AMD Bulldozer Disappoints

Looking for new hardware to run WRF? Intel or AMD? Check this forum.
cmaiers
Posts: 27
Joined: Wed Dec 08, 2010 12:16 pm
Location: Iowa

AMD Bulldozer Disappoints

Post by cmaiers » Thu Oct 13, 2011 11:35 am

So apparently 8 cores in AMD world are less powerful than Intel's 4.

http://www.pcworld.com/article/241812/l ... power.html

Zoyx
Posts: 14
Joined: Fri Aug 21, 2009 7:21 pm

Re: AMD Bulldozer Disappoints

Post by Zoyx » Sat Jan 07, 2012 4:42 am

If you compile specifically for the bulldozer architecture, you get better benchmarks. I will have an FX-8120 machine built shortly, I'll see how it goes.

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: AMD Bulldozer Disappoints

Post by pattim » Mon Jun 25, 2012 3:36 am

Zoyx wrote:If you compile specifically for the bulldozer architecture, you get better benchmarks. I will have an FX-8120 machine built shortly, I'll see how it goes.
Don't we have to ask the EMS maintainer(s) to build for that architecture? (When I installed, it only downloaded prebuilt binaries.)

Zoyx
Posts: 14
Joined: Fri Aug 21, 2009 7:21 pm

Re: AMD Bulldozer Disappoints

Post by Zoyx » Mon Jun 25, 2012 3:41 am

You can do your own compiling of components, then insert it into the package.

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: AMD Bulldozer Disappoints

Post by pattim » Mon Jun 25, 2012 6:02 pm

Zoyx wrote:You can do your own compiling of components, then insert it into the package.
Thank you. I was thinking about doing that... but optimizing for Opterons takes just a tad more know-how than I have, although *maybe* the PGI compilers have optimization switches for Magny-Cours and Bulldozer? If so, they could be built as default binary options (to replace some of the older options for CPUs that are disappearing)?

meteoadriatic
Posts: 1603
Joined: Wed Aug 19, 2009 10:05 am

Re: AMD Bulldozer Disappoints

Post by meteoadriatic » Mon Jun 25, 2012 6:13 pm

pattim wrote:*maybe* the PGI compilers have optimization switches for Magny-Cours and Bulldozer? If so, they could be built as default binary options (to replace some of the older options for CPUs that are disappearing)?
There are optimization routines for AMD in PGI. I read something about that in PGI user manual. However, neither I have PGI licence, neither any AMD computer, so I just have zero exerience with that.

WRFEMS is compiled with PGI. With what optimization flags... no idea. If you're so interested in this, maybe EMS's autor Robert Rozumalski can tell you more about that. You can reach him by e-mail (that can be found on official WRFEMS web page).

Zoyx
Posts: 14
Joined: Fri Aug 21, 2009 7:21 pm

Re: AMD Bulldozer Disappoints

Post by Zoyx » Mon Jun 25, 2012 6:53 pm

It would be cool if someone would give the Open64 compiler suite a try with the latest Opterons. That is bit beyond my level of skill, unfortunately.

http://developer.amd.com/tools/open64/P ... fault.aspx

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: AMD Bulldozer Disappoints

Post by pattim » Mon Jun 25, 2012 7:13 pm

meteoadriatic wrote:
pattim wrote:*maybe* the PGI compilers have optimization switches for Magny-Cours and Bulldozer? If so, they could be built as default binary options (to replace some of the older options for CPUs that are disappearing)?
There are optimization routines for AMD in PGI. I read something about that in PGI user manual. However, neither I have PGI licence, neither any AMD computer, so I just have zero exerience with that.

WRFEMS is compiled with PGI. With what optimization flags... no idea. If you're so interested in this, maybe EMS's autor Robert Rozumalski can tell you more about that. You can reach him by e-mail (that can be found on official WRFEMS web page).
I was hesitant to contact Robert because I know he's terribly busy - so I was going to run it by folks here and see first if there was wisdom to be had... :)

pattim
Posts: 199
Joined: Sun Jun 24, 2012 8:42 pm
Location: Los Angeles, CA, USA

Re: AMD Bulldozer Disappoints

Post by pattim » Mon Jun 25, 2012 7:14 pm

Zoyx wrote:It would be cool if someone would give the Open64 compiler suite a try with the latest Opterons. That is bit beyond my level of skill, unfortunately.

http://developer.amd.com/tools/open64/P ... fault.aspx
Thanks - I didn't even know about that compiler! But I didn't see anything about NUMA or Opterons...

Benchmarks:
http://www.phoronix.com/scan.php?page=a ... en64&num=1

Zoyx
Posts: 14
Joined: Fri Aug 21, 2009 7:21 pm

Re: AMD Bulldozer Disappoints

Post by Zoyx » Mon Jun 25, 2012 7:33 pm

I exchanged emails with Bob about the Open64 compilers back when the Bulldozers first came out, here are the tips he passed along...
I took a look at the site and it appears you could use the AMD compilers; however, I have a few comments:

1. The EMS 64-bit binaries are fully optimized for AMD and INTEL processors. That's not to say that
you can not squeeze out some additional performance using AMD's compiler but it will require
quite a bit of time on your part.

2. The AMD compiler system is relatively new so expect to have some setbacks along the way.

3. Binaries built from the off-the-shelf WRF release will not work correctly with the EMS. You will
need to replace a number of routines in WPS and WRF with those available in the
wrfems/util/wrfems directory. After that you will have the same version that is provided with the
EMS.


4. You will also have to build netCDF and MPICH2 with the same AMD compilers. You may also need
to build libz, libjasper and libpng as well although I think the RPM versions will work. I compile
everything from scratch with the same PGI compilers to be safe.

Post Reply