From owner-freebsd-questions@freebsd.org Fri Nov 13 17:33:53 2015 Return-Path: Delivered-To: freebsd-questions@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 44E85A2E640 for ; Fri, 13 Nov 2015 17:33:53 +0000 (UTC) (envelope-from smithi@nimnet.asn.au) Received: from sola.nimnet.asn.au (paqi.nimnet.asn.au [115.70.110.159]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id AC81D1872 for ; Fri, 13 Nov 2015 17:33:51 +0000 (UTC) (envelope-from smithi@nimnet.asn.au) Received: from localhost (localhost [127.0.0.1]) by sola.nimnet.asn.au (8.14.2/8.14.2) with ESMTP id tADHXhrJ092350; Sat, 14 Nov 2015 04:33:44 +1100 (EST) (envelope-from smithi@nimnet.asn.au) Date: Sat, 14 Nov 2015 04:33:43 +1100 (EST) From: Ian Smith To: john.haraden@yahoo.com cc: freebsd-questions@freebsd.org Subject: Re: est In-Reply-To: Message-ID: <20151114022704.Y55748@sola.nimnet.asn.au> References: <20151016224929.Q15983@sola.nimnet.asn.au> <20151017232247.P15983@sola.nimnet.asn.au> <20151019175116.X15983@sola.nimnet.asn.au> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 13 Nov 2015 17:33:53 -0000 On Wed, 21 Oct 2015 06:41:04 -0700, john.haraden@yahoo.com wrote: John, sorry for the long delay. Please reply both to the list (freebsd-questions@freebsd.org) and myself if you can. > Subject: Re: (strange) Failure Notice (fwd) > (1). I know nothing about these strange emails, and I do not receive them. Ok, don't worry about those. I've restored the subject to original. > (2). Est first appeared in the base system of version 10.1. Not so; EST goes way back. I still have a FreeBSD appliance running FreeBSD 5.5 from just over 10 years ago; cpufreq(4) and est(4) were available then. I have an 8.2 system using cpufreq using acpi_perf (an older P3-Mobile CPU, before EST) with powerd, and a stable/9 Core2Duo system using EST .. none of this is new, though updated over time. > If powerd is not specifically set, powerd assumes the > high-performance option. No, powerd(8) defaults to hiadaptive mode on AC power; check the manual. > With multiple cores and processors, I do not know whether it is a > good thing or a bad thing. If, as you originally suggested, you want to run at maximum performance all the time, and don't care about system temperature, then powerd_flags '-a max' will force that, ie always run at maximum frequency. However there are caveats about doing so, especially on some processors, and '-a hiadaptive' (or hadp) will jump to max freq quickly enough when there's any load, and stay up there till load is very low; further you can adjust -r)unning and -i)dle percentages to best suit your needs. You can also use powerd's -M switch to set a lower maximum, useful on some CPUs (eg Core2Duo) where turbo speed (here "2401") is not desired due to likely overheating on big long jobs, like buildworld. I use -M 2400 as advised by Warner Losh, which actually improves performance while reducing heat here, but perhaps not so with your 24-core Xeon. > I forwarded you an article on both turbo-charging and disabling est. https://www.ateamsystems.com/tech-blog/increase-freebsd-performance-with-powerd/ There are parts of that article that don't ring true to me, not that I know anything about latest Xeons and how they handle Turbo mode. I do know that FreeBSD never, so far, runs different cores at any different frequencies, which seems to be implied there. There may also be some confusion regarding P-states (frequencies) and C-states (idle states) though he doesn't refer to them in those terms. /etc/defaults/rc.conf already has what you likely want for C-states: performance_cx_lowest="HIGH" # Online CPU idle state performance_cpu_freq="NONE" # Online CPU frequency Where HIGH is C1 (hlt) and NONE means don't update CPU freqs, leave it as either originally set on boot, or as since set or managed by powerd. And given that post implies it only turbo-boosts some cores when there's less load on some others (mostly to stop it overheating, I gather) then it makes little sense to fully load all cores then claim a 12% boost in performance .. but it doesn't say compared to what? Colour me skeptical. What it does make clear is that you need EST enabled in BIOS (and not disabled in loader.conf) to take any advantage of Turbo mode, or indeed to be able to set (or report) CPU frequencies at all. Powerd relies on EST on these cpus. I know of no reason to disable EST nor to think that doing so might increase performance. > (3) I am not currently on the subscription list. How do I join the > subscription list. https://lists.freebsd.org/mailman/listinfo/freebsd-questions If still in doubt, I suggest posting afresh to the freebsd-acpi list, perhaps after reviewing recent archives from and/or subscribing at: https://lists.freebsd.org/mailman/listinfo/freebsd-acpi which is one place ACPI, powerd and bus developers tend to hang out. cheers, Ian