URL
https://opencores.org/ocsvn/phr/phr/trunk
Subversion Repositories phr
[/] [phr/] [trunk/] [doc/] [informe-tesis/] [phd-thesis-template-master/] [extra-ref/] [www/] [3D-MAPS Many-Core Processor.html] - Rev 132
Go to most recent revision | Compare with Previous | Blame | View Log
<!-- saved from url=(0054)http://arch.ece.gatech.edu/research/3dmaps/3dmaps.html --> <html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> <title>3D-MAPS Many-Core Processor</title> <meta name="keywords" content="3D IC, Georgia Tech, 3D MAPS, 3D architecture, 3-D IC, 3D-MAPS"> <meta description="description" content="3D IC architecture research using die stacking and TSV at Georgia Tech"> <meta name="author" content="Hsien-Hsin Lee"> <style type="text/css"> A:link{text-decoration:none} A:visited{text-decoration:none} TD{font-family: verdana; font-size: 10pt} </style></head> <body bgcolor="white" text="black" link="red" vlink="red" alink="red"> <font face="verdana" size="2"> <h3> <font color="blue"> The 3D-MAPS Many-Core Processor</font></h3> <h5> <font color="green"> Announcements </font> </h5> <ul> <li> 10/24/11: We taped out 3D-MAPS V2 to MOSIS/Tezzaron MPW run. </li><li> 02/21/12: 3D-MAPS V1 is presented at ISSCC 2012. Here is our <a href="http://arch.ece.gatech.edu/pub/isscc12.pdf" target="new"> paper</a> and <a href="http://arch.ece.gatech.edu/present/isscc12.pdf" target="new">presentation</a>. </li></ul> <h5> <font color="green"> Media Coverages </font> </h5> <ul> <li> <a href="http://www.i-micronews.com/news/3D-MAPS-multicore-processor-closer-look,8706.html" target="new">3D-MAPS multicore processor: A closer look</a> (I-Micronews) </li><li> <a href="http://eda360insider.wordpress.com/2012/03/01/3d-thursday-three-on-3d-papers-from-isscc/" target="new">3D Thursday: Threee on 3D --- papers from ISSCC</a> (EDA360 Insider) </li><li> <a href="http://www.electroiq.com/blogs/insights_from_leading_edge/2012/04/iftle-97-date-in-dresden-synopsys-3d-eda-solution.html" target="new">3D EDA Solution</a> (Solid State Technology) </li><li> <a href="http://www.electroiq.com/blogs/insights_from_leading_edge/2012/03/iftle-93-2-5-3d-at-the-2012-ieee-isscc.html" target="new">IFTLE 93 2.5 / 3D at the 2012 IEEE ISSCC</a> (Solid State Technology) </li><li> <a href="http://www.eetimes.com/electronics-news/4236570/ISSCC--Picture-from-a-silicon-exhibition?pageNumber=1" target="new">ISSCC: Pictures from a silicon exhibition</a> (EE Times) </li><li> <a href="http://www.theregister.co.uk/2011/12/22/isscc_2012_chip_preview/" target="new"> New chippery on parade at ISSCC: CPU and memory makers strut their stuff</a> (The Register) </li><li> <a href="http://www.theregister.co.uk/2012/02/24/3d_chips/page3.html" target="new"> Real apps, real benchmarks. Georgia Institute of Technology's "3D-MAPS: 3D massively parallel processor with stacked memory"</a> (The Register) </li></ul> <h5> <font color="green"> Overview </font> </h5> 3D-MAPS (3D <u>MA</u>ssively <u>P</u>arallel processor with <u>S</u>tacked memory) V1 is a logic+memory 2-tier 3D IC, where the logic die consists of 64 general purpose processor cores running at 277MHz, and the memory die contains 256KB SRAM. <font color="red"> This 3D IC is arguably the FIRST many-core general purpose 3D processor developed in academia.