Date: Mon, 21 Nov 2022 06:37:48 +1100 From: Ian Smith <smithi@nimnet.asn.au> To: Kevin Oberman <rkoberman@gmail.com> Cc: mike tancsa <mike@sentex.net>,questions@freebsd.org Subject: Re: RELENG_13 and min cpu frequency Message-ID: <5E64A5AC-3634-4336-B0A9-E98962D8BE79@nimnet.asn.au> In-Reply-To: <CAN6yY1v59%2BUeHuOZBVeK%2BsN694u=xjA7CyXOiOMEXtkgZfYQFA@mail.gmail.com> References: <9d17ea30-4b10-2aa3-9d09-017da7423844@sentex.net> <BF5CF7F7-DBB3-45A5-9FFA-7595EF41D494@nimnet.asn.au> <CAN6yY1v59%2BUeHuOZBVeK%2BsN694u=xjA7CyXOiOMEXtkgZfYQFA@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 20 November 2022 9:48:04 am AEDT, Kevin Oberman <rkoberman@gmail=2Ecom> = wrote: > On Fri, Nov 18, 2022 at 9:41 PM Ian Smith <smithi@nimnet=2Easn=2Eau> > wrote: >=20 > > On 19 November 2022 8:28:45 am AEDT, mike tancsa <mike@sentex=2Enet> > wrote: [=2E=2E=2E] > > > dev=2Ecpu=2E0=2Efreq_levels: 3400/-1 > > > dev=2Ecpu=2E0=2Efreq: 1302 > > > > So, hmm=2E powerd reads and sets values in dev=2Ecpu=2E0=2Efreq_leve= ls, > hence > > can't change it to anything but 3400=2E > > > > So how is cpu=2E0=2Efreq changing at all? Is cpufreq using est(4)?= =20 > Are there > > more freqs in dev=2Eest=2E0=2Efreq_settings? > > > > Or is it using a _relative_ driver like acpi_throttle or p4tcc?=20 > Default > > /boot/device=2Ehints has both of these disabled for good reason > (Kevin > > Oberman can elaborate) > > > > > sysctl -A debug=2Ecpufreq > > > debug=2Ecpufreq=2Everbose: 0 > > > debug=2Ecpufreq=2Elowest: 0 > > > > > > > > > CPU is CPU: Intel(R) Xeon(R) E-2226G CPU @ 3=2E40GHz > > > (3400=2E00-MHz K8-class CPU) > > > > Also noted that the usual +1 freq for turbo mode isn't there=2E=20 > Strange? > > (I don't speak Xeon though) > In recent processors, things have changed radically WRT CPU frequency > and thermal behavior=2E >=20 > I have an Alder Lake system (12th gen) and a Comet Lake (10th gen)=2E > Both > now show a single frequency=2E On my Alder Lake I see: > dev=2Ecpu=2E?=2Efreq_levels: 2496/-1 > This is shown for all CPU devices (threads) whether P or E cores=2E > Thermal > control does change these seeds continually=2E I have seen E cores at > 900M, > but rarely=2E I see them as 1=2E2G during buildworld=2E The P cores are= at > 1=2E3G=2E > I have never seen ANY core running at over 2=2E5G=2E When I ordered my > T16, the > specs showed minimum, default, and maximum frequencies, but I can't > find > them any longer=2E All I can find are "Minimum" and "Max Turbo"=2E It's > not > clear to me what the significance of this is, but it was clearly a > decision > of Intel=2E To me, it implies that Turbo is really no longer distinct , > but > considered a portion of a continuum of frequencies=2E >=20 > Just looked again=2E buildworld is still running, and see all CPUs are > at > 2094, 2095, 2096, and 2097=2E A few seconds later, 1997 to 2094=2E Al= l > CPUs > are not always running at very similar speeds=2E This is very different > from > anything I have seen on older processors=2E I see no distinction > between P > and E cores=2E I also suspect that powerd is no longer relevant for > modern > CPUs=2E Anyone have a pointer to information on CPU frequency and > thermal > management on newer Intel processors? Thanks Kevin=2E I went hunting through freebsd-acpi archives, and found y= our related comments in https://bugs=2Efreebsd=2Eorg/bugzilla/show_bug=2Ecgi?id=3D264775 Seems we need some overview on what FreeBSD can and can't do regarding cha= nging CPU speeds, or whether it's even appropriate to try doing so on these= latest processors? I'm well behind; few years ago there wasn't even the concept of different = cores running different frequencies=2E=20 Maybe the best we can do with these is making sure that fans are up to kee= ping the system cool enough =2E=2E Not that this helps mike's problem with apparent packet loss, but it's har= d to see if that may be related=2E=20 cheers, Ian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5E64A5AC-3634-4336-B0A9-E98962D8BE79>