Date: Sat, 27 Mar 2021 10:54:53 +0000 From: Santiago Martinez <sm@codenetworks.net> To: freebsd-current@freebsd.org Subject: Re: freebsd 13 ryzen micro stutter Message-ID: <9638abfe-fd9e-6855-b3c7-ee796d1c5ce5@codenetworks.net> In-Reply-To: <54768f44-c110-d58a-1e7b-61843cde754e@codenetworks.net> References: <d793c8b7-e0e2-e0d9-3ecd-923134d15bf3@twcny.rr.com> <20210323101146.18c9a969@bsd64.grem.de> <ED533759-D326-45D5-8AE2-51CB10763A1E@unrelenting.technology> <677da81d7af9781be6b46c7cba4efd05@riseup.net> <54768f44-c110-d58a-1e7b-61843cde754e@codenetworks.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, i have the same output as @Nils B. If i run with steal =2 and dtrace the micro stutter doesn't happen but as soon as i stop the dtrace script it the stutters come back again. Santi On 3/27/21 10:39 AM, Santiago Martinez wrote: > Hi there, my pause are not that often ( with steal_thresh=2) , maybe > is around 30 secs or so. > > If i set the thresh to 1 or 0 they go away. Now im trying with =2 and > schedgraph.d. > > Santi > > > On 3/25/21 10:04 AM, Alastair Hogge wrote: >> On 2021-03-23 17:34, myfreeweb wrote: >>> On March 23, 2021 9:11:46 AM UTC, Michael Gmelin <freebsd@grem.de> >>> wrote: >>>> >>>> On Mon, 22 Mar 2021 23:50:52 -0400 >>>> monochrome <monochrome@twcny.rr.com> wrote: >>>> >>>>> After about 8 months of struggling to narrow this down I did another >>>>> search and saw this: >>>>> >>>>> https://www.reddit.com/r/freebsd/comments/lc8fwo/freebsd_13_vega_64_micro_stutter_in_x/ >>>>> >>>>> >>>>> I haven't seen this come up here so I thought I would bring it up. >>>>> >>>>> My story started sometime before around August last year when synergy >>>>> started getting really annoying with stuttering. Pretty sure it >>>>> wasn't like that when I first started tracking 13-current in around >>>>> May 2020 (I was on 12 with scfb for a long time before that with no >>>>> issues), but since then I have tried to eliminate as many variables >>>>> as possible. First I switched to barrier instead of synergy. Shortly >>>>> after that I realized it was happening all the time and not just a >>>>> network problem, I started using foobillard to verify during tests. I >>>>> tried different RAM combinations, different network cards, a variety >>>>> of RAM timings, stripping rc.conf etc, powerd settings, also scfb, >>>>> with no effect. It is observable with ping -f, a dot or two appears >>>>> every time it glitches. It seemed much better with RC2, but now with >>>>> RC3 it seems to be back with a vengeance, and since its my main >>>>> workstation and barrier/synergy server host for several machines, it >>>>> is unbearable to use. Both Win10 and devuan3 on the same machine are >>>>> smooth with no issues. Any feedback or info would be appreciated. >>>>> >>>>> Hardware: >>>>> ASRock B450M Pro4 >>>>> Ryzen 2400G, no OC >>>>> 32M DDR4-2933 >>>>> Onboard Vega GPU, drm-fbsd13-kmod >>>>> _______________________________________________ >>>> Without knowing much about the issue at hand, just a few ideas: >>>> >>>> Sysctls I would look at: >>>> - raising *.cx_lowest >>>> - different kern.eventtimer.timer >>> None of these should be an issue, but: >>> >>> sysctl kern.sched.steal_thresh=1 >>> >>> For some reason with the default value of 2, I'm seeing weird >>> stuttering in youtube videos, games, etc. on a 5950X system. 1 (or 0, >>> IIRC) works fine. >> I have noticed this too, firefox exhibits the issue frequently, however >> I have >> found it most noticeable in games/dhewm3[0] or Yamagi Quake 2[1]. There >> is a distinct pause every 3 or 4 seconds where the simulation advances >> significantly. Thanks for the kern.sched.steal_thresh pointer that has >> sorted >> the issue out for now. >> >> $ dmesg | egrep '(CPU:|avail memory)' >> CPU: AMD Ryzen 9 3950X 16-Core Processor (3500.02-MHz >> K8-class CPU) >> avail memory = 66779394048 (63685 MB) >> With an AMD NAVI 8GiB GPU, the host should be able to play either game >> well enuf, >> Quake 2 came out in 1997 and Doom 3 2004. I am sure Quake 2 used to play >> without issues on my old AMD bulldozer. >> >> >> 0: https://www.freshports.org/games/dhewm3/ >> 1: https://github.com/yquake2/yquake2 >> _______________________________________________ >> freebsd-current@freebsd.org mailing list >> https://lists.freebsd.org/mailman/listinfo/freebsd-current >> To unsubscribe, send any mail to >> "freebsd-current-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9638abfe-fd9e-6855-b3c7-ee796d1c5ce5>