Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
AMD

Behind the Closed Doors of AMD's Chip Production 151

rokali writes "Tom's Hardware is running an article on AMD's chipmaking procedure, plants, and future. Check out the pictures of Fab 36, their new plant slated to open in 2006, which will put of the next generation of 65nm chips. From the article: 'Currently, AMD's devices in Dresden are still produced on 200 mm wafers; the new APM 3.0 using 300 mm wafers won't be ramped up until Fab 36 opens. Production startup at the new facility is slated for the beginning of 2006, at which point the company will have invested an additional $2.5 billion.'"
This discussion has been archived. No new comments can be posted.

Behind the Closed Doors of AMD's Chip Production

Comments Filter:
  • Motherboards (Score:5, Interesting)

    by superpulpsicle ( 533373 ) on Tuesday April 19, 2005 @06:38PM (#12286920)
    How about manufacturing AMD motherboards. The Intel chip + Intel board is a ridiculously stable combination. AMD should have a combo of their own to counter.

  • Re:Question (Score:3, Interesting)

    by BlacBaron ( 875559 ) on Tuesday April 19, 2005 @06:43PM (#12286970) Homepage
    Not sure how related one could consider this, but it might explain why they chose Dresden and not some other german city.

    History [bbc.co.uk]

    As a result I believe it was rebuilt to be a rather industrial place.
  • Here's a question... (Score:5, Interesting)

    by DrKyle ( 818035 ) on Tuesday April 19, 2005 @06:45PM (#12286993)
    Why is the building so darn yellow inside? Is it important for the process, the workers, the ability to keep the environment clean? It's just so yellow, I think I'd get a huge headache working there.
  • by tofucubes ( 869110 ) on Tuesday April 19, 2005 @06:52PM (#12287065)
    dual cpu and gpu...
    I'm betting the same people who bought SLI configs are going to buy dual core...
    the problem with dual core vs. SLI is that people can buy one video card now and one later...
    which is not the case with dual core

    anyway I wonder if this all started people buying two of the same ram modules for more bandwidth performance

    and I wonder if this trend will continue?

  • by Blapto ( 839626 ) on Tuesday April 19, 2005 @06:53PM (#12287071)
    True fact:
    When I was at school, the walls were painted "bright spark yellow". According to our teacher, studies had been done and it was found that this particular colour made people think more productively. He had entire studies to give us and everything, being 11 I'm not quite sure what we did with them.
  • by G4from128k ( 686170 ) on Tuesday April 19, 2005 @06:58PM (#12287112)
    Looking at the die layout, its easy to imagine that AMD (and Intel) will be produces a good many dual-core chips with one defective core (maybe 10-25% of production). I'd bet that somebody finds a market for those partially-functional chips. I also wonder what will happen when people discover that one core can be overclocked more than another core. For applications/loads that only use a single core, the system could disable the slow core and run the fast core at full speed.
  • by ciroknight ( 601098 ) on Tuesday April 19, 2005 @07:03PM (#12287149)
    The real problem is, AMD's Opteron will probably be done and shipping by the time Intel gets 64-bit dual core Xeons out the door. Not that they couldn't go ahead and shift all of their production capacity to dual core now, and have early chips ready by the end of this year, it's more like they won't.

    More and more I get my hopes up that Intel is doing research into a 64-bit enhancement for the Pentium-M, and I believe this to be the only reason we haven't seen Dual Core Pentium-M's yet. We're just now starting to see a move for the Pentium-M to the desktop, which is a good start, but without the cutting edge memory controllers present on new chipsets, it doesn't stand a chance.

    I believe Intel is also probably investigating adding memory controllers to their next Xeon line, which is definitely going to extend the amount of time in which we expect to see it. Intel really would see this as defeat, but as DDR2 becomes prime, Opteron's with DDR2 controllers will be able to completely smash any Intel offering, simply because it can get the data faster, get it processed, and pumped back out, while the Intel chips still wait for the laggy north bridge memory host to allocate the resources.

    Reliability will always be in Intel's court, simply because they control all factors of production, beginning to end. AMD's trying to take this approach, and by opening new fab facilities, maybe they can get into competition in other chip segments (like the Turion vs the Pentium-M). It also doesn't help that AMD is no longer making chipsets, but I believe a new fab facility will open this up as a possibility once again.

