6 with 133 BCLK vs a multiplier of 4 with 200 BCLK will net the exact same results. This is not the days of 775 where a higher FSB will generate better results.
What is semi-locked to the core speed is the QPI I think, which can either be run at full speed or half speed. It uses a fixed multiplier off of the BCLK.
With SB and IB, there was BCLK overclocking available, but due to the fact that SATA and PCI-E clocks were tied to the BCLK, you couldn't overclock the BCLK above 105 mhz or so. That's because the SATA and PCI-E devices will become unstable at clocks above 105 mhz. Therefore, the only method of overclocking was by multiplier overclocking on unlocked K processors.
You are incorrectly understanding memory speed. Memory speed is based off of a multiplier from the BCLK. In the case of X58, the multipliers are in steps I think from 3 to 9. At the stock BCLK, which is 133 mhz, that would correspond to 800 mhz to 2400. At a BCLK of 200 mhz, that will give you memory speeds of 1200 mhz to 3600 mhz. Since the multiplier of 4 corresponds to 1600 mhz, you were perfectly capable of achieving 4 ghz with 1600 mhz RAM. In fact, that's what I did with my 920 system. In no way does lowering the RAM multiplier "bottleneck" the system. A multiplier of 6 with 133 BCLK vs a multiplier of 4 with 200 BCLK will net the exact same results. This is not the days of 775 where a higher FSB will generate better results.
What is semi-locked to the core speed is the QPI I think, which can either be run at full speed or half speed. It uses a fixed multiplier off of the BCLK.
With SB and IB, there was BCLK overclocking available, but due to the fact that SATA and PCI-E clocks were tied to the BCLK, you couldn't overclock the BCLK above 105 mhz or so. That's because the SATA and PCI-E devices will become unstable at clocks above 105 mhz. Therefore, the only method of overclocking was by multiplier overclocking on unlocked K processors.
Alright I get it now, so does anyone know whether or not Haswell will be like SB/IB or the i7 920s?
