{"history":[],"article":{"id":"90e187a8-2e3a-4708-9308-78a345060785","title":"Connection Machine","slug":"connection-machine","content":":::summary\n**Name:** Connection Machine \n**Manufacturer:** [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation) \n**Release Date:**

1986 (CM-1)
1987 (CM-2)
1991 (CM-5)

\n**Units Sold:** At least 70 \n**Dimensions:** ≈ 6 feet cubed (CM-1 and CM-2) \n**Weight:** 575.0 kg (CM-2) \n**Cpu:** Up to 65,536 1-bit processors (CM-1 and CM-2) \n**Memory:**

512 [Megabytes](https://bliptext.com/articles/megabytes) (CM-2)
10 [Gigabytes](https://bliptext.com/articles/gigabytes) ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5)

\n**Storage:**

Up to 80 [Gigabytes](https://bliptext.com/articles/gigabytes) with eight [DataVaults](https://bliptext.com/articles/datavault) (CM-1 and CM-2)
≈ 2 [Terabytes](https://bliptext.com/articles/terabytes) ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5)

\n**Flops:**

2.5 GigaFLOPS (CM-2)
65.5 GigaFLOPS ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5)

\n:::\n\n# \nThe **Connection Machine** (**CM**) is a member of a series of [massively parallel](https://bliptext.com/articles/massively-parallel) [supercomputers](https://bliptext.com/articles/supercomputer) sold by [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation).\nThe idea for the Connection Machine grew out of doctoral research on alternatives to the traditional [von Neumann architecture](https://bliptext.com/articles/von-neumann-architecture) of computers by [Danny Hillis](https://bliptext.com/articles/danny-hillis) at [Massachusetts Institute of Technology](https://bliptext.com/articles/massachusetts-institute-of-technology) (MIT) in the early 1980s.\nStarting with CM-1, the machines were intended originally for applications in [artificial intelligence](https://bliptext.com/articles/artificial-intelligence) (AI) and symbolic processing, but later versions found greater success in the field of [computational science](https://bliptext.com/articles/computational-science).\n\n# Origin of idea\n[Danny Hillis](https://bliptext.com/articles/danny-hillis) and [Sheryl Handler](https://bliptext.com/articles/sheryl-handler) founded [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation) (TMC) in [Waltham, Massachusetts](https://bliptext.com/articles/waltham-massachusetts), in 1983, moving in 1984 to Cambridge, MA.\nAt TMC, Hillis assembled a team to develop what would become the CM-1 Connection Machine, a design for a massively parallel [hypercube](https://bliptext.com/articles/hypercube-internetwork-topology)-based arrangement of thousands of [microprocessors](https://bliptext.com/articles/microprocessor), springing from his PhD thesis work at MIT in Electrical Engineering and Computer Science (1985).\nThe dissertation won the ACM Distinguished Dissertation prize in 1985, and was presented as a monograph that overviewed the philosophy, architecture, and software for the first Connection Machine, including information on its data routing between [central processing unit](https://bliptext.com/articles/central-processing-unit) (CPU) nodes, its memory handling, and the programming language [Lisp](https://bliptext.com/articles/lisp-programming-language) applied in the parallel machine.\nVery early concepts contemplated just over a million processors, each connected in a 20-dimensional hypercube, which was later scaled down.\n\n# Designs\nEach CM-1 microprocessor has its own 4 [kilobits](https://bliptext.com/articles/kilobit) of [random-access memory](https://bliptext.com/articles/random-access-memory) (RAM), and the [hypercube](https://bliptext.com/articles/hypercube-internetwork-topology)-based array of them was designed to perform the same operation on multiple data points simultaneously, i.e., to execute tasks in single instruction, multiple data ([SIMD](https://bliptext.com/articles/single-instruction-multiple-data)) fashion.\nThe CM-1, depending on the configuration, has as many as 65,536 individual processors, each extremely simple, processing [one bit](https://bliptext.com/articles/1-bit-architecture) at a time.\nCM-1 and its successor CM-2 take the form of a [cube](https://bliptext.com/articles/cube) 1.5 meters on a side, divided equally into eight smaller [cube](https://bliptext.com/articles/cube)s.\nEach subcube contains 16 [printed circuit boards](https://bliptext.com/articles/printed-circuit-board) and a main processor called a sequencer.\nEach circuit board contains 32 chips.\nEach chip contains a [router](https://bliptext.com/articles/router-computing), 16 processors, and 16 RAMs.