Is it a lot slower?

  • Thread starter Thread starter mike3
  • Start date Start date
M

mike3

Hi.

I'm curious. How does a single core of an Intel Core 2 Quad Q6600
processor running at 2.4GHz compare to a Pentium 4 running at 2.8GHz
(the latter is the type of processor I have now)? Does that 400MHz
drop in clock rate give a significant decrease in terms of single-
threaded performance, or does the chip design potentially quell some
of that? Ie. clock-for-clock, does the Core 2 run faster, slower, or
the same as the P4, on average?
 
mike3 said:
I'm curious. How does a single core of an Intel Core 2 Quad Q6600
processor running at 2.4GHz compare to a Pentium 4 running at 2.8GHz
(the latter is the type of processor I have now)? Does that 400MHz
drop in clock rate give a significant decrease in terms of single-
threaded performance, or does the chip design potentially quell some
of that? Ie. clock-for-clock, does the Core 2 run faster, slower, or
the same as the P4, on average?
Click to expand...

Well, I'm sure it depends upon the application, but I de-clocked my
overclocked Q6600 to stock speed, set the core affinity to a single
core, and ran SuperPi Mod 1.5.
http://www.xtremesystems.com/pi/
It completed in 21.235 seconds. At 3 GHz, it did 16.828 s.
(i965 chipset, DDR2 at 4-4-4-10)

How does the P4 2.8 do?
 
Well, I'm sure it depends upon the application, but I de-clocked my
overclocked Q6600 to stock speed, set the core affinity to a single
core, and ran SuperPi Mod 1.5.
Click to expand...

Uh, I meant to say that was the 1M test.
 
Well, I'm sure it depends upon the application, but I de-clocked my
overclocked Q6600 to stock speed, set the core affinity to a single
core, and ran SuperPi Mod 1.5.http://www.xtremesystems.com/pi/
It completed in 21.235 seconds.  At 3 GHz, it did 16.828 s.
(i965 chipset, DDR2 at 4-4-4-10)

How does the P4 2.8 do?
Click to expand...

I got a slow 66 seconds on my machine :(

Dang, this would suggest that a single core of C2Q Q6600 @ 2.4GHz
is *3* times faster than the P4!

Is this really right -- and therefore I won't notice a performance
loss
on most of my applications if I got the Q6600 (and quite likely a
significant performance _gain_) even though the clock is somewhat
slower?
 
mike3 said:
Is this really right -- and therefore I won't notice a performance loss
on most of my applications if I got the Q6600 (and quite likely a
significant performance _gain_) even though the clock is somewhat
slower?
Click to expand...

No, only in single threaded applications. In multithreaded applications,
it will kick its ass all over the place! No, a lot of times I find it is limited
by disk I/O speed, like when I am demultiplexing a .mpg file for editing.
The charts here help you compare, but I'm afraid they are mostly biased
toward multithreaded apps:
http://www23.tomshardware.com/cpu_2007.html
The 2006 chart has the quad core listed, also.
 
mike3 said:
I got a slow 66 seconds on my machine :(

Dang, this would suggest that a single core of C2Q Q6600 @ 2.4GHz
is *3* times faster than the P4!

Is this really right -- and therefore I won't notice a performance
loss
on most of my applications if I got the Q6600 (and quite likely a
significant performance _gain_) even though the clock is somewhat
slower?
Click to expand...

How much L1 and L2 cache does the P4 have? The Q6600 has 128KB + 128 KB
L1 cache and 2 X 4 MB L2 cache. That, plus faster RAM and a 1066 MHz
FSB makes a big difference.
 
No, only in single threaded applications. In multithreaded applications,
it will kick its ass all over the place! No, a lot of times I find it is limited
by disk I/O speed, like when I am demultiplexing a .mpg file for editing.
The charts here help you compare, but I'm afraid they are mostly biased
toward multithreaded apps:http://www23.tomshardware.com/cpu_2007.html
The 2006 chart has the quad core listed, also.
Click to expand...

But is going from the 2.8GHz P4 to the 2.4GHz C2Q Q6600 a good
idea or not? Will I notice a big loss on the 1-thread apps, though, as
many programs are still 1-threaded? Although the primary reasons
for wanting the chip was to be able to use multi-threaded programs,
there's still that problem of running those single-threaded ones.
 
mike3 said:
I got a slow 66 seconds on my machine :(

Dang, this would suggest that a single core of C2Q Q6600 @ 2.4GHz
is *3* times faster than the P4!

Is this really right -- and therefore I won't notice a performance
loss
on most of my applications if I got the Q6600 (and quite likely a
significant performance _gain_) even though the clock is somewhat
slower?
Click to expand...

The Core2 design has a different IPC (instructions per clock) rating
than the Pentium4. Core2 can execute more instructions per clock cycle,
on average. For integer, the speedup is at least 1.5x, if not more.
For floating point, the speedup is still there, but not quite as great
as the integer factor.

