|
|
|
|
|
|||||||
| System Hardware Internal storage, motherboards, PCI cards, stuff inside the case. |
 |
|
|
Thread Tools | Display Modes |
|
#26
October 22nd, 2012, 05:37
|
|||
|
|||
I have an 8 core dual socket AMD machine. I noticed CURRENT was having issues saturating it under load until a week or so ago, not sure what that was about. I do a buildworld -j(NTHREADS * 2 + 1) and it executes very quickly, limited primarily by disk i/o. ZFS rocks on this thing, as does mysql/postgres or anything else that is happy with threads.
I do force multithreaded make jobs for ports, with 0 problems so far (not since mid-2011). I usually only see one thread getting hammered with older software that isn't thread aware. Usually load is split well. The fact of the matter is that the future lies in more cpus with deeper c-states. It allows better power savings based on demand than two to four all-powerful cores. Even with deep c-states, the resolution for demand vs power saving is very low. I think the change will be very rapid as NUMA issues are resolved as well as software that is thread-averse gets fixed. There will always be software that benefits from one fast core more than many slower ones, but that pool of software is shrinking. But like I said, I'm just as happy with my dual i7 (4 threads). It kills single threaded stuff, and most stuff is HT aware (FreeBSD is) so it doesn't bog down two fake cores on one die before giving jobs to the other die. But it doesn't have ECC, which is a downer. Personally for intel I really like the i3's that can run ECC in server boards. They're plenty fast for a file server or web server, and they run a consistent temperature as they don't have a "turbo" (AKA lava) mode. 
|
|
#27
October 24th, 2012, 21:43
|
|||
|
|||
|
The real benefit in running Intel boards, in my experience, is that the BIOS and chipsets follow fairly rigorous rules that FreeBSD kernel/Xorg/etc developers usually only have to code for once. The chipsets for AMD chips have been varied and often have rather odd behaviors that may have to be fixed on a baord-by-board (or at least chipset-by-chipset) basis. Better chance of weird errors, heat issues, and AMD's performance is subpar at present. I'd like to see more parallel AMD systems and run FreeBSD on them, but they're just not here yet.
One note on the Intel G9xx series chipsets - there are some hardware acceleration support issues in 8.X - I'm dealing with them now on a Dell Studio 1745 with the notebook version of the G945 and acceleration is apparently not supported until 9.X. figuring on upgrading very soon to use that 
__________________
It doesn't get happy. It doesn't get sad. It just. Runs. Programs.
|
|
#28
January 26th, 2013, 21:02
|
|||
|
|||
|
I use a biostar g41d3c board. Its not the greatest nor the worst board. It gets the job done along with my 6870 graphics card allows me the play the latest high end games.
|
|
#29
January 27th, 2013, 09:41
|
||||
|
||||
|
Any opinions on the new 8 core amd fx?
|
|
| Thread Tools | |
| Display Modes | |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| The "ultimate" FreeBSD laptop (in 2012) | yha | Mobile Computing | 14 | September 14th, 2012 20:53 |
| CVE-2012-0217 Intel's sysret Kernel Privilege Escalation and FreeBSD 6.2/6.3 | bcrisp | FreeBSD Development | 0 | July 12th, 2012 17:17 |
| Be careful when upgrading FreeBSD 9.0 06-03-2012 | Adgerard | Installation and Maintenance of FreeBSD Ports or Packages | 7 | June 7th, 2012 13:26 |
| [Solved] Motherboards with ECC memory support? | mefizto | System Hardware | 6 | February 9th, 2011 19:35 |
| [Solved] Monitoring IT8718F and IT8720F LPC-IO chipped motherboards | trev | System Hardware | 3 | December 25th, 2009 07:09 |
Linear Mode
