Doug S (foo.delete@this.bar.bar) on January 31, 2013 12:00 pm wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on January 31, 2013 2:59 am wrote:
> > However, I don't think that many MicroCloud users will pick Celeron over inexpensive core-i3.
> > Core-i3-3220T looks especially attractive due to 35W TDP and ECC support.
>
>
> This is a perfect example of the kind of crappy artificial market segmentation that
> Intel will be forced to stop doing once AMD servers are a somewhat viable alternative.
> Even if they never amount to much market share, their mere presence will not allow
> Intel to force people into higher price points just to get basic stuff like ECC.
>

My mistake, sorry.
As pointed by Gabriele, now all desktop (i.e. Socket 1155) IvyBridge dual-core processors marked as core-i3, Pentium or Celeron (in other words, all dual-cores except core-i5-3470T) have ECC support enabled.

So, "stupid segmentation" w.r.t. ECC applies only to relatively expensive i5/i7 parts, but those have a price tag nearly identical to ECC-capable Xeon-E3 counterprats. Typically by buying Xeon instead of identically priced i5/i7 you only lose 3-6% of CPU frequency.

However, even without ECC advantage, I think there are good reasons to prefer $117 Core-i3-3220T as a server CPU over $42 G1610/G1610T - SMT, bigger cache, higher CPU frequency, support for higher memory speed. All those look like natural binning advantages rater than artificial de-featuring.

Now, there *is* artificial de-featuring in Celeron, but it applies to all cheap Intel CPUs, including not-so-cheap core-i3 - AES-NI is not supported. I suspect that it's a requirement of US government rather than Intel's own marketing decision.