Date: Tue, 28 Feb 2006 13:33:32 -0800 From: Peter Wemm <peter@wemm.org> To: freebsd-amd64@freebsd.org Subject: Re: AMD 64 stability Message-ID: <200602281333.33309.peter@wemm.org> In-Reply-To: <43FE010F.8040501@datafarm.de> References: <43FCEF9C.5050308@bluelight.org.uk> <200602230937.44464.vayua@sklinks.com> <43FE010F.8040501@datafarm.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 23 February 2006 10:38 am, V. T. Mueller wrote: > Since the original question was related to 'server' usage... the > compatibilty list tells if the components found on a single board > (along with its BIOS) are supported. It doesn't statethat a specific > combination or future setup will work. > > For example, we just had to find out that iir-based icp controllers > work extraordinarily fine as long as <4G RAM are used. Upgrading > memory beyond 4G turns your big server(s) into one really big > problem. The amd64 platform port shares many (but not all) of the same >4GB ram device driver bugs with the i386+PAE kernels do. Bounce buffer code historically doesn't get much testing on many drivers and if problems are going to happen, they will usually turn up when you are trying to run a 32 bit card with more than 4GB of ram. Many of these problems are mitigated by cards that can do 64 bit addressing, but this also requires driver support. Since not that many of our developers have machines with that much ram, it is an unfortunate fact that those code paths tend to not get quite enough excercise.. sometimes not until it is too late.. ie: after a release. In other words, device driver bugs that appear with >4G ram on amd64 will *also* appear on i386 with >4G ram enabled. The device drivers are mostly shared between the two platforms. -- Peter Wemm - peter@wemm.org; peter@FreeBSD.org; peter@yahoo-inc.com "All of this is for nothing if we don't go to the stars" - JMS/B5
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200602281333.33309.peter>