Fyo, "This contradicts your claim that "the difference
between consecutive scores should not drop faster than
it started at"."
Do not play pgerassi. I did not start at 0 frequency.
I started from 1500-1600-1700-1800 point where the
d²P/df² appears to be flat. And my argument
was that the 3x-jump in second derivative can't happen
unless BIOS made compulsory configuration
changes to something.
"However, in the AthlonXP vs P4 case, there is a huge difference in available bandwidth. This significantly impacts not only performance, but also scaling. Same goes for larger caches, although this is certainly harder to compare directly, since the caches are very different in nature (latency, bandwidth and associativity - along with size)."
It all is true and relevant, but it does not lead us to
any conclusion. The bottom line is the value of memory
wall for each particular platform, see definitions
complang.tuwien.ac.at
ftp://ftp.cs.virginia.edu/pub/techreports/CS-94-48.ps.Z
For example, FSB can be slow but cache may be more
effective, and faster FSB/memory may be coupled with less
effective cache. In both cases the memory wall might
be the same, then a CPU with higher internal IPC will
win, again depending on the top frequency of the
design. So, go figure.
- Ali |