\nThe CM-1 as a whole has a 12-dimensional [hypercube](https://bliptext.com/articles/hypercube)-based [routing](https://bliptext.com/articles/routing) network (connecting the 212 chips), a main RAM, and an [input-output processor (a channel controller)](https://bliptext.com/articles/channel-i-o).\nEach router contains five buffers to store the data being transmitted when a clear channel is not available.\nThe engineers had originally calculated that seven buffers per chip would be needed, but this made the chip slightly too large to build.\n[Nobel Prize](https://bliptext.com/articles/nobel-prize)-winning physicist [Richard Feynman](https://bliptext.com/articles/richard-feynman) had previously calculated that five buffers would be enough, using a differential equation involving the average number of 1 bits in an address.\nThey resubmitted the design of the chip with only five buffers, and when they put the machine together, it worked fine.\nEach chip is connected to a switching device called a nexus.\nThe CM-1 uses [Feynman's algorithm](https://bliptext.com/articles/logarithm) for computing logarithms that he had developed at [Los Alamos National Laboratory](https://bliptext.com/articles/los-alamos-national-laboratory) for the [Manhattan Project](https://bliptext.com/articles/manhattan-project).\nIt is well suited to the CM-1, using as it did, only shifting and adding, with a small table shared by all the processors.\nFeynman also discovered that the CM-1 would compute the Feynman diagrams for [quantum chromodynamics](https://bliptext.com/articles/quantum-chromodynamics) (QCD) calculations faster than an expensive special-purpose machine developed at Caltech.\nTo improve its commercial viability, TMC launched the CM-2 in 1987, adding [Weitek](https://bliptext.com/articles/weitek) 3132 [floating-point](https://bliptext.com/articles/floating-point) numeric [coprocessors](https://bliptext.com/articles/coprocessor) and more RAM to the system.\nThirty-two of the original one-bit processors shared each numeric processor.\nThe CM-2 can be configured with up to 512 MB of RAM, and a redundant array of independent disks ([RAID](https://bliptext.com/articles/raid)) [hard disk](https://bliptext.com/articles/hard-disk) system, called a [DataVault](https://bliptext.com/articles/datavault), of up to 25 GB. Two later variants of the CM-2 were also produced, the smaller CM-2a with either 4096 or 8192 single-bit processors, and the faster CM-200.\nDue to its origins in AI research, the software for the CM-1/2/200 single-bit processor was influenced by the [Lisp](https://bliptext.com/articles/lisp-programming-language) programming language and a version of Common [Lisp](https://bliptext.com/articles/lisp-programming-language), *[Lisp](https://bliptext.com/articles/lisp-programming-language) (spoken: Star-[Lisp](https://bliptext.com/articles/lisp-programming-language)), was implemented on the CM-1.\nOther early languages included [Karl Sims](https://bliptext.com/articles/karl-sims)' IK and Cliff Lasser's URDU.\nMuch system utility software for the CM-1/2 was written in *Lisp.\nMany applications for the CM-2, however, were written in [C*](https://bliptext.com/articles/c), a data-parallel superset of [ANSI C](https://bliptext.com/articles/ansi-c).\nWith the CM-5, announced in 1991, TMC switched from the CM-2's hypercubic architecture of simple processors to a new and different multiple instruction, multiple data ([MIMD](https://bliptext.com/articles/multiple-instruction-multiple-data)) architecture based on a [fat tree](https://bliptext.com/articles/fat-tree) network of [reduced instruction set computing](https://bliptext.com/articles/reduced-instruction-set-computing) (RISC) [SPARC](https://bliptext.com/articles/sparc) processors.\nTo make programming easier, it was made to simulate a [SIMD](https://bliptext.com/articles/single-instruction-multiple-data) design.\nThe later CM-5E replaces the SPARC processors with faster SuperSPARCs.\nA CM-5 was the fastest computer in the world in 1993 according to the [TOP500](https://bliptext.com/articles/top500) list, running 1024 cores with Rpeak of 131.0 G[FLOPS](https://bliptext.com/articles/flops), and for several years many of the top 10 fastest computers were CM-5s.\n# [Thinking Machines CM-2 at the Computer History Museum in Mountain View, California. One of the face plates has been partly removed to show the circuit boards inside. | upload.wikimedia.org/wikipedia/commons/6/66/Thinking///machines///cm2.jpg]\n# [The light panels of FROSTBURG, a CM-5, on display at the National Cryptologic Museum. The panels were used to check the usage of the processing nodes, and to run diagnostics. | upload.wikimedia.org/wikipedia/commons/0/01/Frostburg///%28CM-5%29///-///National///Cryptologic///Museum///-///DSC07914.JPG]\n\n# Visual design\nConnection Machines were noted for their striking visual design.