New Chinese chips pose threat to Intel

[Ophir]

Originally posted by: DAPUNISHER
gadolinium oxide on a gallium arsenide semiconductor

Currently, most wireless applications use gallium arsenide metal-semiconductor field-effect transistors (MESFETs), which lack a gate oxide, but which are necessary because higher frequencies are attainable with gallium arsenide compared to silicon. (That's because electrons can travel five to six times faster in gallium arsenide-based transistors.)

Getting a complicated single-crystal gate oxide to grow on gallium arsenide was totally unexpected because scientists did not think it was possible. But the Bell Labs scientists made an attempt and succeeded. Then, when the two crystals came together perfectly, it was an even bigger surprise. "You would not expect two vastly different materials to have strong, complete bonds and a defect-free interface," said Bell Labs materials scientist Minghwei Hong. "We now are investigating why it happens."

That's 3yrs ago so it reasonable to assume that the tech is commercially viable now and it's not impossible that given it's superiority to silicon that the the claim is true and could explain why no special cooling is required to run@5ghz.

Bell Labs was using MBE for these growths and was growing them layer by layer, this is very slow and a far cry of the throughput necessary to make this process even remotely commercial. There has been much research done to this effect, especially in designing CVD processes for the gate oxide, but nothing has come close to becoming a reality - much less a 90nm reality. There's still too many problems with the deposition. I'd say this is still at least 2-3 years away. University research, even IF they have working substrates, is just that - RESEARCH. Trust me 'cuz I'm doing research in this right now, getting a working sample is really difficult; but, making the process viable commercially is VERY difficult.

 

[Naythn]

Don't Gallium Arsenide substrates release Arsenic gas when they overheat/decompose?

That would give me second thoughts.

 

[murphy55d]

Originally posted by: Naythn
Don't Gallium Arsenide substrates release Arsenic gas when they overheat/decompose?

That would give me second thoughts.

What's a little Arsenic gas when you can now run games at 3209834984638346 fps?

 

[Ophir]

Originally posted by: Naythn
Don't Gallium Arsenide substrates release Arsenic gas when they overheat/decompose?

That would give me second thoughts.

Within normal operating temp, thats a myth. AFAIK arsenic desorption from GaAs does not begin until well over 400C. Any temp even remotely close to this would be far outside a viable operating temp. The devide would surely fail before the GaAs structure would break down. Also, the substrate would most certainly be packaged (not exposed to atmosphere).

 

[Soulkeeper]

impressive !!!

 

[ProviaFan]

Echoing what a few others have said, I will believe it when I see it.

 

[Naythn]

Originally posted by: Ophir

Originally posted by: Naythn
Don't Gallium Arsenide substrates release Arsenic gas when they overheat/decompose?

That would give me second thoughts.

Within normal operating temp, thats a myth. AFAIK arsenic desorption from GaAs does not begin until well over 400C. Any temp even remotely close to this would be far outside a viable operating temp. The devide would surely fail before the GaAs structure would break down. Also, the substrate would most certainly be packaged (not exposed to atmosphere).

Ok, I didn't think it was quite as high as 400C, but that makes much more sense, I guess.

Since they are pin compatible with P4s, I guess that they may be able to take advantage of some of the thermal protection that P4s use. IIRC that's motherboard monitored, or something.

 

[BFG10K]

I'll believe it when I see it.

Also if the chip is from China then it'll probably have some kind of control features built into it to spy on users and/or limit what they're doing.

 

[magomago]

Don't doubt the Chinese- Who knows if this is real perhaps it will be free of the TCPA and Palladium...and since I'd import a chip I'd be free of chinese restrictions.

 

[Schadenfroh]

intel probably has alot of stuff they dont tell the world about. they will be ready for any type of competition.

 

[DAPUNISHER]

Thanx for the info Ophir, I was just working from the premise the announcement seemed to hint at which was that until then it hadn't been commercially viable but then read the quote "The gate oxide was the last castle to conquer in the gallium arsenide kingdom," said physicist Refik Kortan of Bell Labs, the research and development arm of Lucent Technologies. Kortan and his colleagues described their research results in today's issue of the journal Science. which seemed to be saying the last hurdle was behind them and it was just a matter of time till the process was perfected but I didn't realize the process was so incredibly slow.

 

[Snoop]

Originally posted by: Wingznut
1gb of L2 cache, eh? So, the die would be about 1000 times the size of a Prescott?

EDIT: I guess it wouldn't be "1000 times"... more like 600-700 times the size. I'm not even sure you could fit 1gb of cache on a single wafer.

They could name the CPU -Frisbee 1-

 

[Snooper]

1) The Chinese don't have the semicondutor knowledge to pull something like this off (Yet). They are still working on reverse engineering Pentium class CPUs the last I heard. You have to crawl before you walk.

2) GaAs DOES make fast parts. No doubt about that. To bad they can only make 4" wafers (They have been working on 6", but I don't think it is HVM yet) with it. No 8". No 12".

3) The cost of manufacturing GaAs wafers is a LOT more expensive than silicon based CMOS.

4) Even if they figure out a way to make it, you WON'T be able to afford it.

5) Just exactly who has 128 bit OS?

6) Even if it is P4 pin compatible, the bus is only 32 bit. Not going to work to well I would think.

Needless to say, I'm not worrying about Intel's future just yet.

What I AM interested in is quantum computing. That has the potential to turh the semicondutor world upside dow.

