Intel 860 PCI bandwidth problem
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Steffen Persvold sp at scali.comSat Jan 19 05:47:47 PST 2002
- Previous message: Intel 860 PCI bandwidth problem
- Next message: Intel 860 PCI bandwidth problem
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Greg Lindahl wrote: > > On Fri, Jan 18, 2002 at 10:40:37PM -0700, Maurice Hilarius wrote: > > > Implication: The early termination limits the maximum bandwidth to ~90 MB/s. > > That note sounds like it's not talking about DMA operation. You did > look at the Myrinet Experiences website > I think you misunderstood, "memory read multiple" is the PCI instruction used by most DMA engines (SCSI, ethernet, SCI and I would guess Myrinet) when they read from RAM (source machine). On the other side (destination), "memory write invalidate" is normally used. However, this errateadoesn't limit the SCI DMA bandwith to ~90MB/s either, 210MB/s is the most I've seen so far. Regards, -- Steffen Persvold | Scalable Linux Systems | Try out the world's best mailto:sp at scali.no | http://www.scali.com | performing MPI implementation: Tel: (+47) 2262 8950 | Olaf Helsets vei 6 | - ScaMPI 1.12.2 - Fax: (+47) 2262 8951 | N0621 Oslo, NORWAY | >300MBytes/s and <4uS latencyy
- Previous message: Intel 860 PCI bandwidth problem
- Next message: Intel 860 PCI bandwidth problem
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
