IBM Beats Intel To 7nm Process Thanks To Silicon-Germanium Transistors, EUV Lithography

[whassup]

The way the IBM-led consortium that includes Samsung, Globalfoundries for combined R&D for new chip manufacturing node is best way to counter Intel's monopoly in the process node technology. Intel need to fight not one but combined 3 leading semiconductor manufacturers which puts a lot pressure for Intel in semiconductor field for sure.

This massive IBM-led consortium is a helping hand for AMD in the future.

 

[Guest]

AMD's chips have had higher, to notably higher, power requirements compared to Intel. The smaller you go, the more this gets to be a problem. I think they're in a significantly inferior position to move down to this size. (That is, beyond the fact that they've been a generation or more behind Intel with regard to fabrication techniques.)

Fab size is strongly related to power consumption. Lower fab = lower power requirement, ceteris paribus. A big reason why AMD uses more power than Intel is because of their process is so much larger.

According to IBM, going from 10nm to 7nm reduces power consumption by 50%. Ouch for Intel that just cancelled their 10 nm plans. Maybe they'll have to reconsider. Good for us.

 

[fixxxer113]

And the other question is, is this even the right path to follow? Are there other solution paths that will be better for the medium term? There are parallels with hard drives; rather than increase platter density, they added more platters, and then they developed RAID.

How about stacking whole CPUs? Imagine having a dual socket system using only one socket with 2 CPUs directly interconnected. Of course you would have huge cooling issues with one CPU hidden under the other, further away from the cooler... I guess you would need some kind of new material in coolers too, that can conduct huge amounts of heat very very quickly.

 

[cats_Paw]

Keep it coming.
I wonder when we will reach such a small size that it will be basicalu atoms 😀.

 

[mosu]

The main problem here is the small amount of Germanium available , with known resources in the hundreds of tons and being use for LCD screens.For even higher electron mobility I suggest GaAs based semiconductors.

 

[InvalidError]

Your really left with Intel on one side and everyone else, including IBM, on the other.

Even Intel cannot afford to do the fundamental process research on their own. Everyone in the semiconductor business is pouring research and money in a common pot for stuff like practical EUV sources, optics, mask-making, etc.

The main problem here is the small amount of Germanium available , with known resources in the hundreds of tons and being use for LCD screens.For even higher electron mobility I suggest GaAs based semiconductors.

There are quite a few issues with GaAs for use in mass-manufactured logic. The first one is that GaAs is more difficult and expensive to work with. Another big one is that while GaAs has faster electron mobility than silicon, it has slower hole mobility, which means less efficient P-FETs and by extension, less efficient CMOS logic.

 

[JackNaylorPE]

Would love to see an AMD apu built 7nm process. With 7nm process they'd be able reduce power consumption while increasing graphics performance and increasing the number of cores giving Intel a run for their money. How cool would it be to have 6 and 8 core apu's for desktops and mobile apu's that use half the power of current apu's resulting in longer battery life and cooler running.

I several articles, AMD is listed as one of IBM's customers who have reportedly "signed on" for 7nm.

 

[Blueberries]

They seem confident but electromigration at 7nm is going to be a factor, and I should put that in bold letters.

Still, I'll take 10nm sillicon with 16GB of integrated HBM in my home PC.

 

[whassup]

Would love to see an AMD apu built 7nm process. With 7nm process they'd be able reduce power consumption while increasing graphics performance and increasing the number of cores giving Intel a run for their money. How cool would it be to have 6 and 8 core apu's for desktops and mobile apu's that use half the power of current apu's resulting in longer battery life and cooler running.

I several articles, AMD is listed as one of IBM's customers who have reportedly "signed on" for 7nm.

Really ?? Could you mention the link ??

 

[merikafyeah]

I think it will take 10x longer to go from 7nm to 1nm than the time it took to go from 130nm to 7nm. Moore's Law as it was is over.

 

[JackNaylorPE]

Really ?? Could you mention the link ??

Saw 2 articles on Yahoo home page that I had read before visiting THG .... I'll see if I can find one .

There's some history there .... GF (formerly AMD's manufacturing subsidiary) will be making the chip for IBM. Alo remember that AMD abandoned the 20nm from TSMC and will be using Global Foundries for the upcoming 14nm FinFET solution.

Part of that shift was this month’s completion of the sale of IBM’s chipmaking unit to GlobalFoundries, a semiconductor foundry that was spun off from AMD back in 2009. GlobalFoundries is a member of the research group that successfully produced the new 7nm node chips, and it will serve as IBM’s exclusive processor manufacturer for the next 10 years.

Chip was developed in New York so I expect 4 things:

1. Case security will be by means of "The club"
http://www.theclub.com/

2. Trying to install incompatible hardware will result in a "Do that again and I kick you ass" popup

3. Users will get an urge to store their cordless mice by holding them to their shoulder and folding their T-Shirt sleeve over it.

4. BSOD messages will be replaced with "Fuhghedaboudit"

 

[palladin9479]

I think it will take 10x longer to go from 7nm to 1nm than the time it took to go from 130nm to 7nm. Moore's Law as it was is over.

Once your under 10~14nm Quantum Mechanics starts kicking your butt. Classic electronics theory really starts to break down at such incredibly small levels and you need to insert all sorts of safeguards or space things out. At 1nm you've entered the world of Quantum Computing as there is no way you could possibly predict, with any sort of accuracy, that an electron will be where you want it to be.

 

[Blueberries]

I think it will take 10x longer to go from 7nm to 1nm than the time it took to go from 130nm to 7nm. Moore's Law as it was is over.

Once your under 10~14nm Quantum Mechanics starts kicking your butt. Classic electronics theory really starts to break down at such incredibly small levels and you need to insert all sorts of safeguards or space things out. At 1nm you've entered the world of Quantum Computing as there is no way you could possibly predict, with any sort of accuracy, that an electron will be where you want it to be.

Do you have any references as to how these safeguards work? I'd love to read about the process and struggles involved.

 

[aldaia]

The main problem here is the small amount of Germanium available , with known resources in the hundreds of tons and being use for LCD screens.For even higher electron mobility I suggest GaAs based semiconductors.

Global proven recoverable deposits of germanium amount to 8,600 metal metric tons. Global production of germanium was estimated by the US Geological Survey (USGS) to be roughly 120 metric tons in 2011. An estimated 30% of the world's annual germanium production is recycled from scrap materials.
At current production levels germanium reserves can last almost 100 years.

IBM demonstration (wich acording to some sources is the first 7nm "comercially viable" technology) uses silicon-germanium (not pure germanium) as the channel material (not for the whole chip), therefore I expect germanium required will be relatively low.