Silicon lottery

[Konrad]

I'm just going to post this one without much preamble, and invite your insights before volunteering any of my own:

Silicon Lottery

[d_stilgar]

I think it's interesting, and if you need to have a CPU that you know will overclock, then it seems like a decent way to go about getting one. As for this company in particular, someone will probably question their testing methodologies, and they have yet to make a strong track record, but that will come with time.

[x88x]

I would agree with d_stilgar's take on it.

The fact that they can reliably do this lends credence to my understanding of Intel's operating methodology. That is: anything that meets [x] minimum bin QA values goes into that bin, and the top bin has a lot of chips in it which can outperform the minimum requirements.

As a side note, I would not expect you to be able to obtain the same results with Xeon chips. Reason being, Intel does deals with certain big customers for special bins. Case in point the "custom" E5-2666 chips that Amazon has in their C4 hosts, which are actually just higher binned E5-2660's.

[Konrad]

Selling the best-overclocking parts at a premium. Selling the lemon-overlocking parts below base price. And apparently selling enough to be worthwhile - there seems to be demand for these after-binned parts.

Maybe you want the best of the best, price be damned. Maybe you just need a processor running at spec, maybe it's even already overkill, so why not pay less if you won't overclock it anyhow. I assume the Intel warranty is still valid, unless you order a modded or de-lidded part.

I'm thinking Intel might take some interest in this guy's approach. They already know their own yield figures well, so less guesswork and gambling, the odds are already in their favour. This sort of marketing could indicate new revenues which exceed Intel's already overpriced "Extreme" enthusiast-tier price tags.

[x88x]

I'm thinking Intel might take some interest in this guy's approach.

This may just be my bias talking, but I kinda doubt that there is a large enough market for Intel to seriously get into it. Keep in mind, they probably don't actually know which CPUs will fall into which of these higher level bins. Adding more stock bins would require more levels of QA; something that likely would just not be worth it for the relatively small niche market they would serve. At least that's what it seems to me.

[Konrad]

Yeah, probably correct.

And Intel might find themselves in a position where cheap low-clock parts oversell in high demand while a surplus of costly high-clock parts barely sell to a small niche ... leading back to underbinning the parts again, losing revenue, and giving the overclockers more goodies at the low-/mid-end ... vicious cycle.