    Oh I love competition.
  • by ndykman ( 659315 ) on Tuesday April 19, 2005 @07:04PM (#12287154)
    Firstly, after reading the article, I was shocked to note that AMDs processors come out of one fab line, and the American fab line was flash only. If this is the case, well, wow. That seems a bit risky. If you get a tricky or persistent process issue (and it happens, no matter how cool you are), that seems like it could really impact AMDs output and yield a good deal.

    Of course, that's the main question here, and no way you are going to find out that answer. Yield. How many chips are good in a wafer?

    You can guess, but the answer may speak alot about AMD and Intel. It could very well be (here comes the flames) that Intel has an advantage in being consistent in volume and yield that allows them to keep large-scale contracts.

    It is a big question in my mind if AMD can currently provide the large-scale on demand volume that the big companies require in some product lines. Could an HP, a Gateway rely exclusively on AMD for chips? (I don't know)

    Certainly, it seems that have one fab plant only could be a big bottleneck or issue to make major vendors concerned and place a cloud on that question.

    Toss in this which the fact that you can get chipsets (heck, network chips if you'd like) from Intel as well, and you have a real competitive advantage that is tough to beat. All your motherboard bits, one vendor.

    And, sure, Intel chips have disadvantages, but in real-world experiences, the performance of similarly priced AMD and Intel desktop solutions aren't so obviously different that most people will notice enough to overcome those other issues at play.

    Just a thought.
  • by DrLex ( 811382 ) on Tuesday April 19, 2005 @07:16PM (#12287235) Homepage
    I visited an experimental fab here (at IMEC [www.imec.be]) a while ago and there also was a great deal of yellow light in some places. As far as I can remember, it has something to do with the processing. Some methods involve 'developing' photoresist layers on wafers, like developing a photographic film, and this process is insensitive to yellow light -- just like good ol' black & white photographs were insensitive to the typical red light in dark rooms.
    However, I recently visited a new cleanroom in the same fab, made for experimenting with 30mm wafers. Next to the fact that all the tiny rooms from the older fab were replaced by one large 'ballroom', all the light was simply white. I guess the new process is insensitive to visible wavelengths.
    So maybe they just colored the photograph to evoke the typical cleanroom-look of a decade ago. Or the photograph is simply from an old fab.
  • by edwdig ( 47888 ) on Tuesday April 19, 2005 @07:45PM (#12287492)
    The dual core chips with one defective core will simply be sold as single core chips. The single and dual core chips both use the same motherboards, so really the only way someone could tell them apart would be by the writing on it anyway.

    As to overclocking, there's only one bus and one clock driving both chips, so you can't clock them differently.
  • Quite right. The Austin plant hasn't done CPUs since the aluminium Athlon days (think Thunderbird core).

    But we are talking about the Dresden Fab 30, which was for a long time considered the most advanced fab in the world.

    "In May 2001, Fab 30 was awarded the coveted "Fab of the Year" title by Semiconductor International. The magazine recognized Fab 30 as the first facility in the world specifically designed to produce microprocessors with copper interconnects." http://www.amdboard.com/amdfab30.html [amdboard.com]

    With over 150,000 square feet of clean-room, it could, and does, handle the load.

    As a side note, here's AnandTech's tour of Fab 30: http://www.anandtech.com/cpuchipsets/showdoc.html? i=1773 [anandtech.com]

  • by kesuki ( 321456 ) on Tuesday April 19, 2005 @11:26PM (#12289226) Journal
    Doh you opened the door, now this batch is all shot!

    As many have stated here, if the window is tinted yellow, the room inside infact has all yellow (amber) lighting. This is because much as a photo negative will expose under more than the slightest infrared lighting, CPUs will not be etched correctly if exposed to UV rays in the wrong areas.

    The entire building is not yellow, as only certain processes are UV sensative, and once the part has been given the needed chemical baths they are no longer light sensitive.

    White light would burn out the chips about to be etched as surely as opening the door to a dark room before the film/photo paper can be given it's chemical bath to 'crystalize' the paper/films light sensitivity.
  • Re:Motherboards (Score:3, Interesting)

    by Herr_Nightingale ( 556106 ) on Tuesday April 19, 2005 @11:33PM (#12289273) Homepage
    I've found the nVidia nForce 3/4 to be better than Intel's latest; AMD's HTT makes a huge difference for memory-intensive computation :)
    With Intel having HT it's probably 6(1) or 1/2 dozen t'other, but I like my nForce.

"Experience has proved that some people indeed know everything." -- Russell Baker

Working...