IT Infrastructure - eWeek

IT Infrastructure: Inside Intel's Chip Creation and Innovation


Intel is the dominant microprocessor vendor in the world, holding an almost 80 percent share of the world's chip market, according to research firm iSuppli. The company's chips run in most PCs and servers today, and in recent years—thanks in large part to Intel's intense rivalry with Advanced Micro Devices—processors have become smaller, more powerful and more feature-rich. Gone are the days when upgrading chips meant simply cranking up the clock speed. For example, now Intel, AMD and other chips makers—including Sun Microsystems and IBM—are rapidly growing the number of processing cores on each chip. Intel and AMD each have chips with six cores, and both are planning to aggressively grow those numbers in the near future. The processors also now offer such features as power management, on-board virtualization capabilities and increased memory bandwidth, all of which were highlighted in Intel's launch of its "Nehalem" architecture and AMD's rollout of its "Istanbul" Opteron chip. On its Website, Intel recently offered a high-level glimpse of what goes into the making of a CPU. Here are some of the slides from the graphics presentation, which show only the most important steps among the hundreds that are needed to build a processor. Photo credits: Intel
 
  • Inside Intel's Chip Creation and Innovation
    by Jeffrey Burt
  • Starting with Sand

    Sand is the base ingredient for semiconductor manufacturing, given its high percentages of silicon. The silicon is purified through multiple steps to reach a semiconductor manufacturing quality called Electronic Grade Silicon, which means there can only be one alien atom every 1 billion silicon atoms. In the second image, one big crystal is grown from the purified silicon melt. In the last image is an ingot that has been produced from Electronic Grade Silicon. One ingot weighs about 220 pounds.

  • Slicing the Ingot

    The ingot is cut into wafers, which are single silicon disks. As seen in the second image, the wafers are then polished until they have flawless and mirror-smooth surfaces. Intel buys the manufacturing-ready wafers from third-party suppliers. In its 45-nanometer manufacturing process, Intel uses wafers with diameters of 300 millimeters, or about 12 inches.

  • Getting to the Transistor

    The blue liquid in the first image is a photo-resist finish that is poured onto the wafer while it spins. In the second image, the the photo-resist finish--which is similar to the finish used in film photography--is exposed to ultraviolet light, which makes the finish soluble. Then masks--which act like stencils--are used with the UV light to create circuit patterns on each layer of the microprocessor. A lens, seen in the middle of the UV light, reduces the mask's image to the point that what gets printed on the wafter is about four times smaller than the pattern on the mask. The third image shows a transistor, which controls the flow of electrical current in the chip. Intel has developed transistors so small that about 30 million of them can fit on the head of a pin.

  • Removing the Photo Resist

    In the first image, a solvent is used to dissolve the gooey photo resist, revealing the pattern of photo resist made by the mask. Chemicals are then used to etch away material revealed by the dissolved photo resist, and, finally, the photo resist is removed and the desired shape of the transistor becomes visible, as seen in the third image.

  • Implanting the Ions

    Photo resist, in blue, is applied. The photo resist protects material that should not get ions implanted on it. In the second image, through a process called ion implantation--one form of a process called "doping"--the exposed areas of the wafer are bombarded with ions, which are various chemical impurities that alter the way silicon in these areas conduct electricity. The ions are shot into the wafer's surface at speeds of about 185,000 mph. After the doping process, the photo resist is removed.

  • Putting on the Copper

    The transistor in the first image is close to being finished, and three holes have been etched into the insulation layer--in magenta--above it. The holes will be filled with copper for connections to other transistors. In the second image, transistors--which at this point are still on the wafers--are put into a copper sulphate solution, where copper ions are put onto the transistor through the electroplating process. After the electroplating, the copper ions settle on the wafer surface as a thin layer of copper.

  • Applying the Metal

    The excess material is polished off, as seen in the first image. Then metal layers are created for interconnects between the various transistors. The chip's architecture and design teams will determine how the connections are made. Though they appear flat, chips can have more than 20 layers in order to create their circuitry.