\nThe CM-1 and CM-2 design teams were led by [Tamiko Thiel](https://bliptext.com/articles/tamiko-thiel).\nThe physical form of the CM-1, CM-2, and CM-200 chassis was a cube-of-cubes, referencing the machine's internal 12-dimensional [hypercube](https://bliptext.com/articles/hypercube) network, with the red [light-emitting diodes](https://bliptext.com/articles/light-emitting-diode) (LEDs), by default indicating the processor status, visible through the doors of each cube.\nBy default, when a processor is executing an instruction, its LED is on.\nIn a SIMD program, the goal is to have as many processors as possible working the program at the same time – indicated by having all LEDs being steady on.\nThose unfamiliar with the use of the LEDs wanted to see the LEDs blink – or even spell out messages to visitors.\nThe result is that finished programs often have superfluous operations to blink the LEDs.\nThe CM-5, in plan view, had a staircase-like shape, and also had large panels of red blinking LEDs.\nProminent sculptor-architect [Maya Lin](https://bliptext.com/articles/maya-lin) contributed to the CM-5 design.\n\n# Surviving examples\n\n## Permanent exhibits\n\n## Past exhibits, Museum collections\n\n## Private collections\n\n# References in popular culture\nA CM-5 was featured in the film [Jurassic Park](https://bliptext.com/articles/jurassic-park-film) in the [control room](https://bliptext.com/articles/control-room) for the [island](https://bliptext.com/articles/island) (instead of a [Cray X-MP](https://bliptext.com/articles/cray-x-mp) [supercomputer](https://bliptext.com/articles/supercomputer) as in the novel).\nTwo banks, one bank of 4 Units and a single off to the right of the set could be seen in the control room.\nThe computer mainframes in [Fallout 3](https://bliptext.com/articles/fallout-3) were inspired heavily by the CM-5.\n[Cyberpunk 2077](https://bliptext.com/articles/cyberpunk-2077) features numerous CM-1/CM-2 style units in various portions of the game.\nThe b-side to [Clock DVA](https://bliptext.com/articles/clock-dva)'s 1989 single \"The Hacker\" is titled \"The Connection Machine\" in reference to the CM-1.","current_revision":null,"created_by":"kACnuQmH5k9Nr5VNVEpVeg7XpvUmMctD","created_at":"2025-01-25T17:44:43.998Z","updated_at":"2025-01-25T17:44:43.998Z","search_vector":"'connect':1 'machin':2","search_content":" \n**Name:** Connection Machine \n**Manufacturer:** [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation) \n**Release Date:** 1986 (CM-1) 1987 (CM-2) 1991 (CM-5) \n**Units Sold:** At least 70 \n**Dimensions:** ≈ 6 feet cubed (CM-1 and CM-2) \n**Weight:** 575.0 kg (CM-2) \n**Cpu:** Up to 65,536 1-bit processors (CM-1 and CM-2) \n**Memory:** 512 [Megabytes](https://bliptext.com/articles/megabytes) (CM-2) 10 [Gigabytes](https://bliptext.com/articles/gigabytes) ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5) \n**Storage:** Up to 80 [Gigabytes](https://bliptext.com/articles/gigabytes) with eight [DataVaults](https://bliptext.com/articles/datavault) (CM-1 and CM-2) ≈ 2 [Terabytes](https://bliptext.com/articles/terabytes) ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5) \n**Flops:** 2.5 GigaFLOPS (CM-2) 65.5 GigaFLOPS ([FROSTBURG](https://bliptext.com/articles/frostburg) CM-5) \n \n\n# \nThe **Connection Machine** (**CM**) is a member of a series of [massively parallel](https://bliptext.com/articles/massively-parallel) [supercomputers](https://bliptext.com/articles/supercomputer) sold by [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation).\nThe idea for the Connection Machine grew out of doctoral research on alternatives to the traditional [von Neumann architecture](https://bliptext.com/articles/von-neumann-architecture) of computers by [Danny Hillis](https://bliptext.com/articles/danny-hillis) at [Massachusetts Institute of Technology](https://bliptext.com/articles/massachusetts-institute-of-technology) (MIT) in the early 1980s.\nStarting with CM-1, the machines were intended originally for applications in [artificial intelligence](https://bliptext.com/articles/artificial-intelligence) (AI) and symbolic processing, but later versions found greater success in the field of [computational science](https://bliptext.com/articles/computational-science).\n\n# Origin of idea\n[Danny Hillis](https://bliptext.com/articles/danny-hillis) and [Sheryl Handler](https://bliptext.com/articles/sheryl-handler) founded [Thinking Machines Corporation](https://bliptext.com/articles/thinking-machines-corporation) (TMC) in [Waltham, Massachusetts](https://bliptext.com/articles/waltham-massachusetts), in 1983, moving in 1984 to Cambridge, MA.