which one is faster ?

  • Thread starter Thread starter bobman
  • Start date Start date
bobman said:
which one is faster ?
Northwood 2.8G @400fsb or Northwood 2.4G @533fs

thanks
Click to expand...

The 2.8GHz. The core speed is the most important
parameter.

(There are exceptions. For example, there was a guy
who had a "cache buster" type application, where he
claimed performance was dominated by memory type and
speed, but not everyone is running simulation environments
like he uses. A lot of other applications do quite well
with cache, meaning that the CPU can do useful work most
of the time. And even if the memory type stinks, you'll
still get most of the extra performance from the core.)

If you want to do your own research on this, you can
try -

http://www.hwbot.org/browseResults.do

I tried a few processors, and it is actually amazing
how well increased performance tracks increased core
speed. I figured it would begin to tail off (diminishing
return), but even a processor above 7GHz did well.

Memory and FSB have less of an effect.

(Example of torturing processors in the name of science...
A P4 running at 7.6GHz, on liquid nitrogen. That is double
its normal core speed.)

http://www.xtremesystems.org/forums/showpost.php?p=1267944&postcount=161

Paul
 
The 2.8GHz. The core speed is the most important
parameter.
Click to expand...

I used to think that as well until Intel came out with
the core2duo chips a couple yrs ago.

A new 2.4 GHz Core2Duo (E6600) is fully 150%
faster than my old P4D-3.0 GHz - at least as far as
integer and real number math is concerned. Even my
1.733 GHz c2d laptop is as fast as the P4D.

The larger L2 cache allows for computing much larger
numbers before RAM is used - which slows either
chip to a crawl once the numbers get larger than L2
cache can contain.

- Ed
 
Ed said:
I used to think that as well until Intel came out with
the core2duo chips a couple yrs ago.

A new 2.4 GHz Core2Duo (E6600) is fully 150%
faster than my old P4D-3.0 GHz - at least as far as
integer and real number math is concerned. Even my
1.733 GHz c2d laptop is as fast as the P4D.

The larger L2 cache allows for computing much larger
numbers before RAM is used - which slows either
chip to a crawl once the numbers get larger than L2
cache can contain.

- Ed
Click to expand...

But the original question, was comparing two chips of
like architecture, and asking which one to select.
The 2.8Ghz is the right answer for the OP.

*******
When comparing chips of unlike architecture, the performance
metric is:

clock_rate * IPC

where IPC is instructions per clock. The latest Core2 and
AMD processors, have an IPC which is 50% (or more) higher
than the P4 processor. And then, if the OP was selecting
a brand new processor and motherboard, then yes, more
needs to be considered than just the clock rate. A benchmark
may be one way to make a selection in that case, as long as
the reader understands how to read and interpret the (skewed)
set of benchmarks provided. (Hard to find single core benchmarks
here, for determining how old software would run.)

http://www23.tomshardware.com/cpu_2007.html

But when comparing processors of the same architecture,
say an E6600 and an E6700, the E6700 has a slightly higher
core clock rate, and it would win. Seeing as the IPC of the
two processors would be identical.

A benchmark is useful, because it also takes into account
the effects of different sized caches, different FSB,
potentially different speed memories selected, and so on.
Comparing processors is very difficult to do, with a
paper analysis, which is why a benchmark is the best (imperfect)
way to do it. You need an honest benchmarker, someone
who is "interested in the science", and not trying to sell
you something, to get the best value from a benchmark. In
that sense, Tomshardware fails, because they have no interest
in actual architecture or performance comparison. The charts
exist, to sell advertising. Some of the smaller web sites,
have more of an interest in what lies underneath, but they
don't have the amount of testing resources that Tomshardware
does.

The effects of cache are quite variable. Some games benefit from
cache, and some don't. Just because a processor has a larger
cache, doesn't mean it is going to benefit the user. That
will be determined by usage pattern. To give an example, at
my old employer, we had a very expensive "computing device",
running a fixed application, an application which was a
"cache buster". In fact, we had engineers researching how
to do something other than conventional cache, to try to
fix that. The execution model of the software was "run
to completion", mostly linear with little looping, and for
that, it didn't matter if conventional cache was 128KB or 24MB.
Made no difference at all. Certain simulation environments are
like that as well, benefiting poorly from cache.

Paul
 
Back
Top