Rating Laptops vs. Desktops

Is there a formula or way to compare the speed of today's laptops to
older desktop computers? Since the companies seem to downplay
processor speed, it's harder to do.

 

They're called "benchmarks" and they have to be carefully
selected to study a particular aspect of the processor.

On my systems here, I might use "SuperPI" which is a math
calculation. You can set it up so it fits within cache
(calculate a small number of digits) or set it so it
has a memory footprint larger than the cache. If a fair
comparison is desired, then both processors have to be
put at the disadvantage of the program not fitting in cache.

Between my P4 and Core2 based systems, I might have found
a completion time ratio of 1.5x or 1.7x or whatever. The Core2
is faster, clock for clock. A 3GHz Core2 equals a 4.5GHz Pentium 4.

That benchmark compares execution on a single core of a multi-core
processor. As such, it states the more pessimistic gain to be
expected from a processor. If your old processor had a single core,
your old software was single threaded, you got a new multi-core
processor, a single core on the new processor might still be used
to run the old software. SuperPI helps you decide what the "worst"
performance ratio might be. In my case, I have a dual core. If my
old software uses a single core, I see 1.5x to 1.7x. If my software
was multi-threaded, I might see 3.0x to 3.4x, as both cores are
running the software.

Most sites providing benchmarks "drink the Koolaid" (Intel promotional
material) and only show multithreaded benchmarks. This tends to
over-emphasize the performance of the new processor. You might find
that a typical application mix, has very little multi-threaded
software. Some games are multi-threaded, but not in a symmetric
way (core split 100-30-30-30 perhaps). Photoshop has been multi-threaded
for years, but Photoshop only has multi-threading for half of the image
filters. The reason for that, is some algorithms, cannot be effectively
split amongst multiple cores, so they remain single threaded. And thus,
if a person is estimating performance, they could easily use the
SuperPI benchmark result as an indicator.

SuperPI results are recorded on hwbot.org , but the problem with that
site is an emphasize on overclocking the CPU. Which makes the site
worthless for people seeking to compare processors. There are some
processors, where I can't get a rating at the CPU stock speed.

The best compromise is this site. Benchmark is multi-threaded, overestimates
the results, but at least you have *something* to compare with. Each CPU
receives a "PassMark" result. Now if someone could do a "SuperPI" site like
this, that would be a nice complement to the PassMark info. I feel there is
room on a benchmark site, for both kinds of benchmarks.

http://www.cpubenchmark.net/

*******

The Windows Experience Index, is something Windows calculates. There are
forums filled with "well, I got a 7.9". I can't see the point really - I'd
rather have something a bit more concrete, such as numbers I can take ratios of.
Passmark will do that for you. You can't take ratios of the Windows computed
numbers, as no attempt is made to make them linear. If one machine gets 7.9
and another 7.8, does that tell me when my movie render will finish ?

http://en.wikipedia.org/wiki/Windows_System_Assessment_Tool

(From one of the reference links in the article...)

"Achieving the maximum Windows 7 WEI score of 7.9 requires nothing short
of the best of the best hardware components on the market."

HTH,
Paul