Looking into Intel’s new fab

The 40-minute video I linked to earlier with a tour by Intel Senior Fellow Mark Bohr has lots of interesting facts:

  1. The fab is three-football fields big and is the first of several fabs that’ll be exact copies of this one around the world (Israel and Arizona were specifically mentioned).
  2. It’s one of the cleanest rooms in the world.
  3. My camera is too “dirty” to get into the fab.
  4. Intel secret? They won’t tell me how many chips fit on a wafer. They also wouldn’t tell me how many chips a single fab could produce.
  5. 4:33: a piece of hair is huge compared to a nanometer. A red blood cell is 5,000 nanometers big. A transistor on the new chips is 45 nanometers.
  6. The real advance here is High-K dielectrics (link goes to Wikipedia). Good description of what that means is at 6:10.
  7. The new gate, above the dielectric (see technical discussion and photos about 26 minutes into the video) is based on hafnium (link goes to Wikipedia).
  8. Transistors up to now were created using silicon dioxide. Hafnium-based Metal Gates are much harder to manufacture, which is why it took so long. Intel is the first company to do this process.
  9. How many steps does a processor go through to be built? More than 50.
  10. How long does it take to make a 45 nm processor? “A few weeks.”
  11. There are eight copper interconnect layers on the new chips.
  12. The modern chip has 200 to 300 million transistors. The first chip Intel made had only 2,000 transistors.
  13. You can see the “clean/dirty” barrier in the cleanroom at about 11:10 in the video. On one side you need to be suited up and wearing booties to keep dirt on your shoes from contaminating the clean room, on the other, street clothes are OK.
  14. About a thousand technicians and engineers will work in the cleanroom, which works 24-hours a day, seven days a week.
  15. Most equipment is fully automated now. Workers don’t touch the wafers anymore like in the old days. A robot takes wafers around the fab line between “tools,” which, really, are entirely encased machines which do only a portion of the process needed to produce a processor.
  16. If you see a fab from the outside, 15:20, you’ll see a variety of pipes which mostly bring air into and outside of the plant.
  17. Mark has worked at Intel for 29 years and he talks at 17:00 about the differences between processor manufacturing then and now. Discusses how important “yields” are to processor manufacturers.
  18. Fibbing is an focused ion beam that engineers can use to “connect” small details on a chip. Mark says they don’t do that on chips they sell, but do use that on prototype processors to correct design errors for testing purposes.
  19. 19:46, discussion of what a “new fab” means.
  20. At 21:15 you get a look into the fab through one of several windows that exist on the side of the fab. I love the sign on the window. It says “no cameras.”
  21. The fab is three levels, we only got to see one (the clean room).
  22. 22:40, I asked Mark what he’s proudest of on the 45 nm processors. He answered the High-K dielectric which, he says, represents the biggest change in transistor manufacture since the 1960s.
  23. 22:30, discussion of “leakage,” which affects how power-efficient, and cool, a chip can be. Mark says that leakage has been reduced by a factor of five to 10 times. Translation: your battery will last longer!
  24. Discussion of competition at 25:20. “That’s some of the excitement of working in this industry.”
  25. 26:15 photos of “old, 65 nm” and “new, 45 nm” transistors and a technical, but understandable, discussion of how transistors work on a processor.
  26. Admission, at 28:30, that the first chips have already been produced and properly ran Windows.
  27. Intel, Mark says, has been working on this process for about three years.
  28. 30:30: discussion of the people involved.
  29. Discussion of key steps along the way. 31:21.
  30. At 36:30 Mark tells us what the secrets of Intel are.
  31. Mark takes us through some photos of older technology at about 38 minutes into the video.

What would you ask if you got a chance to get a tour of Intel?

Comments

  1. I think I’d want to ask each person there how they describe the significance of what they do, when trying to explain it to people who may not understand it. I think there might be some interesting answers and analogies and stories there.

    Judging by the number of posts on this topic, Robert I’m guessing it was really impacting for you to see all this. Here’s a question for you. You’ve seen a bit and interviewed a lot of people. How does this rate compared to other cool or provocative things you’ve been able to learn about?

  2. I think I’d want to ask each person there how they describe the significance of what they do, when trying to explain it to people who may not understand it. I think there might be some interesting answers and analogies and stories there.

    Judging by the number of posts on this topic, Robert I’m guessing it was really impacting for you to see all this. Here’s a question for you. You’ve seen a bit and interviewed a lot of people. How does this rate compared to other cool or provocative things you’ve been able to learn about?

  3. Toby: these people live in a world that I don’t and that’s always interesting. I mean, their world is so freaking small (just a few atoms across) and they need to understand materials and electrical properties that I just am not capable of learning anymore (and, even when I was younger, just didn’t have any skill in doing that).

