XANSYS Message: 4189 [Go back to message list] [bookmark on del.icio.us]
No rating yet Subject: Re: platforms Author: Martin Liddle Date: 1999-08-25 04:13:00In article <3F8128AC1C8AD2118EF90008C7B171465B67E6@D...>, Tomaszewski, Paul [DPYUS] writes >Fern@s...; Yes, I've been watching the recent flurry of Linux news, but not >ready to make the step since support is not developed. I'll assume from >your medium [<100,000 DOF] vs. large [1,000,000+ DOF] distinction that above >100k the Wintel boxes start to break a sweat.
I wouldn't agree that above 100,000 DOF Wintel boxes have a problem. I've been running models with 350,000 DOF on a year old Dell PC and have been very happy with the run times. The main factor if you are using the PCG solver is available RAM. There is certainly a point where the Intel 32 bit architecture reaches a limit true but providing you have enough RAM and disk the point must be nearer to 1,000,000 DOF. Remember that PC memory, disk and maintenance are MUCH cheaper than for the equivalent workstation items, so thinking in terms of a computer with 1Gb of RAM and 40Gb of disk is a perfectly feasible and economically viable solution. Also a PC on your desk has the big advantage that you have 100% utilisation and control of it. Shared multi processor compute servers do have some advantages but somehow in the real world the other guy always seems to be able to argue that his job is more urgent and gets placed into a higher priority queue.