</font> This 3D processor achieves up to 64GB/s memory bandwidth while consuming 5W power. This project is led by <a href="http://users.ece.gatech.edu/limsk/">Prof. Sung Kyu Lim</a> (PI) and <a href="http://users.ece.gatech.edu/~leehs/">Prof. Hsien-Hsin Lee</a> (co-PI) from the Georgia Institute of Technology and Dr. Gabriel Loh (co-PI) from AMD with funding from the US Department of Defense. There have been 20+ students involved in this project working on architecture, programming, CAD tools, circuit and physical design, packaging, board design, and testing. Our collaborators include KAIST, Tezzaron, Amkor Inc, and Board Lab. <p><img src="./3D-MAPS Many-Core Processor_files/team.JPG"></p><p> </p><p>The fabrication of this chip is completed in July 2011 using the 130nm GlobalFoundies device technology and 1.2um TSV diameter Tezzaron technology. The packaging is completed in August 2011 by Amkor. 8 parallel applications are developed to demonstrate the bandwidth and power benefit of 3D MAPS processor. This processor contains 33M transistors, 50K TSVs, and 50K face-to-face connections in 5mm x 5mm footprint and 0.8mm thickness. </p><p>The core architecture is developed from scratch by our architecture team to benefit from single-cycle access to SRAM. One of the two instructions we issue in one cycle can be memory read/write, so it is possible to access memory at every clock cycle. Our RTL-to-GDSII tool chain is based on commercial tools from Synopsys, Cadence, and Mentor Graphics. Since these tools can only handle 2D ICs, we have developed plug-ins to handle TSVs and 3D stacking. </p><p> Here is our <a href="http://arch.ece.gatech.edu/pub/cicc10.pdf" target="new"> CICC 2010</a> and <a href="http://arch.ece.gatech.edu/pub/3dtest10.pdf" target="new"> 3D-TEST 2010</a> papers on 3D-MAPS V1. </p><p>We are currently working on 3D-MAPS V2 that features 128 cores and 2GB DRAM stacked in 5 dies. Here are the differences:</p><p> <table border="1"> <tbody><tr> <td></td> <td align="center"><font color="blue">3D-MAPS V1</font></td> <td align="center"><font color="blue">3D-MAPS V2</font></td> </tr> <tr> <td># of tiers</td> <td align="center">2, one logic and one SRAM</td> <td align="center">5, two logic and three DRAM</td> </tr> <tr> <td># of cores</td> <td align="center">64</td> <td align="center">128</td> </tr> <tr> <td>logic footprint</td> <td align="center">5mm x 5mm</td> <td align="center">10mm x 10mm</td> </tr> <tr> <td>DRAM footprint</td> <td align="center">-</td> <td align="center">20mm x 12mm</td> </tr> <tr> <td>device technology</td> <td align="center">130nm, Globalfoundries</td> <td align="center">130nm, Globalfoundries</td> </tr> <tr> <td>bonding style</td> <td align="center">face-to-face</td> <td align="center">face-to-face & face-to-back</td> </tr> <tr> <td>TSV technology</td> <td align="center">Tezzaron, 1.2um diam</td> <td align="center">Tezzaron, 1.2um diam</td> </tr> </tbody></table> </p><h5> <font color="green"> 3D-MAPS V1 Specifications </font> </h5> <p><img src="./3D-MAPS Many-Core Processor_files/spec.jpg"></p><p> </p><h5> <font color="green"> 3D-MAPS V1 Measurement Results </font> </h5> 3D-MAPS V1 supports 42 instructions, and we wrote 8 parallel applications and ran them on our chip. Here are the memory bandwidth and power measurement results.<p> <table border="1"> <tbody><tr> <td>application</td> <td align="center">memory BW (GB/s)</td> <td align="center">power consumption (W)</td> </tr> <tr> <td>AES encryption</td> <td align="center">49.5</td> <td align="center">4.032</td> </tr> <tr> <td>edge detection</td> <td align="center">15.6</td> <td align="center">3.768</td> </tr> <tr> <td>histogram</td> <td align="center">30.3</td> <td align="center">3.588</td> </tr> <tr> <td>k-means clustering</td> <td align="center">40.6</td> <td align="center">4.014</td> </tr> <tr> <td>matrix multiply</td> <td align="center">13.8</td> <td align="center">3.789</td> </tr> <tr> <td>median filter</td> <td align="center"><font color="red">63.8</font></td> <td align="center"><font