\nAt TMC, Hillis assembled a team to develop what would become the CM-1 Connection Machine, a design for a massively parallel [hypercube](https://bliptext.com/articles/hypercube-internetwork-topology)-based arrangement of thousands of [microprocessors](https://bliptext.com/articles/microprocessor), springing from his PhD thesis work at MIT in Electrical Engineering and Computer Science (1985).\nThe dissertation won the ACM Distinguished Dissertation prize in 1985, and was presented as a monograph that overviewed the philosophy, architecture, and software for the first Connection Machine, including information on its data routing between [central processing unit](https://bliptext.com/articles/central-processing-unit) (CPU) nodes, its memory handling, and the programming language [Lisp](https://bliptext.com/articles/lisp-programming-language) applied in the parallel machine.\nVery early concepts contemplated just over a million processors, each connected in a 20-dimensional hypercube, which was later scaled down.\n\n# Designs\nEach CM-1 microprocessor has its own 4 [kilobits](https://bliptext.com/articles/kilobit) of [random-access memory](https://bliptext.com/articles/random-access-memory) (RAM), and the [hypercube](https://bliptext.com/articles/hypercube-internetwork-topology)-based array of them was designed to perform the same operation on multiple data points simultaneously, i.e., to execute tasks in single instruction, multiple data ([SIMD](https://bliptext.com/articles/single-instruction-multiple-data)) fashion.\nThe CM-1, depending on the configuration, has as many as 65,536 individual processors, each extremely simple, processing [one bit](https://bliptext.com/articles/1-bit-architecture) at a time.\nCM-1 and its successor CM-2 take the form of a [cube](https://bliptext.com/articles/cube) 1.5 meters on a side, divided equally into eight smaller [cube](https://bliptext.com/articles/cube)s.\nEach subcube contains 16 [printed circuit boards](https://bliptext.com/articles/printed-circuit-board) and a main processor called a sequencer.\nEach circuit board contains 32 chips.\nEach chip contains a [router](https://bliptext.com/articles/router-computing), 16 processors, and 16 RAMs.\nThe CM-1 as a whole has a 12-dimensional [hypercube](https://bliptext.com/articles/hypercube)-based [routing](https://bliptext.com/articles/routing) network (connecting the 212 chips), a main RAM, and an [input-output processor (a channel controller)](https://bliptext.com/articles/channel-i-o).\nEach router contains five buffers to store the data being transmitted when a clear channel is not available.\nThe engineers had originally calculated that seven buffers per chip would be needed, but this made the chip slightly too large to build.\n[Nobel Prize](https://bliptext.com/articles/nobel-prize)-winning physicist [Richard Feynman](https://bliptext.com/articles/richard-feynman) had previously calculated that five buffers would be enough, using a differential equation involving the average number of 1 bits in an address.\nThey resubmitted the design of the chip with only five buffers, and when they put the machine together, it worked fine.\nEach chip is connected to a switching device called a nexus.\nThe CM-1 uses [Feynman's algorithm](https://bliptext.com/articles/logarithm) for computing logarithms that he had developed at [Los Alamos National Laboratory](https://bliptext.com/articles/los-alamos-national-laboratory) for the [Manhattan Project](https://bliptext.com/articles/manhattan-project).\nIt is well suited to the CM-1, using as it did, only shifting and adding, with a small table shared by all the processors.\nFeynman also discovered that the CM-1 would compute the Feynman diagrams for [quantum chromodynamics](https://bliptext.com/articles/quantum-chromodynamics) (QCD) calculations faster than an expensive special-purpose machine developed at Caltech.\nTo improve its commercial viability, TMC launched the CM-2 in 1987, adding [Weitek](https://bliptext.com/articles/weitek) 3132 [floating-point](https://bliptext.com/articles/floating-point) numeric [coprocessors](https://bliptext.com/articles/coprocessor) and more RAM to the system.\nThirty-two of the original one-bit processors shared each numeric processor.\nThe CM-2 can be configured with up to 512 MB of RAM, and a redundant array of independent disks ([RAID](https://bliptext.com/articles/raid)) [hard disk](https://bliptext.com/articles/hard-disk) system, called a [DataVault](https://bliptext.com/articles/datavault), of up to 25 GB. Two later variants of the CM-2 were also produced, the smaller CM-2a with either 4096 or 8192 single-bit processors, and the faster CM-200.