    What was interesting about this is learning about what’s happening beneath my keyboard as I type these characters. I’m a lot more in awe now of the hardware folks than I was before.

    Out of the thousands of people I’ve met in this journey, Mark was among the most interesting. If Douglas Englebart was a “10″ I’d say Mark was a 8.5. I’d rank as a “2″ along that scale, though, so that’s pretty darn high.

    I’d really love to film a conversation between Mark and my dad, though. My dad would have some really interesting discussions about materials and tools and stuff, I’m sure.

    I’d also love to be a fly on the wall when Apple or Dell engineers come to visit with Mark (and when the business types get together and start negotiating over price and quantities).

  4. Toby: these people live in a world that I don’t and that’s always interesting. I mean, their world is so freaking small (just a few atoms across) and they need to understand materials and electrical properties that I just am not capable of learning anymore (and, even when I was younger, just didn’t have any skill in doing that).

    What was interesting about this is learning about what’s happening beneath my keyboard as I type these characters. I’m a lot more in awe now of the hardware folks than I was before.

    Out of the thousands of people I’ve met in this journey, Mark was among the most interesting. If Douglas Englebart was a “10″ I’d say Mark was a 8.5. I’d rank as a “2″ along that scale, though, so that’s pretty darn high.

    I’d really love to film a conversation between Mark and my dad, though. My dad would have some really interesting discussions about materials and tools and stuff, I’m sure.

    I’d also love to be a fly on the wall when Apple or Dell engineers come to visit with Mark (and when the business types get together and start negotiating over price and quantities).

  5. Regarding impact. My dad worked in this field and I know how excited he was when they turned on a new fab using a new high-tech process. I got that same feeling from Mark.

    I lived in Silicon Valley most of my life and I’ve never gotten to tour an Intel fab, so this was a real treat for me and something I realize doesn’t happen very often.

  6. Regarding impact. My dad worked in this field and I know how excited he was when they turned on a new fab using a new high-tech process. I got that same feeling from Mark.

    I lived in Silicon Valley most of my life and I’ve never gotten to tour an Intel fab, so this was a real treat for me and something I realize doesn’t happen very often.

  7. My husband worked for Intel for over ten years, until his death six months ago. He was a Mask Designer here in Hillsboro, on the team that designed the next generation of chips, over and over and over again. It was always awesome following the rapid achievements of Intel’s manufacturing capabilities.

    From a personal perspective, I just have say “Thank God for Intel.” My husband had melanoma that went to his brain. Intel’s health insurance was so good…they paid for everything without question, even “radical” treatments that cost $80,000 dollars a pop. Having such good medical care extended his life by at least a year. I wound up with cancer myself, and Intel paid for the first six months of my Cobra and extended a certain deadline for me that they didn’t have to, so that I can keep the insurance for three years instead of 18 months. Since I can’t work now, his Intel Life Insurance and what they put into his retirement account is what is sustaining me and our kids.

    The past two years have been hell, but having the Intel family’s moral support and the company’s financial support made things infinitely better than they could have been. I don’t know what we’d have done if he hadn’t been working for a great company like Intel who takes such good care of their people. I only wish that everybody could have such resources to draw on during tragic times.
    Some people around my husband left Intel to get larger hourly pay, but their benefits don’t compare at all, and a lot of those jobs have since folded along with the companies that hired them away. People should be careful about things like that, because you never know. My husband was just 46.

  8. My husband worked for Intel for over ten years, until his death six months ago. He was a Mask Designer here in Hillsboro, on the team that designed the next generation of chips, over and over and over again. It was always awesome following the rapid achievements of Intel’s manufacturing capabilities.

    From a personal perspective, I just have say “Thank God for Intel.” My husband had melanoma that went to his brain. Intel’s health insurance was so good…they paid for everything without question, even “radical” treatments that cost $80,000 dollars a pop. Having such good medical care extended his life by at least a year. I wound up with cancer myself, and Intel paid for the first six months of my Cobra and extended a certain deadline for me that they didn’t have to, so that I can keep the insurance for three years instead of 18 months. Since I can’t work now, his Intel Life Insurance and what they put into his retirement account is what is sustaining me and our kids.

    The past two years have been hell, but having the Intel family’s moral support and the company’s financial support made things infinitely better than they could have been. I don’t know what we’d have done if he hadn’t been working for a great company like Intel who takes such good care of their people. I only wish that everybody could have such resources to draw on during tragic times.
    Some people around my husband left Intel to get larger hourly pay, but their benefits don’t compare at all, and a lot of those jobs have since folded along with the companies that hired them away. People should be careful about things like that, because you never know. My husband was just 46.

