smithi at nimnet.asn.au
Mon Nov 16 07:50:01 UTC 2015
On Sat, 14 Nov 2015 08:18:30 -0600, Adam Vande More wrote:
> On Sat, Nov 14, 2015 at 6:33 AM, Ian Smith <smithi at nimnet.asn.au> wrote:
> > > > I do
> > > > know that FreeBSD never, so far, runs different cores at any different
> > > > frequencies,
> > >
> > > FreeBSD has supported TurboBoost for years.
> > Of course. But we do NOT support setting different CPU frequencies on
> > different cores.
> Well of course, but that's not what you said initially which is what
> prompted my response.
You cut a fine line there, but OK. I'll slice a little finer too :)
> > https://www.ateamsystems.com/tech-blog/increase-freebsd-performance-with-powerd/
> > Do I need to detail the several incorrect assumptions at play above,
> > regarding FreeBSD's role in interacting with the CPU/s re TurboBoost in
> > particular and SpeedStep in general?
> I do not believe you need to do that. That link or any other external
> source has never been the basis of my response.
No, but I think it may have contributed to John's initial concern that
enabling EST might somehow degrade performance, which I've been trying
to address all along. I'm not sure how successfully.
> I agree the link contains some dubious claims and methodology, however the
> heart of it is sort of close enough to accurate.
Well, sort of, except that parts of it might make one think that there's
some sort of constant interaction between the OS and the CPU, that would
indeed burn lots of cycles, whereas in really there are many orders of
magnitude between powerd's sampling of the average load over all cores
- at 4Hz by default - and consequent adjustments of CPU base frequency,
and the much faster internal calculations that determine how many cores
are relatively busy or idle, with consequent adjustments to individual
cores' clocking when TurboBoost is enabled - which I gather only occurs
when cpufreq(4) has already selected maximum (XX01) frequency.
E.g: "windows of opportunity quickly close and open, and the CPU and OS
work in rapid conjunction to calculate and exploit them. TurboBoost will
not engage if SpeedStep information is not being received from the OS,
so it is crucial to enable the powerd(8) service and ensure that
SpeedStep is performing properly".
For another thing, powerd is not required if the frequency is always set
to MAX (ie 2601 in John's case) to enable TurboBoost; it will be on all
the time, so up to the CPU/s to adjust (over)clocking internally, no?
Not that I'd ever want to waste power/heat unless to meet a demand.
> You don't have TurboBoost, so what ever results you have can't be applied
> to a TurboBoost generalization. My suggestion for you to re-enable
> TurboBoost was also flawed.
Yeah, well you prompted me do some more research, thanks. On Core2 it
was called Intel Dynamic Acceleration (IDA) and apparently only boosts
one core when the other is in idle state C3. I don't see how that would
work properly even on single-core tasks (unless pinning that task to one
CPU), as the scheduler otherwise shares work between available cores.
> Getting rid of my Core2 stuff was a happy time for me. Don't get me wrong,
> it was great when it came around but it's so slow and power hungry compared
> to even a sandybridge.
Lucky you :) The X200 is still my fastest machine, running about 6 x
faster than my older P3-M workhorse. The SSD helps greatly of course.
> > Don't take my word for it .. please read:
> > http://lists.freebsd.org/pipermail/freebsd-mobile/2015-February/013240.html
> > then feel free to argue with Warner about advice that worked for me :)
> I don't even understand what there would be to argue with him about. His
> particular setup may have heat issues when utilizing TurboBoost for an
> extended period/load. My systems do not and they operate in turbo mode
> much of the time, at least as far as I've checked it. It is not something
> I monitor continuously. My current main workstation, a m6600, can have
> heat issues if the external video slot is filled. Without it, it works
> great under all conditions.
Warner's was a T400, a bigger brother to my X200, also a Core2Duo, so
his advice was salient.
> And the OP's post which started this was nonsensical so I have no comment
> in that regard other than EST is not magic. It doesn't just go off and on.
Indeed. Sometimes I'm driven to try to help; sometimes it backfires ..
More information about the freebsd-questions