Date: Sun, 01 Dec 2019 22:05:37 +0000 From: greg@unrelenting.technology To: "John F Carr" <jfc@mit.edu>, freebsd-arm@freebsd.org Subject: Re: 64 bit ARM systems with more than four cores Message-ID: <e839d38b18d4ba1ff0354ebc6ba8bc9c@unrelenting.technology> In-Reply-To: <94BC4785-B16E-43B2-B222-3D3C8C5DEB50@exchange.mit.edu> References: <94BC4785-B16E-43B2-B222-3D3C8C5DEB50@exchange.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
December 2, 2019 12:37 AM, "John F Carr" <jfc@mit.edu> wrote:=0A=0A> Two = RK3399 based systems, 2 A-72 + 4 A-53: RockPro64 (http://www.pine64.com) = and ROCK Pi 4=0A> (http://radxa.com). I can't use the RockPro64 due to th= e 1.5 MBps serial port. Apparently it works=0A> or almost works with some= special handling if your serial interface can count to 1,500,000. The=0A= > ROCK Pi 4 is in crochet so it must work... right? Does it have a reason= able console port bit rate?=0A=0ABefore I got my 1.5M capable dongle for = the ROCKPro64, I just had an u-boot command to=0Aswitch to 115200 in the = boot script.=0A=0AI consider 2xA72+4xA53 worse than 4xA72 to be honest.= =0AEspecially since the FreeBSD scheduler is not big.LITTLE aware =E2=80= =94 it loves to schedule=0Ae.g. a big single-threaded linker job onto the= slow cores. So you have to mess with=0Acpuset but apply it carefully to = still utilize all cores on parallel tasks.. argh.=0A=0A> One LX2160A base= d system, 16 A-72 cores: HoneyComb LX2K=0A> (https://www.solid-run.com/nx= p-lx2160a-family/honeycomb-workstation). I see relevant files in=0A> sys/= gnu/dts/arm64/freescale but nothing outside of these files imported from = Linux. It's not=0A> mentioned on https://wiki.freebsd.org/arm64. My guess= is that means FreeBSD does not run. Is it a=0A> little job or a big one?= =0A=0AThere's SBSA-ish ACPI-capable firmware so it should boot, but proba= bly with uhh not a lot of devices.=0AThey've been sort of vaguely promisi= ng that PCIe would work in ACPI..=0Aand recently this commit dropped:=0Ah= ttps://source.codeaurora.org/external/qoriq/qoriq-components/edk2-platfor= ms/commit/?h=3DLX2-UEFI-ACPI-0.2&id=3Dec8deb524171a03448395e6b8acc49978bc= f5ce5=0Aso maybe I shouldn't have doubted them, but I missed the $500 dis= count period waiting for this :(=0A=0A> SC2A11, 24 A-53 cores at 1 GHz: S= ynQuacer (https://www.96boards.org/product/developerbox). I don't=0A> see= any evidence of SC2A11 support in the kernel tree or on https://wiki.fre= ebsd.org/arm64. My=0A> guess is that means FreeBSD does not run.=0A=0AThe= re's SBSA-ish ACPI-capable firmware so it should boot, and PCIe should wo= rk.=0AI guess nobody tried because it's very expensive for something with= A53 cores.=0A=0AI would *not* trade my MACCHIATObin for this because A53= is just so slow.=0A=0A> ThunderX in various forms, rack mount or worksta= tion. Nice specs and apparently supported but the=0A> two American system= builders don't seem interested in selling me one.=0A=0AKeep in mind that= the TX1 cores are slow. It's great at super parallel workloads,=0Abut wh= en you have something single-threaded, it's no better than an A53.=0A=0A= =0A=0ANow, if you have cash but not ThunderX2 levels of cash: Ampere eMAG= is the best option.=0AAvailable in Lenovo HR350A and HR330A boxes, also = from Avantek (idk what's the mainboard there??),=0Aalso rentable in bare-= metal-cloud at Packet (they have the Lenovos).=0A=0ASee https://bugs.free= bsd.org/bugzilla/show_bug.cgi?id=3D237055 for the bringup story :)=0AI th= ink most of the bugfixes required for eMAG are in -CURRENT.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?e839d38b18d4ba1ff0354ebc6ba8bc9c>