Date: Wed, 5 Feb 2014 11:35:39 -0800 From: Adrian Chadd <adrian@freebsd.org> To: Kevin Bowling <kevin.bowling@kev009.com> Cc: "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org> Subject: Re: opteron a1100 arm Message-ID: <CAJ-VmomU7gR=TAUOdu%2BZE1aDDmLtdnWRYL1DUG2KPh4T6-zU3Q@mail.gmail.com> In-Reply-To: <lctus7$1a0$1@ger.gmane.org> References: <alpine.BSF.2.00.1401311911120.2427@wojtek.tensor.gdynia.pl> <1391538649.19169.79261269.3C5F49D1@webmail.messagingengine.com> <CAFU734xXWyc_TqBJ7e4MhD2nB01BAejR_1vT9%2B_5Ar5mJncncA@mail.gmail.com> <493DEB39-C4B4-409E-B8B2-B1B11E013754@netgate.com> <60555.1391549390@critter.freebsd.dk> <23B18B88-D888-46B3-99F6-905F86E20FAF@netgate.com> <lcsqov$ui9$1@ger.gmane.org> <DCE7B22A-D51F-471A-B446-4B76D87E775D@netgate.com> <lctus7$1a0$1@ger.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 5 February 2014 10:11, Kevin Bowling <kevin.bowling@kev009.com> wrote: > On 2/5/2014 6:44 AM, Jim Thompson wrote: >> >> >> >>> On Feb 5, 2014, at 1:55, Kevin Bowling <kevin.bowling@kev009.com> wrote: >>> >> [...] >> >>> Where A1100 wins hands down is memory capacity, and possibly even memory >>> bandwidth (DDR4 mentioned in the PR..). >>> >>> Intel's server Atom chips and even the extremely powerful Xeon E3 are >>> quite limited by 32GB RAM (and it's also somewhat expensive vs RDIMMs). >> >> >> I'm not sure what your point is. > > > Simply that this is a better chip than Intel's current low end offerings > because all relevant bandwidths are much higher (RAM, I/O, net) and these > matter more to many DC workloads. > > >> The C2750, C2550, C2558 & C2758 will all address 64 GB of DDR3 RAM. >> >> You'll have to use Registered DIMMs and likely drop back to DDR3 to get to >> 128GB on the A1100. >> >> So I see parity in terms of addressable RAM. > > > That's half, and a big deal if you want to run multiple VMs or a large ARC > or many large JVM heaps or any number of other things. You're trying to > claim the chip is already matched by the current Atom, where it is clearly > not, so I'll fire it back at you - I'm not sure what your point is. > > I'm happy to see the announcement and am looking forward to running FreeBSD > on the platform with others. But that's not enough - some one / some company has to step up and do a port, then run with it long enough in -HEAD to make it stable for general use. Almost all of the ARM stuff in FreeBSD doesn't fall under this category. It's getting better, but in a lot of instances the ports are done for very specific chipsets and/or use cases and that doesn't stress the rest of the system out enough. The MIPS port went through this too, FWIW - it initially was stable if you never bothered doing much in userland, but userland suffered from a lot of random hilarity. so if you were a router that occasionally ran ntpd, it was great. If you were an AP that kept going into hostapd to do things, it would get upset. "make all install clean" in a port directory wasn't great. It was only through repeated head-slamming from Warner, Juli and others that this stuff got ironed out enough. So, I'm all for the platform - who's going to port it? Step up please and be cheered. :-P -a
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmomU7gR=TAUOdu%2BZE1aDDmLtdnWRYL1DUG2KPh4T6-zU3Q>