Date: Sun, 17 May 2020 02:41:02 -0700 From: Mark Millard <marklmi@yahoo.com> To: Philipp Klaus Krause <pkk@spth.de> Cc: freebsd-arm@freebsd.org Subject: Re: Recommended arm hardware (mostly for compilation)? Message-ID: <C088F416-80FA-45E3-A61E-8E7EB61B94ED@yahoo.com> In-Reply-To: <f7a88252-fa01-69a1-d442-747d69541d89@spth.de> References: <1b109252-3232-05c4-e1e0-2fea4739583d@spth.de> <47A5AF8A-1EB9-4721-94D5-B0A0A6FFF69B@unrelenting.technology> <9E10C464-474F-489A-A667-B263A64E78F2@googlemail.com> <EXbrttrQC0_YXokurrTuHT6YcY3XC0JhXIkklq6vP6OmhlDw2DSXWz8B3_e_9kDFZowErA8gjjvQrAe4w5t4M8J1Fmfb-ev4uIgI9r7wW4w=@a9development.com> <f7a88252-fa01-69a1-d442-747d69541d89@spth.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2020-May-17, at 00:21, Philipp Klaus Krause <pkk@spth.de> wrote: > Am 14.05.20 um 17:30 schrieb Dan Kotowski: >> I have a SolidRun HoneyComb workstation board based on the LX2K and >> can confirm it's quite powerful, however I have only been able to get >> Linux-based systems running - I have not been able to get FreeBSD >> working, but am working on it. SolidRun provides all the sources >> necessary to do so, it's just a matter of interpreting all the >> documentation. > > Do you have any experience of how this board performs vs the MACCHIATObin? > > Both have A72 cores at 2 Ghz. The LX2K has 16 cores vs 4 cores on the > MACCHIATObin, but I wonder if typical compilation workloads are able to > use all cores vs. being I/O or memory-bandwidth-starved. > > The other main advantage of the LX2K would be dual-channel memory. But I > wonder if that worth the price difference. I do not have LX2K experience, but . . . An illustration of the type of question: I have access to both a MACCHIATObin DoubleShot (A72 based) and a OverDrive 1000 (A57 based). Turns out the OverDrive 1000 is faster at parallel builds overall vs. the MACCHIATObin being so for single-threaded activity. The OverDrive 1000 has dual channel memory. Both are configured to have SATA SSDs of the same type and the I/O does not seem to contribute differently for builds. (The file system context is UFS for both.) As an example (from memory), a poudriere bulk build session set up to do the same on both got results like: OverDrive 1000: 15 hours MACCHIATObin DoubleShot: 18 hours (in round figures). The poudriere configurations for both allowed 4 jobs at once, each job allowed to use 4 processes per job. (So much of the time the load average was more than 4.) For both there was enough RAM that the builds avoided swapping/paging, if I remember right. Sounds like a LX2K would be an interesting thing to compare with. === Mark Millard marklmi at yahoo.com ( dsl-only.net went away in early 2018-Mar)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?C088F416-80FA-45E3-A61E-8E7EB61B94ED>