Being in 'the loop' has its advantages. More often than never, I find overclocking results/rumours in my mailbox indicating the capabilities of upcoming products; in this case Sandy Bridge. One of the things I regularly found reported was the strange behavior of the upcoming CPUs; rather than being limited in terms of overclocking capabilities, these samples were limited by CPU multiplier.

This is what I found on a forum recently:

The x47 multiplier is the maximum, this retail has the same problem as the ES: they are limited to a frequency and will not do more than this limit. The limit is different on every CPU

...

Yeah, some can do only x46 and some can do more. The 2600K D1 sample what I had did x51 as maximum, the best i5-2500K ES I had did x48. I don´t know what it is, you can boot also higher multipliers but at the windows loading screen the system freeze if a multiplier isn´t working.

...

I tested 8 2500K CPUs (1 retail and 7 ES) and all had a limit that doesn´t depend on voltage or cooling. For example, one sample can do 4800MHz (48x100) but can´t boot 4809MHz (48x100,2). And this can also not boot at x49 multiplier etc. I don´t know if it is a bios or cpu bug, but all cpus have a specific limit. ask the other guys that test the ES CPUs, they´re tell you the same. Btw, I tested also one with LN2 @ 1.6v VCore but no difference; reached same clock as on air!

In reports, people related the multiplier issue to the ES samples being locked by Intel. Turns out ... retail Sandy Bridge has the exact same issue. Ironically, the K-sku is said to be 'unlocked' for overclocking; with no scaling under cold by CPU ratio-limit I'd argue the term unlocked cannot really be applied here ...

In any case, this particular retail sample can run 4.7GHz:

- Chrisch's 3DMark06 32576 marks with 2x GeForce GTX 460 (256bit) at 925/1025MHz
- Chrisch's 3DMark Vantage - Performance 29257 marks with 2x GeForce GTX 460 (256bit) at 925/1025MHz