From owner-freebsd-questions Thu Mar 28 08:48:24 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id IAA17451 for questions-outgoing; Thu, 28 Mar 1996 08:48:24 -0800 (PST) Received: from horst.bfd.com ([204.160.242.10]) by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id IAA17443 for ; Thu, 28 Mar 1996 08:48:21 -0800 (PST) Received: from harlie.bfd.com (bastion.bfd.com [204.160.242.2]) by horst.bfd.com (8.7.3/8.7.3) with SMTP id IAA27460 for ; Thu, 28 Mar 1996 08:49:21 -0800 (PST) Date: Thu, 28 Mar 1996 08:50:22 -0800 (PST) From: "Eric J. Schwertfeger" To: questions Subject: Triton EIDE interface support Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk I'm currently running 2.1R on a triton based motherboard. I'm in the process of converting over from Linux on this machine, and I was curious. In linux, using hdparm, I can enable multi-sector transfers and 32 bit transfers on my primary hard drive, which considerably increases the drives performance (I'm not running the 1.3.X tree yet, so I don't know what effect the triton DMA interface would have on the feel of the system). I plan on compiling some benchmarks tonight to find out where my performance is under FreeBSD, but I don't have similar numbers for Linux. So, between 2.1R, stable, and current, how much support for these features are in FreeBSD? Details: (irrelevant unless you want to know why I'm asking) My machine is a P100, 16M EDO ram, 256K burst cache, currently with both an ST5850A and an ST32140 on it. I'm switching from Linux to FreeBSD, because PPP on linux has been hanging, and in my FreeBSD tests, it never hung (though it did hang up a few times when I thought I had the timeouts disabled). Anyway, linux has a hdparam program which has the ability to benchmark the sequential throughput of a device (bypasses the file system) reading a large (32M?) portion of the hard drive. It also lets you tune things like turning on multisector transfers, 32 bit transfers, and in the case of the triton chipset and 1.3.X kernels, DMA transfers. The ST5850A as master benchmarked at about 5M/sec at the start of the disk, in both single sector, 16 bit transfers, and multisector 32 bit transfers. the ST32140 as slave benchmarked at about 2.5M/sec single/16, and over 5M/sec multi/32. I *WANT* that performance, and will become a kernel hacker if necessary. My wife's ST5850A died, so I have to loan her mine (yes, I love the ST5850A. 5400 RPM and runs like it. Got one at work too). When it gets back, I plan on puting my swap on it, and the rest will be low access files (ports/src?). (Or very low access, DOS/WIN :-)