64bit/Dual Core Issues and Recommendations?

[Gerry Creager N5JXS]

Bill, et al,

Bill Binko wrote:
>>Bill,
>>
>>I don't have direct experience unfortunately, but I have heard a
>>few complaints about 64bit AMD support as well for GDAL.  If I
>>were you and looking for a "safe" bet I would stick with a 32bit
>>solution.

With only a little learning curve, 64-bit Opteron isn't too bad.

>>If anyone out there has an Opteron system running linux that
>>they would like to provide me with a testing account on, I would
>>be interested in building and testing some components (ie
>>GDAL, MapServer).

Didn't see who posted this (things have been hectic lately) but I can 
likely provide an opteron system account.  Contact me offlist to 
accomplish this.

> I'd like to thank everyone for the feedback on this.  I must say it's
> mixed :-(

I'd not be too mixed.  I've been pleased enough with the Opteron to have 
essentially migrated my HPC cluster in that direction (having said that, 
we just ordered 16 nodes of Pentium D to test the architecture!) as well 
as my graphics systems, fileservers, and data acquisition systems.

> I am in an interesting situation, however: while my hardware needs will
> (hopefuly) grow to be considerable, my initial risk (in terms of real $$)
> is fairly low.  Even an occasional down time would be acceptable until
> about January, when I expect to ramp up usage.
> 
> Therefore, I'm considering the following strategy.  I think it will
> provide the highest potential reward, while leaving me a fall-back
> position without looking too much investment.
> 
> I plan on going ahead with a dual 64-bit Opteron system.  I haven't
> decided on dual-core or not.  I also have considered starting with a
> single dual-core CPU and adding another near the end of my test cycle.
> That might save me cash if I decide I have to scrap this and go to a more
> conventional system -- I am leaning toward just getting two CPUs up front.

The conservative approach would be to start with 2 single-core chips and 
migrate to dual core, but I don't think your approach will be too bad, 
either!

> From everything I have heard, I can run 32-bit Linux on these without any
> problems from any of the core open source GIS apps.  (Please correct me if
> I'm wrong on that impression).

Correct.  We've done that while we were working the kinks out of a 
single-core, single CPU system... Motherboard issues mostly, as our 
compute cluster was merrily running 64-bit linux the whole time on a 
different motherboard set and "just doin' it".

> I will start with a 64-bit Linux, and rebuild my core apps 64-bit.  That
> would give me from my purchase date (probably about 2 weeks out)  through
> my development (until Nov 10) and through my testing cycles (through Jan
> 1) to find and fix as many 64-bit issues as I can.  If I find it's
> unstable for my purposes, I will fall back to 32-bit Linux and 32-bit
> apps.

Learning curve for 64-bit compiles.  kinda steep but not too high.  you 
should do fine.

> Assuming I take this route, I would be happy to provide SSH access to all
> of the core developers of the systems I'm using (GDAL, Mapserver, PostGIS,
> etc.) if they would like to test on that platform (hint, hint, Frank).

Why didn't *I* think of that?  I can likely privide a system where we 
can do the same, either an Opteron or an Athelon-64, which is performing 
about the same as an Opteron in a graphics-intensive environment.

> I am making the assumption that all of the core apps are 80-90% working on
> 64-bit systems.  I'm willing to be guinea pig and bug-hunter for the
> remaining 10-20% given the situation I'm in and the potential benefit of
> running 64-bit (and much faster from the annecdotal feedback I've gotten).

Haven't done most of the GDAL/Proj-intensive stuff.  Mapserver compiles. 
  PostGIS just works.  Most of our focus was on Unidata's Gempak s/w, 
and now, IDL, for some meteorological apps.  MapServer, for us, still 
runs on a one-lung Celeron system!

> Again, thanks for the input,

Sorry I didn't respond earlier!
gerry
-- 
Gerry Creager -- gerry.creager at tamu.edu
Texas Mesonet -- AATLT, Texas A&M University	
Cell: 979.229.5301 Office: 979.458.4020 FAX: 979.847.8578
Page: 979.228.0173
Office: 903A Eller Bldg, TAMU, College Station, TX 77843