- Joined
- Nov 1, 1998
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
FRONTPAGE Is Intel's Building the World's First 450mm/10nm Fab?
- Thread starter Overclockers.com
- Start date
- Joined
- Feb 1, 2009
- Location
- Illinois
Sweeeeeeeet! 
- Joined
- Jul 31, 2004
Now if only Intel would quit skimping on virtualization features in their chips...
- Joined
- Jan 10, 2012
- Location
- clean room-grenoble (france)
@Culbrelai:
Decrease cost per chips comes from the fact that what the manufacturer pay is the processing of the wafer, not directly the chip. If the size of the wafer increases, then you get more chips per wafer that's to say more chips for the same cost
10nm node means lower power consumption per transistor. You usually want to keep this as low as possible while performing well enough. That's why it is quite common to see slow increase or even sometimes decrease in transistor counts.
Temps have to be considered regarding the overall design of the chip, not simply the transistor's count
Regarding the EUV, it seems they are still pretty far from the 250W target... Unless they were able to reduce the number of mirrors while keeping good resolutions, I don't see them ready for 2015 at industrial scale.
I would rather see Intel go for pillars/nanoimprint or even DSA instead of EUV for the next node since those technologies don't require major modifications on tools.
Decrease cost per chips comes from the fact that what the manufacturer pay is the processing of the wafer, not directly the chip. If the size of the wafer increases, then you get more chips per wafer that's to say more chips for the same cost
10nm node means lower power consumption per transistor. You usually want to keep this as low as possible while performing well enough. That's why it is quite common to see slow increase or even sometimes decrease in transistor counts.
Temps have to be considered regarding the overall design of the chip, not simply the transistor's count
Regarding the EUV, it seems they are still pretty far from the 250W target... Unless they were able to reduce the number of mirrors while keeping good resolutions, I don't see them ready for 2015 at industrial scale.
I would rather see Intel go for pillars/nanoimprint or even DSA instead of EUV for the next node since those technologies don't require major modifications on tools.
Last edited:
- Joined
- Apr 1, 2010
- Location
- Wakefield, QC
I need to add some precision to the article:
Intel's new 450mm fab will most probably be 10nm-capable or smaller when it starts production in 2015/2016 time frame. What is unsure is if ASML will have Extreme UV equipement by then. If not, Intel will have to adapt current methods (double patterning with standard UV lights) for its move to 10nm with quadruple patterning/multiple exposures. This could extend current manufacturing methods and equipments for a couple more years but the move to shorter wavelengths will have to happen sooner or later. 10nm is possibly the last process node that could be manufactured using standard 193nm UV light.
The move to 450mm is a huge investment at first but will, over time, give a massive advantage to Intel in the world of foundries. Other players will have to follow if they want to stay competitive.
Intel's new 450mm fab will most probably be 10nm-capable or smaller when it starts production in 2015/2016 time frame. What is unsure is if ASML will have Extreme UV equipement by then. If not, Intel will have to adapt current methods (double patterning with standard UV lights) for its move to 10nm with quadruple patterning/multiple exposures. This could extend current manufacturing methods and equipments for a couple more years but the move to shorter wavelengths will have to happen sooner or later. 10nm is possibly the last process node that could be manufactured using standard 193nm UV light.
The move to 450mm is a huge investment at first but will, over time, give a massive advantage to Intel in the world of foundries. Other players will have to follow if they want to stay competitive.
- Joined
- May 10, 2009
Smaller transistors can often switch faster, but not always. They don't directly increase performance.
What they do increase is how many transistors can be packed onto a given size slab of silicon. More transistors to play with = better clock/clock performance.
What they do increase is how many transistors can be packed onto a given size slab of silicon. More transistors to play with = better clock/clock performance.
- Joined
- Jul 31, 2004
Translation: Intel is a drag queen.
- Joined
- Nov 19, 2011
- Location
- The Cold North
Lower costs for Intel! Goody that means there CPU's will be less expensiv.... Oh wait nvm
- Joined
- Oct 25, 2012
So this is why Intel will keep beating AMD?
Their transistor density is better? I always get confused when it comes to AMD vs Intel, Intel has lower clocks but better efficiency because there are more transistors in their cores than AMD? This must be where I've heard people say "AMD cores are not real cores"
Shorter version of what I said, and what you said...
- Joined
- Apr 1, 2010
- Location
- Wakefield, QC
It is said it costs $300-400 million to develop a new CPU architecture from the ground up. Intel has tons of money to invest in development, AMD sold their HQ a few months back. So yeah... Intel has better CPUs.
I think it would be interesting to see what AMD could do with if they could manufacture at the same transistor size as Intel. But that won't ever happen.
I think it would be interesting to see what AMD could do with if they could manufacture at the same transistor size as Intel. But that won't ever happen.
- Joined
- Jul 31, 2004
- Joined
- Apr 1, 2010
- Location
- Wakefield, QC
