Power Consumption

[mrow]

My point was that a chip that has, for example, a tdp figure of 35 watts may have an idle figure of 5 watts or so.  At that level the other parts of the system also make a big difference.  Yes, a processor with 17 watts tdp is likely to idle at lower power than a chip with 35 watts tdp if they use the same silicon manufacturing node and processes, but unless you take care over every other part of the system, the real benefits may not be achieved.

 

On your other point, I cannot speak for Intel (I was a Texas Instruments employee), but I doubt that "Intel mobile chips are just desktop chips binned as mobile.".  Substantially lower power requires design and manufacturing process changes as well as "binning" during the production  flow.

 

 

Intel has been doing it for over 2 decades. They pick the parts that can run at lower voltages and bin them as mobile. In fact, during the Pentium 4 days you could actually buy an Intel mobile Pentium 4 in a standard socket 478 and put it in your desktop motherboard and overclock the hell out of it. Bump the voltage only to that of a desktop P4 and the sky was the limit. I had a 2.4GHz P4 mobile chip running in a desktop board at 4GHz with low temps on air at voltages below what a stock desktop 2.4GHz chip used.

 

Honestly I'm surprised this is news to you. This is what AMD does for CPUs and GPUs and nVidia does for GPUs. You take the dies from the wafer and bin them based on their power and performance characteristics.

 

In all likelihood the Xeon E3-1220L v2 is a i7 mobile part with some server oriented features enabled. You can find i7 mobile parts with similar speed ratings, core counts, etc. both rated at 17 watts TDP. Maybe this is a coincidence but I doubt it.

 

Not sure why you put binning in quotes. If you worked in a semiconductor division of TI this wouldn't be an unfamiliar term. http://en.wikipedia.org/wiki/Product_binning

[S80_UK]

Not saying Intel don't - just that it depends on the goals, the volumes, etc.  There's more than one way to do these things - if there's enough latitude in the process and the results, then it makes sense.  It depends on the application - at the lower power end, transistor leakage becomes more of an issue and process changes can help where binning can't.  Binning in quotes - I know it's industry jargon - manufacturers use it all the time - some also use it in published data sheets.  It was just a personal choice, nothing more.

[garycase]

It's not just a binning process, as the low power units are throttled to the specified maximum wattage.

 

And there's VERY little difference in the power usage at idle.    I built an HTPC for one of my extra TVs a couple years ago using a 35w Core i5-2390T; but later decided I wanted to do a bit more with the box (some video transcoding), so I replaced the CPU with a Core i7-2700K.    My Kill-a-Watt measurements showed NO difference in the idle power consumption after the swap ... but a significant bump in max power consumption when doing CPU-intensive activities that consumed 100% of the CPU power.