color="red">4.007</font></td> </tr> <tr> <td>motion estimation</td> <td align="center">24.1</td> <td align="center">3.830</td> </tr> <tr> <td>string search</td> <td align="center">8.9</td> <td align="center">3.876</td> </tr> </tbody></table> </p><p>The theoretical maximum memory bandwidth 3D-MAPS V1 can achieve is 70.9GB/s, which is computed by 277MHz x 64 (cores) x 4 Bytes (1 word). One of our applications, median filter, got very close to this theoretical value at the lowest power consumption. As a comparison, here are the maximum achievable bandwidth values of the state-of-the-art processor and memory technology (as of Sep 2011): </p><ul> <li> Intel i7 Extreme Edition + Samsung DDR3 1600 MHz = 1600 MHz x 2 ch x 8 Bytes = 25.6 GB/s </li><li> Intel Xeon E7 + Samsung DDR3 1066 MHz = 1066 MHz x 4 ch x 8 Bytes = 34.1 GB/s </li></ul> 3D-MAPS V1 is fabricated in 130nm technology in 5mm x 5mm footprint. If 3D-MAPS V1 is fabricated in 45nm in 15mm x 15mm footprint (as in Intel i7), the maximum memory bandwidth skyrockets as follows: <ul> <li> 277 MHz X 5 (speedup from 45nm) x 64 ch x 9 (more area) x 4 (smaller cores) x 4 Bytes = <font color="red">12,764 GB/s</font> </li></ul> This truly demonstrates the enormous memory bandwidth benefit of core+memory 3D IC. <h5> <font color="green"> 3D-MAPS V1 Photos </font> </h5> <font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig8.jpg"><br>FIGURE 1: This is the stacking information of 3D-MAPS V1. We use bonding wires and TSVs for the package-to-chip signal and P/G delivery. Chip-to-chip communication is done using F2F pads. The bonding wires did not break the TSVs underneath during manufacturing. Each IO cell contains 204 redundant TSVs.<p> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig1.JPG"><br><font color="blue">FIGURE 2: The topside of 3D-MAPS V1 is actually the backside of the core die that is thinned down to 12um. With bare eyes, we can only see dummy TSVs and IO cells.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig3.JPG"><br>FIGURE 3: SEM image of Tezzaron TSVs and face-to-face bond pads.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig10.jpg"><br>FIGURE 4: More SEM images of TSVs and F2F pads.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig4.png"><br>FIGURE 5: The above image is obtained using an infrared microscope with 6um depth. Since the top surface of 3D-MAPS V1 is the thinned substrate of top die, we had to use an IR microscope to reveal the circuitry that is buried under this substrate. The white dots are dummy TSVs we had to add to satisfy the TSV density rule set by Tezzaron.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig7.jpg"><br>FIGURE 6: Some details of single core and single IO cell.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig2.JPG"><br>FIGURE 7: Bare die and its package side-by-side.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig5.JPG"><br>FIGURE 8: The open TSV above, fortunately, does not cause any problem because all of the TSVs shown are redundant. The top die (= core die) is thinned down to 12um, and the bottom die (= memory die) height is 765um, making the total thickness to be roughly 0.8mm.</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig6.JPG"><br>FIGURE 9: A dummy TSV</font></p><p><font color="blue"> <img src="./3D-MAPS Many-Core Processor_files/3D-MAPS-V1-fig9.jpg"><br>FIGURE 10: Layouts of full-die (core and memory) and single core/memory tile.</font></p><p><font color="blue"> </font> </p><p> <img src="./3D-MAPS Many-Core Processor_files/Count.cgi"> </p></font><p><font color="blue"> </font> </p></font></body></html>
Go to most recent revision | Compare with Previous | Blame | View Log