Looking at buying new laptop, advice on architecture

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Hi all.

My laptop is currently 18 months old, and is a Toshiba P30 (can't remember
model #). It's a 3.2Ghz P4 with 512mb ram, 64mb graphics card etc... It's
just not cutting it any more and I'm looking at getting a new beast.

I'm looking at the HP dv9008tx which has a Core2 Duo @ 2 Ghz. Even with the
significantly lower clock speed, and irrespective of the dual cores, is the
processor going to be significantly faster than the 3.2G P4? I'm aware the
FSB is faster but realistically will I notice a huge difference?

I'm not massivley up on hardware and am much more focused on software. I
predominatley use my PC for coding, graphic design and the occassional DV

Re: Looking at buying new laptop, advice on architecture

A 2 Ghz Core2 processor has about the same performance as a 3 Ghz or 3.2 Ghz
Pentium 4. You won't see any improvement unless your applications use both
processor cores or you run several applications at the same time. FSB speed will
affect performance but is not as important as processor or disk speed.

OPM wrote:
Quoted text here. Click to load it

                   Mike Walsh
            West Palm Beach, Florida, U.S.A.

Re: Looking at buying new laptop, advice on architecture

OPM wrote:
Quoted text here. Click to load it

Mainly, you want to find actual benchmarks of hardware, rather than
opinions. Or actual owner comments. In my experience, architectural
analysis is a pretty inaccurate way to answer the question. But the
extrapolations I had to do below aren't much better.

The T7200 processor is here:

2GHz/FSB667/4MB L2 TDP=34Watts  Dual Core

The nearest desktop with 4MB cache, would be the E6600.
The reason for my interest in this processor, is you are more likely
to find benchmarks for it.

2.4GHz/FSB1066/4MB L2 TDP=65Watts  Dual Core

This might be your present processor. It could be a Prescott.

3.2GHz/FSB533/1MB L2 TDP=88Watts  Single core, Hyperthreaded.

First thing to notice, is the new box should run cooler. Power management
might be a bit better. So you get longer battery life.

In terms of threads of execution, a single core with Hyperthreading disabled
could be considered 1x performance. When you turn on Hyperthreading, the OS
thinks there are two processors. But the virtual processors fight with one
another for hardware resources. They are resource constrained. When some
applications run on such hardware, performance may rise to, say 1.1x . If
there is thrashing, performance could drop to 0.95x . If performance was
an important aspect, you would want to benchmark the high runner application,
to be able to make the decision to enable or disable Hyperthreading.

With dual core, you get 2x performance, minus any resource contention. The
eneny of performance, is cache coherence. If both cores wish to be informed
about the state of a common chunk of system memory, then depending on the
architecture, cache snooping between cores can eat up bus bandwidth. But
we can assume for simplicity sake, that this is not an (important) factor,
and we get closer to 2x.

You get to see the 2x in Photoshop. Photoshop should scale pretty well, with
the number of processors. Other applications could be single threaded, and
run on a single core. For those applications, you mainly see the speed of
one core, so back to 1x. User level concurrency (shrink a DVD with one core,
while using MSWord on the other core), allows the user to get better use of
their 2x capable hardware.

In comparing the T7200 to the E6600, we can ignore the FSB. Yes, the FSB
does have an effect, but remember that the processor has a large cache,
and a low miss rate. It is not going to the bus that often for data. The
FSB effect will show up in Photoshop, but for many other applications, not
to nearly the same degree.

Where does that leave us ? We find a benchmark for an E6600, and scale the
number we find by 2.0/2.4 or 0.83 times.

Tomshardware has some benchmarks, but they tend to emphasize multithreaded
applications. I don't know if any benchmark chart here, is representative
of single core performance. A Prescott 540 3.2GHz might be representative
of your existing computer. The E6600, scaled by 0.83x, gives performance
for your new platform.


About the best I can do for you, is Sandra Integer and Floating Point benchs:

E6600=22038, x0.83==> 18291  Prescott_540=7743             (ALU - Integer)

E6600=15406, x0.83==> 12787  Prescott_540=9862             (MFLOPS)

To compare these two processors on a single threaded application, I'll assume
(and this could be wrong), a perfect scaleup on those two benchmarks. I'll
take 1/2 of the E6600 numbers, to represent the power of one core, when only
one core is being used. 18291 * 0.5 = 9146 compares to 7743.
12787 * 0.5 = 6394 compared to 9862.

