From mboxrd@z Thu Jan 1 00:00:00 1970 From: GoatZilla Subject: Re: Manual PST settings for "unrecognized CPUs" Date: Tue, 17 Aug 2004 14:30:00 -0400 Sender: cpufreq-bounces@www.linux.org.uk Message-ID: <39e348480408171130537c34e6@mail.gmail.com> References: <39e3484804081616133c2765ce@mail.gmail.com> <20040817062634.6C2404BE94@nathan.muc.de> Reply-To: GoatZilla Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20040817062634.6C2404BE94@nathan.muc.de> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: cpufreq-bounces+glkc-cpufreq=gmane.org@www.linux.org.uk Content-Type: text/plain; charset="us-ascii" To: cpufreq@www.linux.org.uk Ah, I just found your patch back in June; looks like good stuff. I'll give it a try when I get Linux onto my machine. It's odd that you couldn't get below 9x on the multipliers. With CPUMSR I was able to take my Duron all the way down to 3x (I underclocked the FSB as well to 100MHz). The curious thing was it was stable at 3x, locked up at 4x, and was stable again at 5x. I would think the core voltage that runs the CPU at full speed should be acceptable at the lower speeds. As far as the max FID is concerned, you've probably figured it out by now, but the max FID that can be set by software is controlled by the L6 bridges on the CPU package. On Tue, 17 Aug 2004 08:26:34 +0200 (CEST), Harald Milz wrote: > GoatZilla wrote: > > I saw a patch fly by a short while ago where a user added some > > settings for an unrecognized config. Is this the only way to hack in > > a PST? > > Actually there were (at least) two patches - Bruno's patch which > implemented additional /proc fs entries to send frequency settings to, and > mine which uses command line parameters for the powernow-k7 module. You can > find them in the ML archive. Internally, they do basically the same thing. > Mine is a little inferior as far as sanity checks. But It Works For Me > [TM]. > > Please be advised that you _could_ potentially fry your CPU if you use them > :-) But IMHO it's quite unlikely. > > What these patches can't do is set the VID on most (?) desktop boards > because of a lack of hardware support. But you should be able to lower the > FID to a certain extent, and the power consumption is linearly proportional > to the frequency, after all. > > -- > "Proper Preparation Prevents Poor Performance." > > _______________________________________________ > Cpufreq mailing list > Cpufreq@www.linux.org.uk > http://www.linux.org.uk/mailman/listinfo/cpufreq >