Date: Fri, 26 Aug 2005 08:55:14 -0700 From: "Kevin Oberman" <oberman@es.net> To: "Dan Langille" <dan@langille.org> Cc: freebsd-mobile@freebsd.org Subject: Re: long flight; need battery power! Message-ID: <20050826155514.98AB05D07@ptavv.es.net> In-Reply-To: Your message of "Thu, 25 Aug 2005 22:42:47 EDT." <430E4967.20355.1BF72257@localhost>
next in thread | previous in thread | raw e-mail | index | archive | help
> From: "Dan Langille" <dan@langille.org> > Date: Thu, 25 Aug 2005 22:42:47 -0400 > Sender: owner-freebsd-mobile@freebsd.org > > In September, I'll have a couple of long flights. I'll be using my > IBM ThinkPad T41 to catch up on some long delayed FreshPorts > enhancements. Apart from charging whenever I can, I'm afraid I'll > get only about 140-150 minutes from my battery. > > Apart from buying a second battery, any ideas? This gets a bit long, but I've been spending a lot of my spare time testing this stuff and it's proven interesting. What version of FreeBSD are you running? V6 has a lot better power management than V5 and current is better, still. If you are running V6, get the powerd from HEAD and use that or just keep the CPU fairly slow. Background: Most newer laptop CPUs have the capability of running at two or more actual clock rates. This makes a BIG difference in power consumption. It can also be throttled (by various means) to a portion of the actual clock speed. This can save power, but the impact is far less than with reducing the actual CPU clock rate. 'sysctl dev.cpu.0' will show (among other things) the available performance settings. Please remember that these are 'pseudo-frequencies' obtained by combining actual CPU clock settings with throttling and not just actual clock speed changes. You can use several tools to see the actual CPU speed. I use the gkx86info2 plugin for gkrellm2. On my T30 I have determined that the a busy system running at low performance uses more power than a system with the same load at a higher performance. I mean the same actual load. I was testing by playing MP3s which provide a fairly steady load. At 300 MHz, it runs about 65% of capacity. At 1.2 GHz it runs about 17%. At 1.8 GHz it runs only 10%. Lowest power utilization was at 1.2 GHz. My system has 2 actual clock speeds (1.8 and 1.2 GHz) and uses TCC (similar to throttling) to provide the added speeds. As long as the system is close to idle, the slower the clock, the better. But, if it's busy, the higher rate with the 1.2 GHz clock speed is the winner. This is not really intuitive, but makes sense if you look at just what the CPU is doing. Running slowly saves power, but halted saves even more and, when the MP3 is plying a a low CPU performance, it seldom is halted. When it's playing at 1.2 GHz, it finishes each operation and can halt for a while until the next buffer is processed, thus running at lower power. If you are just editing, a very low speed is the winner. Turn off powerd (if you are running it) and just set he CPU to the lowest performance available. If you are compiling, bump it up to the lowest native CPU speed. For debug, it depends, but I suspect you want lowest, again. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050826155514.98AB05D07>