Most instructions executed on the new computer will be integer. Comparing single
core performance, the new processor would be 1.18 times as fast on a single
application. It is 18291 / 7743 = 2.36 times as fast on a multithreaded
application (Photoshop or other multimedia applications that happen to scale
well). The floating point, on the other hand, looks a bit weaker. Maybe running
a Prime95 benchmark on a single core, would return lower results. If an
can do floating point on both cores, then it is a win.

In the above benchmarks, I've had to make entirely too many assumptions
for this to be predictive of performance.

There is another benchmark here. The T7400 runs at 2.16GHz, so the T7400
numbers have to be scaled by 0.926x . SPECINT and SPECFP are single
threaded, so the idea is, this is comparing single cores. HT was disabled
in ths second benchmark, presumably as a means to get the best numbers.
These spec numbers probably overestimate the advantage.
2284 x 0.926 = 2115 versus P4 = 1348
2173 x 0.926 = 2012 versus P4 = 1438


Dell    Precision Workstation M90 (Intel Core 2 Duo processor T7400)
        2 cores, 1 chip, 2 cores/chip             CINT2000_BASE=2284
Dell      Precision Workstation M90 (Intel Core 2 Duo processor T7400)
        2 cores, 1 chip, 2 cores/chip             CFP2000_BASE !73

Dell    Precision Workstation 360 (3.20E GHz Pentium 4)
        1 core, 1 chip, 1 core/chip (HT disabled) CINT2000_BASE=1348
Dell    Precision Workstation 360 (3.20E GHz Pentium 4)      
        1 core, 1 chip, 1 core/chip (HT disabled) CFP2000_BASE 38

So is there anything I can say in conclusion, based on some worthless
benchmarks ?

1) You get two real cores.
2) Comparing one core, to your present processor, you get a slight
   boost as near as I can tell. The difference shows up for "stopwatch"
   applications, where a long compute job takes less time. The desktop
   will not be any snappier (unless graphics/disk are faster etc).
3) Your battery life could be longer. Of course, the two notebooks
   won't be an apples-to-apples comparison when all hardware is
   taken into account.

The rest of the laptop (graphics subsystem, hard drive etc) also play
a part in user experience and usability, which I've ignored here.


Re: Looking at buying new laptop, advice on architecture

OPM wrote:
Quoted text here. Click to load it

Depending on what you mean by "just not cutting it any more" you may
very well not _need_ a new computer at all (leaving any hardware lusts
out of the decision). Do you mean that doing the same tasks as eighteen
months ago they are taking far longer? Or do you mean that the tasks I'm
performing now are radically different than the ones from eighteen
months back and the computer cannot perform them well? In the former
case, and possibly in the latter, you may simply be suffering from the
detritus accrued during eighteen months -- fragmentation, too many
background processes running, misconfigured hardware and drivers, and
perhaps even malware running in the background.

Coding, even using the most inefficient tools, is not terribly
demanding. Most graphic design is more demanding but unless complex
filters are being applied constantly to huge images shouldn't be
overwhelming. DV editing can be very demanding, especially on memory
bandwidth and swap space. Complex shoot-em-up gaming is probably the
most demanding task that most PCs experience but you don't say that you
are doing that.

I suspect that given the relative newness of your system and the tasks
you list that doubling or tripling your RAM and doing a thorough cleanup
of the system (possibly even a compete OS re-install although I don't
normally encourage anyone to do that) may give you more than enough
power to keep using the same system for a couple of more years.

John McGaw
[Knoxville, TN, USA]

Re: Looking at buying new laptop, advice on architecture

John McGaw wrote:
Quoted text here. Click to load it

I agree.  I am presently running a P3 at 450 Mhz with 512 G of
memory under W98.  I keep the system pretty clean.  I cloned it
here from a 486 at 80 Mhz with 64 Meg, which was too slow.  However
that machine dated from the mid to late 80s, the P3 dates from the
late 90s.  Upgrading the memory in the P3 from 128 M to 512 M with
ECC leaves the swap file almost unused.

I am quite satisfied with the performance.  I don't do games,
video, etc.

Chuck F (cbfalconer at maineline dot net)
   Available for consulting/temporary embedded and systems.

Site Timeline