  9. Thanks for the insight into the new fab. Sometime working here we get stuck in our silo’d areas of business and forget the incredible work others are doing here. I miss being on the engineering side but happy to be enabling some portion of the company. Thanks as well to Dawn for sharing her story. Best wishes.

  10. Thanks for the insight into the new fab. Sometime working here we get stuck in our silo’d areas of business and forget the incredible work others are doing here. I miss being on the engineering side but happy to be enabling some portion of the company. Thanks as well to Dawn for sharing her story. Best wishes.

  11. Thank you, Robert. Next time you’re in town I would love to take you to lunch or something. I know the best places around here.

    Thanks for your well wishes, too, Nathan. (Great name! We gave it to our son.)

    Praveen, the waffer is produced round, because Intel never cuts corners. HA! LOL Sorry…couldn’t resist. ;)

  12. Thank you, Robert. Next time you’re in town I would love to take you to lunch or something. I know the best places around here.

    Thanks for your well wishes, too, Nathan. (Great name! We gave it to our son.)

    Praveen, the waffer is produced round, because Intel never cuts corners. HA! LOL Sorry…couldn’t resist. ;)

  13. Video peek inside Intel’s new 45nm fab

    Normally I don’t like to honk the Intel horn here, as that’s why we employe a phalanx of marketing types, but Scoble posted a great video tour of our new 45nm fab and a breakdown of some of the interesting…

  14. Robert welcome to my world, I have worked in this industry for over 11 years now. Although it is interesting to see the 12 inch fabs as I was laid off during that transistion and the downfall of the industry a few years ago. 9.11 hit hard. Now back in the industry although working in a older fab now. (equipment still running dos or even windows 3.1)

    Praveen, calculate the area difference between a round wafer and a square wafer, in either case you will loose yield at the edge of the wafer. and you will fit more chips onto that round wafer then on a square wafer. Yield is the big deal here and surface area is king. Won’t go into specfics but take a 3mm exclusion zone on the edge and you will find that you will have significant more surface area on 300 mm diameter wafer then a 300 mm square.

    if there are 8 interconnect layers that would be a min of 16 steps there. and that is just the deposition steps. be it silicon or metal. typically there would start with a interconnct insulation layer, then an mask layer, an etch to make holes into that layer, a metal deposition, a cmp to remove the excess metal, a second insulation layer, another mask, another etch, and so on. this is highly simplified. as you have cleaning and metrology and inspections etc.

    I worked in CMP for 8 years. currently back in the industry working in EPI. Thanks for the show.

    One thing I wanted to add on the switch to a High K and different metal layers. Not only is it harder to do so it took a while it is expensive to make that change. An example would be, a typcial main manufactoring tool is in the million dollar range. and that doesn’t count running cost. retooling the process for that change is significant. when you already have a process that is effective. another transition that shows this is when the industry changed from aluminum and titanium to copper in the Interconnect layers. Or the move to Silicon on Insulator technologies, although much less so.

    Dawn thanks for sharing that gives me great satifaction that we do still have companies that do care about their workers.

  15. Robert welcome to my world, I have worked in this industry for over 11 years now. Although it is interesting to see the 12 inch fabs as I was laid off during that transistion and the downfall of the industry a few years ago. 9.11 hit hard. Now back in the industry although working in a older fab now. (equipment still running dos or even windows 3.1)

    Praveen, calculate the area difference between a round wafer and a square wafer, in either case you will loose yield at the edge of the wafer. and you will fit more chips onto that round wafer then on a square wafer. Yield is the big deal here and surface area is king. Won’t go into specfics but take a 3mm exclusion zone on the edge and you will find that you will have significant more surface area on 300 mm diameter wafer then a 300 mm square.

    if there are 8 interconnect layers that would be a min of 16 steps there. and that is just the deposition steps. be it silicon or metal. typically there would start with a interconnct insulation layer, then an mask layer, an etch to make holes into that layer, a metal deposition, a cmp to remove the excess metal, a second insulation layer, another mask, another etch, and so on. this is highly simplified. as you have cleaning and metrology and inspections etc.

    I worked in CMP for 8 years. currently back in the industry working in EPI. Thanks for the show.

    One thing I wanted to add on the switch to a High K and different metal layers. Not only is it harder to do so it took a while it is expensive to make that change. An example would be, a typcial main manufactoring tool is in the million dollar range. and that doesn’t count running cost. retooling the process for that change is significant. when you already have a process that is effective. another transition that shows this is when the industry changed from aluminum and titanium to copper in the Interconnect layers. Or the move to Silicon on Insulator technologies, although much less so.

    Dawn thanks for sharing that gives me great satifaction that we do still have companies that do care about their workers.

