Re: [Beowulf] IB DDR: mvapich2 vs mvapich performance

2008-04-23 Thread Eric Thibodeau
Mikhail Kuzminsky wrote: In message from Greg Lindahl <[EMAIL PROTECTED]> (Wed, 23 Apr 2008 00:36:44 -0700): On Wed, Apr 23, 2008 at 07:04:51AM +0400, Mikhail Kuzminsky wrote: Is this throughput difference the result of MPI-2 vs MPI implementation or should I beleive that this difference (abou

Re: [Beowulf] MPICH vs. OpenMPI

2008-04-23 Thread Greg Lindahl
On Wed, Apr 23, 2008 at 08:37:06PM +0200, Jan Heichler wrote: > - support for a lot of Interconnects beside TCP/IP based communication mpich is used as the basis of almost all MPIs distributed by interconnect vendors. OpenMPI also supports most of these interconnects. Having 2 implementations can

Re: [Beowulf] IB DDR: mvapich2 vs mvapich performance

2008-04-23 Thread Mikhail Kuzminsky
In message from Greg Lindahl <[EMAIL PROTECTED]> (Wed, 23 Apr 2008 00:36:44 -0700): On Wed, Apr 23, 2008 at 07:04:51AM +0400, Mikhail Kuzminsky wrote: Is this throughput difference the result of MPI-2 vs MPI implementation or should I beleive that this difference (about 4% for my mvapich vs m

Re[2]: [Beowulf] MPICH vs. OpenMPI

2008-04-23 Thread Jan Heichler
Hallo Mark, Mittwoch, 23. April 2008, meintest Du: >> Can anyone give me a quick comparison of OpenMPI vs. MPICH? I've always >> used MPICH (I never liked lam), and just recently heard about OpenMPI. >> Anyone here using it? MH> openmpi is a genetic descendent of lam (with significant other cont

Re: [Beowulf] MPICH vs. OpenMPI

2008-04-23 Thread Mark Hahn
Can anyone give me a quick comparison of OpenMPI vs. MPICH? I've always used MPICH (I never liked lam), and just recently heard about OpenMPI. Anyone here using it? openmpi is a genetic descendent of lam (with significant other contributors, of course). I have not seen any good comparison which

Re: [Beowulf] Thermals question

2008-04-23 Thread Mark Hahn
This link has a lot of good information and might answer your question: http://www.mediarecovery.com/hot_topics/cooling_problems.php as long as you're alert to the fact that it's an APC marketing document. for instance: It is unlikely that a traditional data center, with computer room

Re: [Beowulf] Re: Comparison of high performance interconnects

2008-04-23 Thread Eric Thibodeau
Then the answer is 42. andrew holway wrote: I neglected to add, across a range of different codes. On Tue, Apr 22, 2008 at 5:35 PM, andrew holway <[EMAIL PROTECTED]> wrote: Hi, Has anyone got a benchmark comparison of IB vs 1Ge vs Myri or so? Ta Andy ___

[Beowulf] MPICH vs. OpenMPI

2008-04-23 Thread Prentice Bisbal
Can anyone give me a quick comparison of OpenMPI vs. MPICH? I've always used MPICH (I never liked lam), and just recently heard about OpenMPI. Anyone here using it? -- Prentice Bisbal Linux Software Support Specialist/System Administrator School of Natural Sciences Institute for Advanced Study Pr

Re: [Beowulf] Thermals question

2008-04-23 Thread Prentice Bisbal
Gerry Creager wrote: > We're building up a new high thruput cluster. Anticipate delivery with > nice shiney new racks. We're anticipating a hot-aisle/cold-aisle > concept. We also hope to go toward chilled-water cool-doors eventually, > and as we expand the installation. The racks we're getting

[Beowulf] May NYCA-HUG

2008-04-23 Thread Douglas Eadline
If you live or work in the New York City North Jersey metropolitan area, the May High-Performance User Group (HUG) has invited SiCortex to demonstrate their desk top/side 72 processor Catapult system. If you are interested please stop by. There is also plenty of cluster discussion as well. Details

[Beowulf] HP ProLiant DL165G5

2008-04-23 Thread Eugen Leitl
Anyone had HP ProLiant DL165G5 2.1GHz quad-core in their fingers yet? Either the SAS or the SATA variant? Any caveats? Is the NC7782 NIC any good, or would you recommend buying Intels? -- Eugen* Leitl http://leitl.org";>leitl http://leitl.org

RE: [Beowulf] Re: Comparison of high performance interconnects

2008-04-23 Thread Dan.Kidger
Andrew, You also need to state what *size* cluster you are targeting. Some interconnects can provide hardware broadcast and do adaptive routing - features which only start to show significant application speedup at larger job sizes. caveat: slightly biased view - I work for a vendor Daniel

Re: [Beowulf] IB DDR: mvapich2 vs mvapich performance

2008-04-23 Thread Greg Lindahl
On Wed, Apr 23, 2008 at 07:04:51AM +0400, Mikhail Kuzminsky wrote: > Is this throughput difference the result of MPI-2 vs MPI > implementation or should I beleive that this difference (about 4% for > my mvapich vs mvapich2 at SC'07 ) is not significant - in the sense > that it is simple becaus