64bit/Dual Core Issues and Recommendations?

[Shawn Wallbridge]

Sun just released some nice 1U Opteron boxes and they look to be  
_VERY_ competitively priced, they start at $795. They run Solaris,  
Linux or Windows.

At work we have 120 Dual Opteron 246 machines. They are using the  
Tyan S2891 board in a Chenbro RM214 case. They have been running at  
100% CPU utilization for three months solid without a single problem.  
We benchmarked the Xeon EMT64 as well and the Opteron was faster and  
ran a _LOT_ cooler (and used less power).

Of course none of this is GIS/Mapserver related, so that doesn't help  
with that part of your question.

shawn

On 14-Sep-05, at 11:51 PM, Bill Binko wrote:

> Hi everyone,
>
> I am in the market for a server, and am trying to choose a CPU  
> platform.
> I have seen recently a number of issues coming up with various GIS
> applications (qgis/GDAL/Mapserver/PostGIS) all pointing to spurious
> segfaults when running on AMD 64 platforms.  Also, I've heard  
> rumblings
> from friends that getting any of the linux distros on AMD 64 is  
> somewhat
> troublesome.
>
> I will be doing fairly intensive mapping applications, involving
> Mapserver, Mapscript, GDAL (For aerials), and lots of PostGIS dynamic
> queries against a pretty hefty database.
>
> Since I don't want to pay a Dell/HP premium, I have looked at the 1U
> barebones systems on newegg, and basically, there are three options:
>
> 1) Dual Opteron 64 - dual core is optional
> 2) Pentium (4 or D) - Hyperthreading and/or Dual Core is optional
> 3) Dual Xeon EM64T
>
> From what I can tell, these all have different sweet spots, and all  
> will
> perform high enough for my needs.  What I DON'T want is to be fighting
> fires due to problematic 64bit issues.
>
> Can I get some input on what others out there are using for their  
> server
> needs?  Does anyone have a running list of Opteron issues in the GIS
> arena?
>
> Any help would be appreciated.  I have a Nov. 10 drop-dead date to  
> have
> these up and running so I'm not exactly dying, but I'd really  
> appreciate
> the input up front.
>
> Thanks!
>
> Bill
>