  • Finding the 'Right Answer'

    Wafers now undergo their first functionality tests, where test patterns are fed into every chip. The response from the chip is then monitored and compared to what engineers say is the "right answer." As seen in the second image, the wafer is then cut into pieces, called dies. Then the dies with the right answer to the test pattern are sent forward to the next step, which is packaging. Faulty dies--those that did not respond with the right answer--are discarded.

  • Completing the Process

    The first image shows an individual die--in this case, a die of an Intel Core i7 chip--cut from the wafter. Then the substrate, the die and the heatspreader are put together to create a completed processor. The substrate, in green in the second image, creates the electrical and mechanical interface through which the chip interacts with the rest of the PC system, and the silver heatspreader will be covered with a cooling solution to keep the processor cool while it's operating. The end result is the completed processor, shown in the third image.

  • Preparing for Retail

    The chip goes through its final tests for such key characteristics as power dissipation and maximum frequency. As shown in the second image, processors with the same capabilities--as determined through class testing--are put into the same transporting trays. Finally, the manufactured and tested chips are sent to system manufacturers in trays or to retail stores in a box, as shown in the third image.

xڽZ[s۸~~d_nw7ŗXbwIHB ,JVeSjy|8oi&4oAuB>K9lHbc.!KvfzC7o``ƛVYQ|v$UK*7-;RLDL1OwҨxLO^>.B࿎F:|=F#mX021q #yF.#An{R7Yh[((6j{A،LQLDruzH&$fvȿ+|>o~QLg\M<4#9 "L7'f*Ή{'_tt6g4Wln(3сױ1,?ۈnS3aܰ)M)vtZ4ܪwP臑WhNdm\)I;-<><:1)+Vʢ M&x9tKS:f:bLÃnF1>PC ';!{0#p\V5Ra5G\攪1O|؈{G{(a(;}eq+)P0nC-U̥zU\Ɋn'p` Ձ ^2R`-*T հq셜Þݙo&)?>d R_E!*aR= ׼>Iqv* .[1aCuTD,O4͌3Ky W&?׀c:sBrZ)ĺQĴdv-[9Ϛ 7s.9n ٝօrI/ ah~I%_zlgŎvjf]ffSO0[h>\ !I$STؖN ik'?aWyQq& XR?Ó,)uN~t@|@>c/1)rd|>s\MbSH;{$+}7icqD8 +9 (< ^0ZV {7VeLt"dY> "6#h9p\+ɻtqO@"f.[㑙CB0e_.Ն^ovﶍ#+𫄐Х$^Xr:= vA͘wsN%!JfI+knsqeQapS#Z7>g67c 3O#$vjڧDw% vӂֽWˆלWestOuO~)\9c.4ga-q a cp+)=jѤ:>*7p}SUA^8/VK@V ncFјfTT>g_Y#xO3+(XP.u1sgֽ]0? V|țjUxמo4g{l qEig`oiXI%k#׉x`+,B!S]OLmSgNVӪ2讞]BU:m-.U3g`GG*ar;LmTkqء9j[Xo2cQ JvV_ܔ&˼C]`I`z#p7PD'RMApuPy#y&յy8$Pc.kEb oDz j*cnW~#Z3H䈘 M؞(.D-bt7")th8ɗn &5 ;όF)xp98R.'2]([yw~ޑZuU )RX^c݆̪f;OY*@Λ5`(3<>яv&NXyB)@ SPpxdMADfTaz8ak`9 &*nZS4,PB qٓu"fˆI^KerDdTǪl or AmkB Cs.)(iͩu ?YdbPds[6d$ͺ4X9X؅$ wwv3]»C.0x` acmA4sǑg,Rm.!?^>qL玩&G \01cX{}=+rx>qekZFEbp\|rvP~|x{/: L"s}{zap'xtOO{Wp QXa,m7:~y3Ne`)7}x/\t| Zh3Ow?7mXΊF+c+Ѩ~=ŴYI2UeTF,7vZFSƼ ܟ.^: w%16:tq@N㭲P ƫJڄ>,