Received: from casbah.acns.nwu.edu by karazm.math.UH.EDU with SMTP id AA22771 (5.65c/IDA-1.4.4 for ); Tue, 22 Oct 1991 19:54:37 -0500 Received: from neworder.ils.nwu.edu by casbah.acns.nwu.edu (4.1/SMI-ACNS-1.03) id AA25265; Tue, 22 Oct 91 19:50:27 CDT Return-Path: Received: by neworder.ils.nwu.edu (4.0/SMI-4.0) id AA00261; Tue, 22 Oct 91 19:40:56 CDT From: newton@neworder.ils.nwu.edu (Dave Newton) Message-Id: <9110230040.AA00261@neworder.ils.nwu.edu> Subject: Transputers... To: glove-list@karazm.math.uh.edu Date: Tue, 22 Oct 91 19:40:55 CDT X-Mailer: ELM [version 2.3 PL11] Howdy... I'm trying to figure out why nobody on here has proposed using those $400 T400 transputer cards hooked into a slave-PC to use for rendering... My idea was basically have a T400 card w/ 1Meg ($396) and perhaps a few slave cards ($296?). Each transputer has 10Mips sustained, 20 Mips peak. The C compiler will automagically take advantage of other processors if the code is set up in a parallel fashion. Now, color me stupid, but wouldn't this be a cheap way to get a whole bunch of power? The transputers could either be used in the main PC or put out in a slave PC somewhere. I was planning on getting two dirt-cheap PC boards, three transputer boards (one for the main CPU, one for each slave) and communicate to the slave T400 boards using the on-board 20M data link... Let the slave boards handle the image calculation and spit it out via J. Random NTSC-board.