wave@media-lab.MEDIA.MIT.EDU (Michael B. Johnson) (03/25/91)
In article <18967@milton.u.washington.edu> you write: >>In article <18646@milton.u.washington.edu>, frerichs@ux1.cso.uiuc.edu >>(David J Frerichs) writes: >> >>> ... Talk about the Connection Machine ... >> >>> frerichs@ux1.cso.uiuc.edu _/_\__/_\__/_\_ >>> frerichs@well.sf.ca.us \_/ \_/ \_/ ] >> >> >>I thought that this was just a little too harsh on Cray Research, Inc. and I >>think I'll insert a few facts here (read "blatant advertising material"): >> Oh boy. I'll try not to turn this into a "my computer is bigger than yours", but I thought I might as well defend my old friend, the CM-2. I'll try to keep this germane to this newsgroup also. >> And you get up to 2 100 MegaBYTE per second channels, many, many, many >> 12.5 MegaBYTE per second disk drive channels (which can run simultaneously). >> Hey, how about 4294967000 Bytes of MAIN memory? (Yep that's 4.2 _billion._) >> Oh how about 8 CPUs capable of 2.2 Gigaflops? (That's 2.2G for all 8, not >> for each processor; and this is a demonstrated number for a real science >> application, not a matrix multiply loop) Here's some more facts: On a full CM-2 (65536 1 bit ALUs, 2048 64-bit Weiteks), you have 4 100 MegaBYTE per second channels, any of which could be hooked up to a 50 MegaBYTE per second disk farm (aka a DataVault), or any of various high speed devices. For that matter, some folks in Pittsburgh hooked their CM-2 to their Cray Y-MP the other day. As for main memory, a full CM-2 has 8 gigabytes of main memory. Also, as was reported in the popular press the other day (i.e. NY Times, Boston Globe) the CM-2 just won a contest at (I believe) the University of Tennesee where it displaced the former fastest machine in the world (a Fujitsu machine) with a sustained performance of 5.5 Gigaflops. That was a real science application, not a matrix multiply loop. Oh, as for the matrix multiply loop, a guy at Thinking Machines has one of those little loops that does a sum of squares at about 27 Gigaflops on a full machine. >> >> As an added bonus you get real live UNIX. And those pesky "ps" commands >> run on a real supercomputer cpu, not a teensy weensy processor. (Anyone >> tried multitasking ps over 65000 cpu's yet?) Well, on a CM-2, you run the UNIX on your front end machine, like a SUN-4 or a VAX, and leave the CM-2 for the sort of stuff it was built for. Seems a trifle silly to be running things like ps if it doesn't vectorize or microtask... >> >> You want to try virtual reality with REALLY complicated worlds and tons >> of I/O? How about demonstrated TCP/IP performance, Cray to Cray, of over >> 300 Megabits per second! What? 300 Megabits with TCP/IP? No way!! >> Yep. We do it every day. Well, for VR work, I'm a little more concerned about I/O to a Silicon Graphics or a Stardent. Which you can do from a CM-2, albeit at only a few megabytes a second. Course, that's not really the CM-2's fault, since it's tough to shove bits off the net much faster into one of those machines. Imagine running your radiosity renderer on the CM-2, and downloading the scene to an SGI to let the hardware scan convert it as you move through the space. Imagine running complex simulations and getting the results of that simulation over to a workstation quickly enough to interact with it in real time. You can do that now on a CM-2. -- --> Michael B. Johnson --> MIT Media Lab -- Computer Graphics & Animation Group --> (617) 253-0663 -- wave@media-lab.media.mit.edu