Re: I Should Surrender to Opteron


"anonymous" <no@xxxxxxxxx> wrote in message
news:pan.2006.03.22.17.21.22.956764@xxxxxxxxxxxx
On Mon, 20 Mar 2006 15:31:59 +0100, Joachim Worringen wrote:

John Savard wrote:
This is informative. So the X-1 CPU is ahead of the SX-6 CPU. And I
do remember an article about the SX-6 chip which noted that it had no
cache, but instead a very high-speed path to external memory.

Vector architectures do not provide CPU-cache on purpose - the
coherence
protocol would kill the vector performance. Actually, you just don't
need the
type of cache that scalar CPUs rely on to get some performance.
However, the
SX-8 (the current NEC SX generation at 16Gflops) has a "bank-cache" to
reduce
the potential performance impact of bank conflicts (multiple CPUs
accessing the
same memory banks at the same time - vector machines have thousands of
memory
banks).

Which recent machines have thousands of banks? None of the ones
in the top 500 supercomputers list do, as far as I can tell!
--
Andronico

In some sense the many clustered machines have thousands of banks. They
just happen to be attached to thousands of processors, eg Blue Gene.

del



.

Relevant Pages

  • Re: 54 Processors?
    ... > My memory is cloudy but I seem to recall these statements around the ... a big problem was strong memory consistency model and cache ... were saturating their machines ... ... the common L2 cache interfaced to the SCI memory access port. ...
    (bit.listserv.ibm-main)
  • Re: Calculation Errors - Machine Dependent?
    ... Do the different machines have a single CPU or multiple? ... before cache has been flushed to memory (this can be a cache controller ...
    (comp.lang.cobol)
  • Re: I Should Surrender to Opteron
    ... do remember an article about the SX-6 chip which noted that it had no ... but instead a very high-speed path to external memory. ... type of cache that scalar CPUs rely on to get some performance. ... same memory banks at the same time - vector machines have thousands of memory ...
    (comp.arch)
  • Re: volatile and win32 multithreading
    ... Volatile makes the compiler produce code that reads from memory when ... the value is read and writes to memory when the value is written. ... What goes into the cache doesn't normally get written ... we can expect nearly all machines to start to have more and more ...
    (microsoft.public.vc.language)
  • Re: I Should Surrender to Opteron
    ... but instead a very high-speed path to external memory. ... type of cache that scalar CPUs rely on to get some performance. ... same memory banks at the same time - vector machines have thousands of memory ...
    (comp.arch)