So why does this matter. In servers it's usually price/performance.

And reliability, which I wonder at as dies shrink and unpredictable phenomena at that size start to become more prominent.

They can deliver on this promise, as we hope and pray that this chip has enough horse power to kick Intel in the nads...

I hope DAMIT launches Shanghai better than their last launch. My nads are still hurting from Barcelona.

Smaller dies reduce the manufacturing costs by increasing the economies of scale (more chips per silicon wafer). Furthermore, they increase reliability (to a point) by reducing the amount of material the current needs to pass through (reducing heat generated through resistance). This allows manufacturers to either put more into the chip at no additional cost or lower its TDP. This is why 45nm Core 2's are so popular. Intel took the first route--they have more cache than their 65nm counterparts yet they still have the same TDP. Also, apart from the high-demand CPUs, they are priced at only slight premiums to their 65nm counterparts, making them very tempting steps up.

water cooled servers

"by reducing the amount of material the current needs to pass through (reducing heat generated through resistance)"

That is just a little bit backwards. Decreasing LENGTH of the lines would indeed decrease resistance. But the decrease in width and thickness (also part of that die shrink) will very much INCREASE resistance and INCREASE heat output, IF they keep the same voltage. Why do you think operating voltages of these things have dropped from 3.3V down to just one volt now?

In this case, the decrease in area (width x height) of the assorted elements more than offsets the linear decrease length.

And there is a WHOLE lot more to a new process than just transistor widths. Power consumption is going to be driven far more strongly by gate leakage and clock frequency that by simple resistance.

Heck, that is one reason the original Pentium 4 never made it that far: the .13 micron process SUCKED at gate leakage and they never could fix it. That leakage wasn't truely "fixed" until the 65nm process and it was greatly improved upon with the 45nm process.

So, just because AMD has a 45nm process, it doesn't automatically mean it will allow them to make lower TDP parts.

I hope they DO get it right though. Intel tends to become VERY complacent when they don't have a competitor nipping on their heels (or even passing them up like during the P4 days!)

The interesting point to watch for is going to be the leakage power - new process nodes are about much more than just shrinking geometry - if you just took bulk CMOS from 90nm to 65nm you don't save much because the process static power (leakage) is so high compared to dynamic power.

Modern 45nm is only low power because it has low leakage - and this comes from a lot of physical layer wizardry (strained silicon, silicon-on-insulator, high-K, etc). Get it right and you get great results - but there is a lot in there to screw up, and geometries are getting small enough that it isn't always easy to predict what works.

Another interesting point will be the overclock potential - Core2 is really shipped underclocked to give a green "eco friendly" TDP, but in reality the physical device is capable of going a lot faster. If Intel get pushed into a corner all they have to do is tweak their clock multiplier, and they have a core that goes 50% faster with air-cooling. If AMD can't match that they they'll have a problem with architecture scalability in the long term.

I'm all for a bit of competition, so good luck AMD =)

Trails? AMD doesn't even make a 45nm part yet. 4th Q.? They might try a vaporware shipment on December 31st...

AMD have a pretty reasonable record on deliveries to schedule, though they do seem at times to race to meet schedules and deliver a buggy first edition as they did with Barcelona. From a business-user perspective I'm not too fussed - Intel are far enough ahead at the moment that I won't be racing to try Shanghai upon release, though we might get a trial unit in from someone like HP if they have them in the summer next year.

As a home PC enthusiast, though, I'm hoping Shanghai is good as this year's upgrade was the first year I chose an nVidia graphics card over an ATi one for my games rig, and only a timely price-cut kept an AMD CPU on the shopping list. I'm hoping a good Shanghai launch will mean more time spent on the next gen of AMD desktop cores, otherwise the next upgrade will be the first time I have bought an Intel CPU for home use since the early P4 days!

Or not. But has anyone else noticed that rearranging Nehalem gives us "male hen?"