comparing duel core vs. single core

In the spring I purchased a Dell with a core duo 2.134 GHz CPU. Now I find
that a video editing software package I want to buy (Pinnacle Studio
Ultimate) requires a duel 2.4 GHz minimum.

Elsewhere I see some writers saying a minimum of a 3 GHz single core is
needed for editing hi definition video.

So, now I'm curious, how do you compare the numbers? My 2.134 GHz CPU is
about equivalent to what in a single core?

Joe

Joe wrote:
> In the spring I purchased a Dell with a core duo 2.134 GHz CPU. Now I
> find that a video editing software package I want to buy (Pinnacle
> Studio Ultimate) requires a duel 2.4 GHz minimum.
>
> Elsewhere I see some writers saying a minimum of a 3 GHz single core is
> needed for editing hi definition video.
>
> So, now I'm curious, how do you compare the numbers? My 2.134 GHz CPU is
> about equivalent to what in a single core?
>
> Joe

There is a review here. There is quite a difference in requirements
between a 2.4GHz Core2 Duo and a 3GHz P4. The difference is a factor
of 2.4x, which means the Corel editor doesn't need quite as much
processor for some reason.

http://www.pcworld.com/article/id,133300-page,1-c,topics/article.html

You cannot compare a single core and a dual core very readily. The
reason is, there are different kinds of workloads (old software
programming style versus new).

To give an example, take a look at this AMD processor chart. AMD uses a
rating system, a single number for each processor. If I compare one
of their 2.6GHz single core processors, to one of their dual core 2.6GHz
processors, both with the same cache per core, they rate one as a
4000+ and the other as a 5000+. The implication is, the second core
allows a 1.25x speedup, rather than a simple doubling.

The reason for that, is the mix of software used in their benchmark.
Something like Microsoft Office, might tend to run on one core. Whether
it is running on a single core processor or a dual core processor, it only
uses a single core.

The other extreme, is software that uses both cores equally. Photoshop
contains a mixture of filter types. Some filters only run on one core.
Some filters run on multiple cores. When they run on multiple cores,
the load on each core is equal. Thus, if I pick the right filters
in Photoshop, the dual core processor finishes the task in half
the time of the single processor. If you were a Photoshop user,
you'd be going "that one is 4000+ and that one is 8000+".

Now, we consider a computer game and its behavior. The other day, I
found some results for a couple games. The games were run on a quad
core processor. The user claimed he saw 100% loading on one core,
while the other three cores were around 30% each. This unbalanced
result, is because tasks spawned by the game don't have equal
responsibilities. There might be AI, Physics, and Rendering, but
a boss task has to synchronize all that, and that would be the one
running at 100%. If could well be the render/boss is the thing running
at 100%. Now, with those numbers, I would well hypothesize that the
three 30% activities could be squeezed onto a dual core. On a dual
core, maybe the load becomes 100% on one core, and 90% on the other.
In other words, it is possible in that case, that the user with the
dual core, and the user with the quad core, see equal frame rates.
The dual core is pretty well completely used in that case, with no
appreciable margin left over.

So what you're asking me to do in effect, is guess at what kind
of programming the dude at Pinnacle did. Is he a single threaded
traditionalist ? Has he taken an Intel course on multithreading ?
How good is he at spreading the load equally ? That determines
how a single core compares to a dual core, and whether the dual
should be awarded a 5000+ rating or an 8000+ rating. It depends
on how well the software is able to use a second or subsequent
cores.

The other aspect is the instruction level parallelism, on any
give architecture. If we compare a P4 at 3GHz to a Core2 Duo
at 2GHz, then core for core they are about equal. That means
my "single core conversion factor" is about 1.5x. That means one
of your 2.134GHz cores, is about the same as a 3.2GHz P4. Now,
that comparison isn't really fair, because the different architectures
are like apples and oranges. For example, the conversion factor
for floating point, is not nearly the same as the 1.5x I quoted
for integer stuff. But the integer number remains the most likely
to be applicable to a wide range of stuff.

The single biggest improvement anyone ever gets, is by sending
the software developers back, to recode and optimize their product.
To give an example, when Microsoft Flight Simulator came out,
at least some users went on a crusade to try and get acceptable
frame rates while they were playing. They tried all sorts of
stuff, fancy processors, SLI graphics, 4GB installed RAM, fast
hard drives, the works. But they never really "cracked it".
Performance remained as elusive as ever. Microsoft then came
out with a Service Pack for the game, called SP1. That gave
an immediate improvement over the initial product. And
an improvement "for free".

So at least sometimes, the right answer is to try a different
application, rather than purchasing yet another computer. If
an application is a pig, sooner or later, the software maker
will receive market feedback to that effect. Another explanation,
is that one tool is not as capable as the other, or the
quality of the rendering is different between them. The above
PCWorld review doesn't really use a lot of science in evaluating
the results.

In your shoes, I'd just buy the product and try it. There may be
enough user adjustments, to allow you to get some use from the
thing. I don't expect it'll fall on its face, because you're
short by 11 percent.

Something I like to see, is trial versions of products
available. That is the quickest way to see if the product
simply is not going to work out. If you don't see the trial
on the site, send them an email and ask why not. Receiving
email like that, will help speed the process along. Eval
software makes a big difference to decision making.

Paul