Recent AMD760MPX chipset PCI measurements??
paullu at cs.ualberta.ca
Tue Feb 12 16:57:06 PST 2002
On Fri, Feb 08, 2002 at 09:46:09AM +0100, Joachim Worringen wrote:
> Paul Lu wrote:
> > DMA rate for 4096 Byte transfers (64bit / 66MHz bus)
> > 1st: 8 pages from bogus sdma pg, 8 to bogus rdma
> > bus_read (send) = 315 MBytes/s
> > bus_write (recv) = 341 MBytes/s
> > which is a bit disappointing. I guess the numbers are not outright
> > horrible, but they are not nearly as good as the Serverworks-based boards
> > we have now (yes, this may be an apples-oranges thing :-)
> Depending on what you intend to use your PCI-bus for (I assume for the
> interconnect?), this bandwidth may well be enough. I.e., Myrinet has 2Gb
> links, and thus the bottleneck when using a Myrinet board in this bus is
> the network link, not the PCI bus.
Agreed. Greg Lindahl also pointed this out in a separate reply.
I was hoping for greater bandwidth to handle the case of a full duplex
transfer across the Myrinet (which could, in theory, saturate the PCI bus
given the full 4 Gb/s bandwidth of duplex operation).
In an email to me, Greg made the point that 4 Gb/s duplex
is difficult to achieve in practice with the current drivers.
I haven't tried it, so I defer to Greg (and you) on this issue,
but I was hoping to have some extra headroom (for duplex transfers
*and* in scenarios where there is also lots of, say, disk I/O on
the PCI bus too).
Thank you and Greg for the responses.
More information about the Beowulf