Eli Marmor wrote:
> Diego Iastrubni wrote:
> > In precompiled kde 2.2.2 from mdk8.1 (i586) and self-compiled 3.1
> > (i686) I did not find any speed improove and in startup even a 5
> > seconds less:
> > load -> kde2.2.2 10 sec, kde3.1 15sec
> > konqui both about 2 secs.
> Do you mean that no progress has been achieved by Intel/AMD in the
> last 15 years, except for more Hertz's?
Assuming that gcc fully utilizes the i686, this may mean that the amount
of "useful work" per INSTRUCTION hasn't increased since the i386, rather
than the amount of work per MHz. And this is not far from the truth,
since the instruction set is almost identical, apart from the SIMD
extensions and some ring 0 stuff. If you're not a kernel, a multimedia
app or (thanks to Eli Biham) a block cipher, you won't see much difference.
The number of instructions executed per clock cycle has definitely
increased, but this test is oblivious to the fact.
> If the latest processors can't perform PER HERTZ more than a 15 years
> old processor (386), then what have they done with the billions they
> burned for R&D?
Even if this were the case (and it's not), it would be far from trivial.
First, there's the hardware R&D. Making smaller and faster transistors
is quite hard -- fab costs are measured in billions nowadays. Second, as
everything shrinks and clock rate increases, the design parameters
change (e.g., wire delays become a major factor). The old designs no
longer work and you have to use a lot of smarts (and transistors!) to
compensate for the new limitations.
Regards,
Eran
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
