From owner-freebsd-questions@FreeBSD.ORG Sat Sep 3 20:38:36 2011 Return-Path: Delivered-To: questions@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id BCFF2106564A for ; Sat, 3 Sep 2011 20:38:36 +0000 (UTC) (envelope-from brett@lariat.net) Received: from lariat.net (lariat.net [66.119.58.2]) by mx1.freebsd.org (Postfix) with ESMTP id 55B2D8FC1C for ; Sat, 3 Sep 2011 20:38:36 +0000 (UTC) Received: from WildRover.lariat.net (IDENT:ppp1000.lariat.net@lariat.net [66.119.58.2]) by lariat.net (8.9.3/8.9.3) with ESMTP id OAA27146 for ; Sat, 3 Sep 2011 14:38:33 -0600 (MDT) Message-Id: <201109032038.OAA27146@lariat.net> X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9 Date: Sat, 03 Sep 2011 14:38:32 -0600 To: questions@freebsd.org From: Brett Glass Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed Cc: Subject: Huge interrupt overhead reported after RAM added to Atom-based system X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 03 Sep 2011 20:38:36 -0000 I've just seen something very peculiar. I have here a dual Atom (D525) system which was running with 1 GB of RAM, and this morning I put a 4 GB module into the system instead. Suddenly, the systat(8) and top(8) commands were both reporting bursts of interrupt overhead as high as 25% of total CPU capacity. Yet, in the display from the systat -vmstat option, no additional interrupts were appearing on the right hand side of the screen where interrupt sources were listed. The system is running FreeBSD 8.1-RELEASE. The documentation for several motherboards online mention that the Atom can be run with 4 GB of RAM with a "64 bit operating system." I can't tell whether something non-obvious is going on under the hood -- either in the chipset or in the CPUs -- that's racking up overhead, or if the interrupt overhead doesn't exist at all and the reported CPU load is an artifact of some weirdness in the kernel. I need to know, though, before I deploy the system... so I'd appreciate any advice or ideas from any kernel experts who might be reading messages here. --Brett Glass