[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Opteron SPEC CPU numbers...

On Fri, Jul 11, 2003 at 10:49:28AM -0400, Kurt Keville wrote:

 > AMD has posted some Opteron SPECint and SPECfp numbers at
 > http://www.spec.org/cpu2000/results/res2003q2/

 Ooooo, good... but FP is a bit of a turn off when compared to Itanium2
 numbers (look at the data for the Altix from SGI).  I can't beleive
 that Intel seems to have done something right for a change :-)  The
 Altix has a 1 GHz Itanium2 and that gets beaten by this system's
 Opteron @ 1.8 at CINT2000, but on CFP2000 the Opteron get horribly
 humiliated with its ~ 1100 (base) against the Altix's ~ 1400 (and don't
 even look at the newer data for the Altix @ 1.5 GHz).  The one thing
 that SGI is still doing right is scalability, but that's hardly news.
 The scalability of the Opteron machines on the other hand...  ouch.
 Going from Opteron 144 @ 1.8 to 844 @ 1.8 means a 3.7x change in the
 CINT2000 rates... on Windows.  Where are the numbers for a 4-way
 Opteron on Linux?

 How does GCC fare compared to the Intel compiler here?

 While I have your attention, does anyone here know of a board _with_ an
 AGP slot?  NVIDIA has drivers for Linux x86-64 and Linux 2.5 has
 agpgart support for AMD's 8151 bridge as well as VIA's K8T400 and SiS'
 755, so I guess _someone_ has seen the mythical creature... in the
 wild?  There's been much speculation regarding the actual availability
 of it, and I'd _love_ to get my hands on a 4-way system with at least
 one AGP slot, but one with two would be heaven (the kernel supports up
 to 8 GARTs but more are just a #define away).


Reply to: