zeke@shamash.cdc.com (Robert Scott) (09/14/90)
In article <25833@shamash.cdc.com>, rrr@u02.svl.cdc.com (Rich Ragan) writes: > > Stuff deleted... > > I heard a 6603 story (possibly apocryphal) that at some site > the heads crashed and the motors kept on turning. Supposedly > the magnesium platters soon ignited and the equivalent of > "meltdown" occurred in the computer room. > > -- > Richard R. Ragan (408) 496-4340 > Control Data Corp., Silicon Valley Operations I have another meltdown story to relate regarding CDC equipment. A friend of mine was doing performance testing on our then in-house Cyber 205, serial #501 (the first production machine). At this time, the machine had been installed in the Arden Hills computer room for several years, and was frequently used as a testbed for new hardware updates, such as new memory configurations. It seems that the in-house engineers had developed a non-standard method of removing memory boards in a new type of memory unit, and over time the connectors became slightly loose. My friend began to run a very memory intensive test that read/wrote patterns through various banks of memory as fast as possible. After this program had been running for some time, someone in the computer room smelled smoke, and discovered that there was smoke coming forth from the vents at the top of the 205 memory bay. An emergency powerdown was accomplished and everything "put out". It was later surmised that the increased current flow to/from the memories during the performance test had heated the chips, then the boards, and then the connectors to the point where some of the loose connectors began to melt. Now that's a hot computer! Zeke (Apologies to Bernie if I got the details a little wrong, but I stand by the scenario) -- ~~~~~~~~~~~ From the Shrine of the "Last Gasp of ETA Systems" ~~~~~~~~~~~~~ Extra zesty disclaimer: MINE! MINE! ALL MINE! <chortle snort froth drool> Robert K. "Zeke" Scott internet: zeke@eta.cdc.com Control Data Corp, Supercomputer Support Group
bksmith@aix.aix.kingston.ibm.com (Bernie King-Smith) (09/18/90)
In article <25835@shamash.cdc.com> zeke@shamash.cdc.com (Robert Scott) writes: > >I have another meltdown story to relate regarding CDC equipment. A friend >of mine was doing performance testing on our then in-house Cyber 205, >serial #501 (the first production machine). At this time, the machine >had been installed in the Arden Hills computer room for several years, and >was frequently used as a testbed for new hardware updates, such as new memory >configurations. > >It seems that the in-house engineers had developed a non-standard method of >removing memory boards in a new type of memory unit, and over time the >connectors became slightly loose. My friend began to run a very memory >intensive test that read/wrote patterns through various banks of memory as >fast as possible. After this program had been running for some time, someone >in the computer room smelled smoke, and discovered that there was smoke >coming forth from the vents at the top of the 205 memory bay. An emergency >powerdown was accomplished and everything "put out". It was later surmised >that the increased current flow to/from the memories during the performance >test had heated the chips, then the boards, and then the connectors to the >point where some of the loose connectors began to melt. > >Now that's a hot computer! > > >Zeke > Yes, these details are pretty much correct. The program that was executing heated up the memory boards the farthest from the heat sensors on the machine. The rest of the memory frame apparently cooled the total frame so that the heat sensor never went off. When the board finally overheated, the resin in the board set the whole memory frame on fire. In addition, when the machine caught on fire, the smoke poured from the machine and went right past the smoke detectors. THEY NEVER WENT OFF !!!! However, two years later a fellow countryman of mine from down under continued the tradition on a air cooled ETA10. On this machine, also a prototype, the CPU board (There is only one board per CPU in a ETA10) was wirewrapped. There were several routing corrections made on the surface of the 44 layer board. One night both of us were each running on one of the two CPU's in the machine when he burned a hole the size of an American quarter through the CPU board. To my knowledge these were two of the three fires caused in the CYBER205/ETA10 machines. The other one was a production CYBER 250 at ETA Systems, Inc. They were really "HOT" machines from a performance and experience point of view. Bernie King-Smith (The WOMBAT) Former System Performance Analyst ETA Systems, Inc. Disclaimer: This posting is only from my experience and does not reflect on the opinions of my current employer. P.S. Sorry Peter if the details are not completely correct !!!!