I am upgrading FSB to 800 and ddr to 400 to decrease memory bottleneck, and
thought of this....
Also ,does L2 cache to 1mb vs 512kb change it?
I found info for GFLOPS only being related to cpu hertz and no other
details.
Go here:
http://www.amdcompare.com/us-en/desktop/
Click "view all products"
You can either open up separate windows, like these:
http://www.amdcompare.com/us-en/desktop/details.aspx?opn=ADA4000DAA5BN
http://www.amdcompare.com/us-en/desktop/details.aspx?opn=ADA3800DAA4BP
or you can use the compare function, by ticking separate
boxes on the "view all products" results page, giving this:
http://www.amdcompare.com/us-en/desktop/SideBySide.aspx?opn=ADA4000DAA5BN&opn=ADA3800DAA4BP
Notice how a processor with 1MB cache rates as a 4000+,
while a processor with a 512MB cache rates as a 3800+.
Both processors have the same core clock.
That tells you, that the increased cache is the equivalent
of a 200MHz faster core. That is the typical effect of
the cache. The AMD benchmarks use some combination of
applications, but I doubt I could find a web page on
the AMD site that describes the current application mix.
If a test runs fully within the L1 cache, the L1 runs at
full speed (same speed as the core). Then, the test will
only depend on the core clock speed. A faster core gives
more operations per second.
As the size of a test extends into L2 or into main memory,
the performance will depend more and more, on the speed
of the cache and the memory itself. But few real world
applications are so pathological, that they "fly" inside
a 1MB cache, but "die" within a 512KB cache. You can
engineer a synthetic benchmark to make the 512KB cache
look bad, but the truth is, a 512KB cache is just fine
for running average type stuff on Windows.
Paul
)