For integer, in this chart, I selected a Pentium D 960 versus the
Core2 E6600. The core clocks have a ratio of 1.5x, and the Core2 still
does better. So for whatever this test is doing, the factor is
more than 1.5x.

http://www23.tomshardware.com/cpu_2007.html?modelx=33&model1=887&model2=877&chart=410

On floating point, there is less of a lead of Core2 over P4 than
for integer. I seem to remember the lead was less at one time,
so I don't know if this has to do with the version of Sandra or
not.

http://www23.tomshardware.com/cpu_2007.html?modelx=33&model1=887&model2=877&chart=411

To do such a comparison properly, a lot more single threaded software
should have been tested. Toms is only interested in multi-threaded
performance, as they're careful to use the latest multi-threaded
version of the packages that they can find. I don't think you'll
catch them running SuperPI 1.5 any time soon. Maybe someone is paying
them to do that, because a lot of ordinary users would be most
interested (curious even) as to how single threaded things work.
Obviously they've put 100's of hours into those benchmarks, but
didn't seem to be very creative in their selection of things to
try.

Paul
 
mike3 said:
Hi.

I'm curious. How does a single core of an Intel Core 2 Quad Q6600
processor running at 2.4GHz compare to a Pentium 4 running at 2.8GHz
(the latter is the type of processor I have now)? Does that 400MHz
drop in clock rate give a significant decrease in terms of single-
threaded performance, or does the chip design potentially quell some
of that? Ie. clock-for-clock, does the Core 2 run faster, slower, or
the same as the P4, on average?
Click to expand...
One core of a 2.4 GHz Core 2 Duo or Quad will run circles around a
2.8 GHz P4. The reason is that the Core 2 performs more instructions
per clock cycle than the P4. The extra cache also adds to the
perfomance. That's before considering the fact that the RAM, hard
drives, and video cards in Core 2 systems will most likely have higher
perfomance than those in older P4 systems. Add multithreaded apps
to the mix and the Core 2 Duo/Quad will leave any P4 in the dust.
 
mike3 said:
But is going from the 2.8GHz P4 to the 2.4GHz C2Q Q6600 a good
idea or not? Will I notice a big loss on the 1-thread apps, though, as
many programs are still 1-threaded? Although the primary reasons
for wanting the chip was to be able to use multi-threaded programs,
there's still that problem of running those single-threaded ones.
Click to expand...

SuperPi Mod 1.5 is single threaded, that's why I chose it. I set the core
affinity to one core because it would otherwise use another core slightly
to update the GUI. An app that doesn't use all four threads would be
faster with a C2D with a faster clock speed. Any particular app that you
are concerned about?
 
Fishface said:
Well, I'm sure it depends upon the application, but I de-clocked my
overclocked Q6600 to stock speed, set the core affinity to a single
core, and ran SuperPi Mod 1.5.
http://www.xtremesystems.com/pi/
It completed in 21.235 seconds. At 3 GHz, it did 16.828 s.
(i965 chipset, DDR2 at 4-4-4-10)

How does the P4 2.8 do?
Click to expand...
Obviously this benchmark is not the be all and end all to judging CPU
performance but it is still a tad depressing that I can only achieve about
30 seconds with my AMD X2 6000 (3.0GHZ)
:-/
 
Intel was supposed to release some affordable 45 nanometer quad core
CPUs about now, but I think that got pushed back a bit.http://techreport.com/discussions.x/13878

At this point, I think I would wait for one of those.
Click to expand...

Hmm. However, the 2.83GHz one is still a little too pricey for my
liking. The next two on the list I probably could do, but do they
at least match the performance of the P4 on the single-thread
applications (ie. going to them will not result in a significant
drop in performance)?
 
SuperPi Mod 1.5 is single threaded, that's why I chose it. I set the core
affinity to one core because it would otherwise use another core slightly
to update the GUI. An app that doesn't use all four threads would be
faster with a C2D with a faster clock speed. Any particular app that you
are concerned about?
Click to expand...

Well, the apps that are my reason for getting the chip support
multithreading
up to and even beyond 4 threads, however I still make use of some
older
1-thread apps like a video editor that I've got and I don't want to
lose a
significant amount of performance on them.

The question is how do the Core 2 cores compare to the Pentium 4
cores, clock-for-clock: does a Core 2 get more done at a given clock
rate
than the Pentium 4, and if so, does it make up for the 400MHz drop in
clock speed between the 2.8GHz P4 and the 2.4GHz C2Q?
 
mike3 said:
Hmm. However, the 2.83GHz one is still a little too pricey for my
liking. The next two on the list I probably could do, but do they
at least match the performance of the P4 on the single-thread
applications (ie. going to them will not result in a significant
drop in performance)?
Click to expand...

Isn't this one cheap enough ? This looks like a pretty sweet deal to me.

INTEL Core 2 Duo E8400 3.0 GHZ EM64T 1333MHZ RETAIL $205
http://www.mwave.com/mwave/viewspec.hmx?scriteria=MB-BA24501&RSKU=MB-BA24501

Has a 65W power dissipation.
http://processorfinder.intel.com/details.aspx?sSpec=SLAPL

Paul
 
Isn't this one cheap enough ? This looks like a pretty sweet deal to me.

INTEL Core 2 Duo E8400 3.0 GHZ EM64T 1333MHZ RETAIL $205http://www.mwave.com/mwave/viewspec.hmx?scriteria=MB-BA24501&RSKU=MB-...

Has a 65W power dissipation.http://processorfinder.intel.com/details.aspx?sSpec=SLAPL
Click to expand...

It is, but I wanted the quad-core for the multithreaded applications
that can take advantage of all those cores. Like rendering 3D
graphics, for example.

So that's why I want to know how the 2.4GHz cores in the Core2
Quad stack up to Pentium 4 cores.
 
Back
Top