Doug:
You are sorely mistaken,
Shanghai will be on 45nm, going to be cheaper.
This is not the case until such time as AMD yields more die per 45nm wafer than they do from a mature 65nm-process wafer. In addition, the 45nm process using immersion may be more costly per wafer.
Intel uses a DEDP (Double Exposure Double Patterned) method to extend its 65nm tools to do a bastardized 45nm process, really a 35nm by 65nm process, trading 9 to 12 months quicker TTM, Time To Market, for lower wafer throughput, due to many more steps, higher required alignment (more time per scan step), more time for each scan because of twice the exposure time, and less yield per wafer again because of those extra steps. All this compared to standard processing using 193nm immersion tools that AMD is using. Yes, Intel gained some back by having a lower die size compared to its 65nm process, but they lost some more by needing even more steps for HiK/MG. All in all, they likely get less good dies per month at equal clean room size than AMD does, even if AMD goes to HiK/MG themselves.
And if you don't take my word for it, links I provided when Intel's process first appeared in production, ASML says the exact same thing with a more detailed presentation as does the SIA.
Shanghai is already in production.
AMD's definition of "in production" is... rather liberal. AMD hopes to ship for revenue in early Q4.
So is Intel's as they are paper launching Nehalem at the same time. Which by most people are far more liberal than AMD's. AMD's Shanghai will be quite available at launch. Perhaps even like the Radeon 4850s and 4870s that were available at retail before the official launch. Nehalem will be more like a promise to be available sometime in the future.
Shanghai consume less power.
It does appear to consume less power at 2.3GHz. Beyond that, we don't know at this time.
Yet Nehalem which is to paper launch at the same time, isn't known at all and due to that, likely to be much higher given that it would be touted all over the place, if it was lower. Some estimates have it higher than Barcelona at the same clock.
Indication from several Chinese sites it will run at 2.8Ghz or more.
Link the "several" sites, please. I've seen one experiment showing that with one core stressed, it completed a benchmark at default voltage at 2.8GHz.
Even less is known about Nehalem. Yet you tout it as if it was the second coming.
Intel reaction. Nehalem(2.6GHz) at $286
Why must this be "a reaction"
Nehalem is being paper launched. That shows that they are reacting, not setting their own pace. They must be very worried now that they have seen tests by third parties. If they weren't, they would wait until they could do a true launch.
Intel bring forward Nehalem launch to September even though Nehalem is not ready for launch.
I can make things up, too!
You are doing that too much lately as it is. The dearth of benchmarks and the careful tailoring of those that were done away from certain aspects smacks of problems when they actually appear at retail.
Current Core2 Quad @3.2Ghz is selling for $999. Why does Intel want to sell Nehalem at $286 during launch.
Because as time goes on, Intel increases performance, and prices waterfall? By your "logic", Core2 pricing should start where the highest P4 price left off, and then some. The market doesn't work like that.
Unless Nehalem isn't up to snuff when it finally does appear. Ever think that the price is an actual reflection of its true performance wrt current C2Q on stuff people care about?
it does not make business sense
You don't appear to understand that demand is price-sensitive. Another way of saying this is that CPU ASPs over time have not risen in direct parity with CPU performance.
It does make business sense, if Nehalem was known to be slower than a 3.2GHz C2Q. A couple of bins is enough to drop it that far. Intel is well known to cherry pick the parts it sends out. You just don't like the implications of Nehalem being really worth only $286 to the current market.
Pete |