\nDue to its origins in AI research, the software for the CM-1/2/200 single-bit processor was influenced by the [Lisp](https://bliptext.com/articles/lisp-programming-language) programming language and a version of Common [Lisp](https://bliptext.com/articles/lisp-programming-language), *[Lisp](https://bliptext.com/articles/lisp-programming-language) (spoken: Star-[Lisp](https://bliptext.com/articles/lisp-programming-language)), was implemented on the CM-1.\nOther early languages included [Karl Sims](https://bliptext.com/articles/karl-sims)' IK and Cliff Lasser's URDU.\nMuch system utility software for the CM-1/2 was written in *Lisp.\nMany applications for the CM-2, however, were written in [C*](https://bliptext.com/articles/c), a data-parallel superset of [ANSI C](https://bliptext.com/articles/ansi-c).\nWith the CM-5, announced in 1991, TMC switched from the CM-2's hypercubic architecture of simple processors to a new and different multiple instruction, multiple data ([MIMD](https://bliptext.com/articles/multiple-instruction-multiple-data)) architecture based on a [fat tree](https://bliptext.com/articles/fat-tree) network of [reduced instruction set computing](https://bliptext.com/articles/reduced-instruction-set-computing) (RISC) [SPARC](https://bliptext.com/articles/sparc) processors.\nTo make programming easier, it was made to simulate a [SIMD](https://bliptext.com/articles/single-instruction-multiple-data) design.\nThe later CM-5E replaces the SPARC processors with faster SuperSPARCs.\nA CM-5 was the fastest computer in the world in 1993 according to the [TOP500](https://bliptext.com/articles/top500) list, running 1024 cores with Rpeak of 131.0 G[FLOPS](https://bliptext.com/articles/flops), and for several years many of the top 10 fastest computers were CM-5s.\n# [Thinking Machines CM-2 at the Computer History Museum in Mountain View, California. One of the face plates has been partly removed to show the circuit boards inside. | upload.wikimedia.org/wikipedia/commons/6/66/Thinking///machines///cm2.jpg]\n# [The light panels of FROSTBURG, a CM-5, on display at the National Cryptologic Museum. The panels were used to check the usage of the processing nodes, and to run diagnostics. | upload.wikimedia.org/wikipedia/commons/0/01/Frostburg///%28CM-5%29///-///National///Cryptologic///Museum///-///DSC07914.JPG]\n\n# Visual design\nConnection Machines were noted for their striking visual design.\nThe CM-1 and CM-2 design teams were led by [Tamiko Thiel](https://bliptext.com/articles/tamiko-thiel).\nThe physical form of the CM-1, CM-2, and CM-200 chassis was a cube-of-cubes, referencing the machine's internal 12-dimensional [hypercube](https://bliptext.com/articles/hypercube) network, with the red [light-emitting diodes](https://bliptext.com/articles/light-emitting-diode) (LEDs), by default indicating the processor status, visible through the doors of each cube.\nBy default, when a processor is executing an instruction, its LED is on.\nIn a SIMD program, the goal is to have as many processors as possible working the program at the same time – indicated by having all LEDs being steady on.\nThose unfamiliar with the use of the LEDs wanted to see the LEDs blink – or even spell out messages to visitors.\nThe result is that finished programs often have superfluous operations to blink the LEDs.\nThe CM-5, in plan view, had a staircase-like shape, and also had large panels of red blinking LEDs.\nProminent sculptor-architect [Maya Lin](https://bliptext.com/articles/maya-lin) contributed to the CM-5 design.\n\n# Surviving examples\n\n## Permanent exhibits\n\n## Past exhibits, Museum collections\n\n## Private collections\n\n# References in popular culture\nA CM-5 was featured in the film [Jurassic Park](https://bliptext.com/articles/jurassic-park-film) in the [control room](https://bliptext.com/articles/control-room) for the [island](https://bliptext.com/articles/island) (instead of a [Cray X-MP](https://bliptext.com/articles/cray-x-mp) [supercomputer](https://bliptext.com/articles/supercomputer) as in the novel).\nTwo banks, one bank of 4 Units and a single off to the right of the set could be seen in the control room.\nThe computer mainframes in [Fallout 3](https://bliptext.com/articles/fallout-3) were inspired heavily by the CM-5.\n[Cyberpunk 2077](https://bliptext.com/articles/cyberpunk-2077) features numerous CM-1/CM-2 style units in various portions of the game.\nThe b-side to [Clock DVA](https://bliptext.com/articles/clock-dva)'s 1989 single \"The Hacker\" is titled \"The Connection Machine\" in reference to the CM-1.","revisionCount":0},"hasMore":false,"nextCursor":null}