Going through Linux Mint "Cpu-z" style program, named "dmi-decode" just now, there might be another hidden limiter, that we do not normally see (or maybe its dmidecode's own "invention") --
Using the command:
sudo dmidecode | more
and looking through that, under the :
Handle 0x0004
Max Speed : 4000MHz
And the other givens
MinMultiplier: 12
MaxAllCoreMultiplier: 34
MaxMultiplier: 38
For my system the following seems to be true (take the values with a grain of salt, they are not necessarily set in stone):
Standard BCLK : 100
Max. (stable) BCLK : 105 (with stable means boot-up with 1 or fewer resets)
Max. (unstable) BCLK : 107 (with unstable means boot-up having to set-save-exit BCLK to 100, 105, and then 107, with upto 3-4 reboots between each set-save-exit cycle) (it is though stable enough when in the OS)
So, doing a little math we get:
MaxMultiplier * MaxUnstableBCLK = 4066
MaxMultiplier * MaxStableBCLK = 3990
This lends credit to the dmi-decode value of Max = 4000MHz (though this is only circumstantial evidence)
If we could somehow get the motherboard to supply 2 BCLK's for the CPU, 1 for the MaxMultiplier and one for the MaxAllCoreMultiplier as follows:
MaxMultiplier * BCLK_1 < 4000 -- here BCLK_1 < 4000/38 = 105.26
MaxAllCoreMultiplier * BCLK_2 < 4000 -- here BCLK_2 < 4000/34 = 117.65
If this was possible, and it really is this "internal max = 4000" that hinders further overclocking, one would be able to achieve quite a bit higher processing power.
Remember this is still highly speculative, and not at all proven.
Best regards : CyberstormXIII
<message edited by CyberstormXIII on Saturday, April 28, 2012 4:54 AM>
Motherboard : 1x EVGA Classified SR-X - BIOS 015, serial number ...007
CPU : 2x Intel Xeon E5-2687W SR0KG
GPU : 2x EVGA GeForce GTX 690
RAM : 12x Hynix DDR3 16GB 1333MHz ECC Registered Dual Rank
PSU : 1x Enermax Platimax 1500W
CPU Coolers : 2x Noctua NH-D14 SE2011
Cabinet : 1x Lian Li PC P80N Armorsuit (Black Anodized Aluminum)
SPU : 1x ASUS Xonar Essence One + Sennheiser HD800