Re: Should tests run on a two CPU system?
- From: Logan Shaw <lshaw-usenet@xxxxxxxxxxxxx>
- Date: Tue, 04 Dec 2007 21:14:53 -0600
On Dec 4, 2:47 am, David Schwartz <dav...@xxxxxxxxxxxxx> wrote:
If you follow school of though 1, then yes. If 2, then quad-core would
likely be best. Of course, some bugs pop up best with a single CPU
(especially certain types of livelock where another CPU can come along
and 'save' you).
Okay so i will try to get a quadcore as my next system and some of
larger SUN Servers with more then 4 cpus. I've seen a E3800 with 10
(slow 400Mhz each) for 200 Euro on EBay the last time. Slow but maybe
the ultimate strange system for testing.
I bet you spend more than 200 Euro on efforts to get extra heat out of
your machine room after you install that bad boy. :-)
Speaking of ultimate strange systems, I believe you can get the old
MBus-based SPARC systems to run with MBus processor boards of different
speeds. So go get yourself an old SPARCcenter 2000 with 10 MBus slots
and go to town with possibly 40 Mhz, 45 MHz, 50 MHz, 60 MHz, 75 MHz,
and 85 MHz CPUs, some of which are SuperSPARC-I type and some of which
are SuperSPARC-II, all running together on the same system. Since the
operating system doesn't (AFAIK) know to prefer faster CPUs, you may
get some charming scheduling effects. But good luck getting just the
right configuration to get the system to run. (It's not exactly