From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 4 Sep 2001 16:58:53 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 4 Sep 2001 16:58:44 -0400 Received: from postfix2-2.free.fr ([213.228.0.140]:7432 "HELO postfix2-2.free.fr") by vger.kernel.org with SMTP id convert rfc822-to-8bit; Tue, 4 Sep 2001 16:58:28 -0400 Date: Tue, 4 Sep 2001 22:55:52 +0200 (CEST) From: =?ISO-8859-1?Q?G=E9rard_Roudier?= X-X-Sender: To: Jason Howard Cc: Subject: Re: AMD 760 PCI Problem In-Reply-To: <3B95348E.F8AE6E08@spectsoft.com> Message-ID: <20010904224658.Y2154-100000@gerard> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 4 Sep 2001, Jason Howard wrote: > Hello All, > > I am having trouble getting decent PCI throughput on either of my AMD > system (both using the AMD 760 chipset.) I am moving a high definition > frame into a pci frame buffer. This process takes somewhere around 100 > ms on my Dual Xeon system, however on the two AMD systems I have tried > it is taking about 1 sec (1000 ms). > > The two systems I have tried are a single processor athlon system and a > dual processor athlon system. (Both use the AMD chipset .. the single > uses the AMD760 and the dual uses the AMD760MP). Both boxes are running > kernel 2.4.6. > > Any suggestions? Check the settings of the MTRR's against the frame buffer window. If they appear not to be configured correctly, you just have to ask the kernel to fix the thing (a MTRR defined for Write Combining must contain the address range of the frame buffer). see: linux/Documentation/mtrr.txt by Richard Gooch for details. Gérard.