  16. @Praveen – I asked the question about why wafers are round (on behalf of one of my readers), and got some great answers in the comments:

    http://www.tinyscreenfuls.com/2006/12/why-are-silicon-wafers-round-instead-of-rectangular/

    @Dawn – I’m sorry to hear of your loss, and it makes me happy to hear that Intel took good care of you and your husband.

    @LayZ – Since 45nm was developed here in Oregon, maybe it would be more appropriate to rename the Portland area “Hafnium Valley” instead of the Silicon Forest. ;-)

    @Robert – great videos and posts – thanks! :-) I’m bummed that I missed you when you were up here. Definitely catch you next time!

  17. @Praveen – I asked the question about why wafers are round (on behalf of one of my readers), and got some great answers in the comments:

    http://www.tinyscreenfuls.com/2006/12/why-are-silicon-wafers-round-instead-of-rectangular/

    @Dawn – I’m sorry to hear of your loss, and it makes me happy to hear that Intel took good care of you and your husband.

    @LayZ – Since 45nm was developed here in Oregon, maybe it would be more appropriate to rename the Portland area “Hafnium Valley” instead of the Silicon Forest. ;-)

    @Robert – great videos and posts – thanks! :-) I’m bummed that I missed you when you were up here. Definitely catch you next time!

  18. [...] “>Looking into Intels new fab Yes. i really like it. I hope it hope people. 100% aggree with you. Good luck.The 40minute video I linked to earlier with a tour by Intel Senior Fellow Mark Bohr has lots of interesting facts: The fab is threefootball fields big and is the first of several fabs that#8217ll be exact copies of this one around the world Israel and Arizona were specifically mentioned. It#8…I think it’s good article.Link to original article [...]

  19. Wafers are round because they are sliced from a solid silicon ingot, which is formed by dipping a single tiny silicon crystal into molten silicon and slowwwwly pulling it out as the silicon crystalizes. Of course the natural shape will be a round cylinder, not square, thus the resulting wafers are round as well. The wasted portion of the wafer is fairly small and unavoidable.

  20. Wafers are round because they are sliced from a solid silicon ingot, which is formed by dipping a single tiny silicon crystal into molten silicon and slowwwwly pulling it out as the silicon crystalizes. Of course the natural shape will be a round cylinder, not square, thus the resulting wafers are round as well. The wasted portion of the wafer is fairly small and unavoidable.

  21. I read an article several years ago when Intel was talking about putting in the High-K at 65nm,and at the time, they said the dielectric constant they’ve got is a 3x improvement and that they wanted to delay introduction as long as possible. Back then, 2007 was a late target, but they hoped to hold it as a trump card until 2009. The key part of the article that jumped out at me is that Intel has ANOTHER dielectric they’re working on (2nd gen) that pushes the constant to 5x

  22. I read an article several years ago when Intel was talking about putting in the High-K at 65nm,and at the time, they said the dielectric constant they’ve got is a 3x improvement and that they wanted to delay introduction as long as possible. Back then, 2007 was a late target, but they hoped to hold it as a trump card until 2009. The key part of the article that jumped out at me is that Intel has ANOTHER dielectric they’re working on (2nd gen) that pushes the constant to 5x

  23. Doug Husemann,

    Can you explain in detail how you conclude that there is more surface area on a 300mm circle than on a 300mmx300mm square?

  24. Doug Husemann,

    Can you explain in detail how you conclude that there is more surface area on a 300mm circle than on a 300mmx300mm square?

  25. [quote]
    Doug Husemann,

    Can you explain in detail how you conclude that there is more surface area on a 300mm circle than on a 300mmx300mm square?

    Comment by D1D ENG — February 9, 2007 @ 8:04 pm
    [/quote]

    I don’t know what he’s talking about, but the reason that they use circular wafers is because that is the way the ingots are grown as someone said above. Also, the circular wafers are more durable as they resist shearing forces better than a square wafer would. Somebody please correct me if I am wrong about that.

  26. [quote]
    Doug Husemann,

    Can you explain in detail how you conclude that there is more surface area on a 300mm circle than on a 300mmx300mm square?

    Comment by D1D ENG — February 9, 2007 @ 8:04 pm
    [/quote]

    I don’t know what he’s talking about, but the reason that they use circular wafers is because that is the way the ingots are grown as someone said above. Also, the circular wafers are more durable as they resist shearing forces better than a square wafer would. Somebody please correct me if I am wrong about that.

  27. Having had an interest in Titanium for many years, I find your post an interesting read, Whilst I am not totally in agreement with some of your points, I must say it is well written.

  28. Having had an interest in Titanium for many years, I find your post an interesting read, Whilst I am not totally in agreement with some of your points, I must say it is well written.