Just a sample of the Echomail archive
Cooperative anarchy at its finest, still active today. Darkrealms is the Zone 1 Hub.
|    LINUX    |    Torvalds farts & fans know what he ate    |    8,232 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 7,622 of 8,232    |
|    Mike Miller to Benny Pedersen    |
|    xeon vs epyc    |
|    06 Sep 22 08:16:41    |
      MSGID: 1:154/30 63174c30       REPLY: 2:230/0 6310a171       TID: SBBSecho 3.15-Linux ensemble/e7dfe4318 Sep 1 2022 GCC 4.8.5       BBSID: ENSEMBLE       CHRS: CP850 2       TZUTC: -0500       CHRS: ASCII 1       Hello Benny!              01 Sep 22 12:08, you wrote to all:                      BP> Linux mx 5.15.63-gentoo-dist #1 SMP Thu Aug 25 12:40:44 -00 2022        BP> x86_64 Intel(R) Xeon(R) CPU E5-2697 v4 @ 2.30GHz GenuineIntel        BP> GNU/Linux Linux localhost 5.19.6-gentoo-dist #1 SMP PREEMPT_DYNAMIC        BP> Wed Aug 31 18:48:13 -00 2022 x86_64 AMD EPYC 7642 48-Core Processor        BP> AuthenticAMD GNU/Linux                      We've been switching over to Epyc boxes at work, and it's been a bit of a       nightmare, although mostly that's been due to software limitations.               We started out with dual CPU 64-core Epyc CPUs, and ran into limitations with       applications that couldn't deal with 256 processors. We had to manually pin       each thread/appliaction to a core / set of cores.               We eventually switched over to purchasing single-CPU 64 core EPYC boxes,       which resolved our issues with CPU pinning for the most part.                     However, every single EPYC box we're running has to have IOMMU disabled in       BIOS. Otherwise, after about 3 months of running the servers will start       spewing "AMD-Vi: Completion wait loop timed out" errors. This will cause the       pcie devices to rapidly disable/re-enable, which knocks out networking. We've       yet to nail down the actual cause of the issue, and it doesn't seem to matter       what kernel version we're running. The odd thing is that it will happen in       short bursts, groups of servers (usually assigned to the same application)       will start blowing up every hour or two, one after another. Not a fun thing       when I'm on call, because I swear it starts happening overnight every time. :D                                   Mike                     ... Dancers do it with rhythm.       === GoldED+/LNX 1.1.5-b20220504       --- SBBSecho 3.15-Linux        * Origin: War Ensemble - warensemble.com - Appleton, WI (1:154/30)       SEEN-BY: 1/19 123 15/0 16/0 19/37 90/1 103/705 105/81 106/201 114/709       SEEN-BY: 120/340 616 123/10 130 131 124/5016 129/305 331 153/757 7001       SEEN-BY: 153/7715 154/0 10 30 40 50 70 700 203/0 218/700 220/90 221/0       SEEN-BY: 221/6 226/18 227/114 201 229/111 112 113 206 275 310 317       SEEN-BY: 229/400 424 426 428 452 470 550 616 664 700 240/77 420 5411       SEEN-BY: 240/5824 5832 5853 266/512 280/464 5003 5006 282/1038 292/854       SEEN-BY: 292/8125 301/1 310/31 317/3 320/119 219 319 322/0 757 326/101       SEEN-BY: 341/66 234 342/200 396/45 423/120 460/58 633/280 712/848       SEEN-BY: 770/1 2320/105 2432/390 2452/250 2454/119 3634/12 5020/545       PATH: 154/30 10 280/464 240/5832 320/219 229/426           |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca