Date: Tue, 18 Apr 1995 20:42:02 -0700 (PDT) From: Julian Elischer <julian@ref.tfs.com> To: rcarter@geli.com (Russell L. Carter) Cc: asami@cs.berkeley.edu, hackers@FreeBSD.org, rcarter@geli.com Subject: Re: Case for FreeBSD presentation docs? Message-ID: <199504190342.UAA06958@ref.tfs.com> In-Reply-To: <199504190325.UAA01659@geli.clusternet> from "Russell L. Carter" at Apr 18, 95 08:25:55 pm
next in thread | previous in thread | raw e-mail | index | archive | help
could you let us know what some of these items are? dgemm (data general emmulator?) BLACS? BLAS? LAPACK? pvm3? mpich? nxlib? julian > > In addition to the other observations: > > 1. Linux has dreadful nfs performance. I can supply ample details. > But maybe you should ask the Linux folks. There doesn't appear to > be hope anytime soon, either. > 2. Linux ncr scsi (up until a month or so ago, maybe still) is broken. > It is a sad thing to watch the ncr list. And what about performance 8-0. > 3. I am running a 4 cpu cluster with scalapack and FreeBSD-somewhat-current. > Just starting to benchmark it. BLACS, PB-BLAS, LAPACK all build and test > fine. > 4. My company just bid for a project that uses Intel Paragons up till now, > but price/performance with FreeBSD and the ASUS TP4-PB is 8x better (for > the same performance). > 5. I have pvm3, mpich, and nxlib up and running real problems. > 6. And hot off the presses: dgemm runs 14-23 MFlops/sec with gcc-i2.6.3. > 7. You can get > 5MB/s from commodity disks. (Maybe that weird non-local > memory caching idea is not so important ;-) > 8. Matt's comments on 100BaseT vs. FDDI performance are true, but the > price is not similar. But the upshot is there are at least 2 10MB/s > technologies supported. (Thanks to Matt). > > Succintly, I can build scalable 2k Flops/sec/$ systems using FreeBSD. Not even > the most recent J-90 comes close. > > Cheers, > Russell >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504190342.UAA06958>