 

[Ophir]

Originally posted by: DAPUNISHER
Thanx for the info Ophir, I was just working from the premise the announcement seemed to hint at which was that until then it hadn't been commercially viable but then read the quote "The gate oxide was the last castle to conquer in the gallium arsenide kingdom," said physicist Refik Kortan of Bell Labs, the research and development arm of Lucent Technologies. Kortan and his colleagues described their research results in today's issue of the journal Science. which seemed to be saying the last hurdle was behind them and it was just a matter of time till the process was perfected but I didn't realize the process was so incredibly slow.

No prob. The gate oxide was indeed the biggest problem in building GaAs CMOS. This means that it takes the process from theoretical/impossible to doable. The MAJOR hurdle now is to scale the process up to commercial levels. The throughput necessary to maintain a commercial process financially viable is nothing short of incredible. That is one thing most researchers don't mention.

It is, however, a very exciting developement. If GaAs processing becomes commercial it integrates CMOS processing up to photonics and wireless applications - with every component on one die. Not only will this reduce cost, but size as well. Imagine a cell-phone with an integrated high performance processor, wireless LAN, and radio.

 

[AnitaPeterson]

Just as someone here posted earlier...

Even if the promises are unrealistic, the fact that someone else but the three (Intel, AMD, Via) would make processors could very well mean lower prices for all of us, as well as the end of TCPA, Palladium and/or the "Fritz" chip... Yes, last I heard they had 100 MHz processors, but that is a good start...

 

[Vash341]

What's a little Arsenic gas when you can now run games at 3209834984638346 fps?

The 3209834984638346 fps would mearly be the number you see due to all the breain damage form the arsenic.

On another note. Even if this story is correct its not like people dont already make crazy fast chips. My friends dad manufactures heatsyncs for the 15Gh chips used in apatchi helecopters. And if the chinese did come out with these chips and they some how became a threat to intell, as if intell wouldent advance, then i wouldent doubt that the government would put heavty fines on importing them, if not banning it. If they realy have the potential to release arsenic gas that would even give them an exuse to do it.

 

[Ophir]

Originally posted by: Vash341

What's a little Arsenic gas when you can now run games at 3209834984638346 fps?

The 3209834984638346 fps would mearly be the number you see due to all the breain damage form the arsenic.

On another note. Even if this story is correct its not like people dont already make crazy fast chips. My friends dad manufactures heatsyncs for the 15Gh chips used in apatchi helecopters. And if the chinese did come out with these chips and they some how became a threat to intell, as if intell wouldent advance, then i wouldent doubt that the government would put heavty fines on importing them, if not banning it. If they realy have the potential to release arsenic gas that would even give them an exuse to do it.

Please try not to post misinformation. GaAs does NOT produce arsenic gas under normal operating conditions. GaAs is currently used extensively in lasers/diodes, as well as HBT, BJT, and MESFET structures for communication/photonic applications.

Also, there is a BIG difference between 15GHz (or even 150GHz) MESFET, HBT technology and 5GHz CMOS. Just because you can produce one, in no way means that you can produce the other.

 

[grrl]

Sounds like another "Great Leap Forward."

 

[Sid59]

ooh .. my clicks for minesweeper will be faster with this chip, i can't wait.

 

[KpocAlypse]

"the chips will pose a significant threat to Intel's dominance"

But then what will it do to AMD?

Either way its intresting to see if they can make it, designing it in a lab is one thing, however building it is another.

Wouldn't be surprised to see the team designing this working for IBM, Intel, or AMD in a few years.

 

[DAPUNISHER]

Originally posted by: Ophir

Originally posted by: DAPUNISHER
Thanx for the info Ophir, I was just working from the premise the announcement seemed to hint at which was that until then it hadn't been commercially viable but then read the quote "The gate oxide was the last castle to conquer in the gallium arsenide kingdom," said physicist Refik Kortan of Bell Labs, the research and development arm of Lucent Technologies. Kortan and his colleagues described their research results in today's issue of the journal Science. which seemed to be saying the last hurdle was behind them and it was just a matter of time till the process was perfected but I didn't realize the process was so incredibly slow.

No prob. The gate oxide was indeed the biggest problem in building GaAs CMOS. This means that it takes the process from theoretical/impossible to doable. The MAJOR hurdle now is to scale the process up to commercial levels. The throughput necessary to maintain a commercial process financially viable is nothing short of incredible. That is one thing most researchers don't mention.

It is, however, a very exciting developement. If GaAs processing becomes commercial it integrates CMOS processing up to photonics and wireless applications - with every component on one die. Not only will this reduce cost, but size as well. Imagine a cell-phone with an integrated high performance processor, wireless LAN, and radio.

That would be sweet! and it makes sense the researchers don't mention the commercial viability is more than likely financially prohibitive as they'd risk having their funding pulled.

 

[Lizardman]

we will never see this chip on store shelves or at newegg. It will be not make it.

 

[clicknext]

Guess we'll see about these things. If it really poses a threat to Intel/AMD, the user could only benefit.

 

[pm]

There's a reason why everyone in the industry isn't using GaAs - it's not like it's a new and revolutionary technology that no one has ever heard of. Rather it's a old technology that has continued to be proclaimed "the replacement to silicon" and never gets off the ground because there are inherent material problems with the technology. GaAs wafers are brittle so wafers have to be small (6"), the wafers are expensive, yields are low, and power is typically high since there is substantial static leakage due to the inability to make MOS transistors. If they aren't using CMOS, then one has to wonder what they are using. MESFETs with direct coupled FET logic would be the typical guess based on the power dissipation goals (which are low based on the article).

Still, there are a number of claims in this report that makes me wonder if the Inq hasn't published another article that is only loosely tied to reality. I had thought Mike had been doing pretty well recently on the factuality of his data, but this one seems remarkably implausible to me.