Date: Tue, 2 Feb 2021 21:23:03 -0800 From: Mark Millard <marklmi@yahoo.com> To: freebsd-ppc <freebsd-ppc@freebsd.org> Subject: Re: Heat-death of the last of the 2-socket/2-cores-each PowerMac G5s that I have access to Message-ID: <330EC96A-6234-4558-A409-3B1BE7768B69@yahoo.com> In-Reply-To: <B38F1AF4-B63F-440C-A07A-0B81D25112EA@yahoo.com> References: <27E8A3B1-7278-464E-A284-0B294337B3DA@yahoo.com> <B38F1AF4-B63F-440C-A07A-0B81D25112EA@yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2021-Jan-11, at 20:38, Mark Millard <marklmi at yahoo.com> wrote: > On 2021-Jan-11, at 17:42, Mark Millard <marklmi at yahoo.com> wrote: > >> The last of the 4-core PowerMac G5s that I have access to now shuts >> down for "CPU B0 DIODE TEMP" that "exceeds critical temperature >> (90.0 C)" when I try to rebuild/update ports or such. The other >> 4-core G5 failed for such reasons in similar contexts a few months >> ago.,Interestingly, the two G5s have very different liquid cooling >> systems despite the similar time frame for the failures. >> >> Without the faster G5s, I may just use cross-built world/kernel >> material and see if there is a tolerable but minimal set of ports >> for supporting boot testing/dump inspection and basic operation of >> the slower 2-socket/1-core-each and 1-socket/1-core-each PowerMacs >> that I have access to, avoid things like building devel/llvm* >> ports that take so long. (I have fairly strong time preferences.) > > I've done some more testing and, while use as a (full load/speed) > builder machine is a no-go, it looks like this 4-core G5 can > still be used for boot testing and basic operation without > overheating. The prior failing machine overheated more easily > but might have a similar status if I test it just for such use. > > How long the recently failed G5 will be useful for boot and basic > operation testing, I do not know. But probably longer than for > the originally-failing G5. Looks like the problem has progressed quickly, so booting without instead overheating is now unlikely. It does not appear that I'll be able to provide any testing of 2-socket/2-core-each G5 contexts any more. As for the 2-socket/1-core-each G5, care to guess which goes with which machine, G5 vs. Rock64 (Cortex-A53, not RockPro64), allowing as many cpus to be used for the job as the executing machine has (2 vs. 4): [00:15:31] [01] [00:10:57] Finished ports-mgmt/pkg | pkg-1.15.10: Success vs.: [00:12:27] [01] [00:10:35] Finished ports-mgmt/pkg | pkg-1.15.10: Success Yep, the Rock64 configuration that I use takes about the same time to build pkg as the G5 does, although the Rock64 is a little faster than the G5 for that activity. (pkg builds by itself, so there is no competing job in the above.) The RPi4B configuration (Cortex-A72), MACCHIATObin Double Shot configuration (Cortex-A72), and the OverDrive 1000 configuration (Cortex-A57) that I use are all faster than the 2-socket/1-core-each G5 for doing self-hosted, parallel builds. All 3 are faster than the Rock64 for such activity. The OverDrive 1000 is the fastest of these machines at doing parallel builds, apparently largely due to RAM caching differences and other memory subsystem distinctions. (The cpu clock rate is slower than the A72 configurations are using.) (For doing aarch64 and armv7 port builds, I generally build on the OverDrive and the MACCHIATObin. Of the configurations reported on above, the MACCHIATObin one is the 2nd fastest for parallel builds.) === 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?330EC96A-6234-4558-A409-3B1BE7768B69>