IA64 and emulator performance
|
From: Greg Lindahl on 24 Oct 2006 01:10 In article <eu7t04-k51.ln1(a)vimes.paysan.nom>, Bernd Paysan <bernd.paysan(a)gmx.de> wrote: >Yes, reminds me on a report from Leibniz Rechenzentrum (here in Munich) >which got an Itanic-based supercomputer. They are still waiting for the >processor upgrade, while the price/perfomance ratio of their machine is >about four times worse than of a comparable Opteron-based system Well, let's be fair: LRZ claimed that they considered shared memory to be a huge plus, and so there is no comparable Opteron-based system. Then again, most centers seem to somehow muddle through with MPI codes, and few widely-used codes don't have an MPI version, so it's really hard to evaluate whether shared memory matters.(*) -- greg (*) to anyone besides NASA Ames, whose Overflow code has a huge load imbalance, but then again, did they really try to create a pure MPI version of that code?
From: John Dallman on 24 Oct 2006 15:48 In article <453d4770$1(a)darkstar>, eugene(a)cse.ucsc.edu (Eugene Miya) wrote: > >* People who have taken on HP's line that it's the natural successor > > to PA-RISC, and feel that it will be better for them than switching > > vendor. > Do you mean object-code compatibility? No, and neither do HP - although the emulator doesn't work too badly. HP are selling Itanium as the natural upgrade from PA-RISC, sometimes in odd ways. I was startled to discover one can buy a Superdome with some processors Itaniums and some PA-RISC. --- John Dallman jgd(a)cix.co.uk "Any sufficiently advanced technology is indistinguishable from a well-rigged demo"
From: Gavin Scott on 24 Oct 2006 19:05 John Dallman <jgd(a)cix.co.uk> wrote: > In article <453d4770$1(a)darkstar>, eugene(a)cse.ucsc.edu (Eugene Miya) wrote: > > >* People who have taken on HP's line that it's the natural successor > > > to PA-RISC, and feel that it will be better for them than switching > > > vendor. > > Do you mean object-code compatibility? > No, and neither do HP - although the emulator doesn't work too badly. Though it's not great either. Generally we have found that customers find the performance of PA-RISC code running under HP-UX on IPF to be unacceptable if they care at all about its performance in the first place. G.
From: Jan Vorbrüggen on 25 Oct 2006 04:46 [about price/performace being significantly better] > The keyword is "significantly". > Transputers would be around more were that true. The transputer had such significant other problems so that it never got to the point where one could actually make the test - unfortunately. The next processor (21364) that would have fit was similarly murdered for quite distinct reasons. Jan
From: Nick Maclaren on 25 Oct 2006 05:09
In article <4q8mieFlt6ipU1(a)individual.net>, =?ISO-8859-1?Q?Jan_Vorbr=FCggen?= <jvorbrueggen(a)not-mediasec.de> writes: |> [about price/performace being significantly better] |> > The keyword is "significantly". |> > Transputers would be around more were that true. |> |> The transputer had such significant other problems so that it never got to the |> point where one could actually make the test - unfortunately. The next |> processor (21364) that would have fit was similarly murdered for quite |> distinct reasons. In both cases, the problems were far more political than technical, though there were technical issues. But, if the political will is lacking, even soluble technical problems don't get solved - or, at least not properly and on time. Regards, Nick Maclaren. |