Slashdot Mirror
Centrino Mobile Equals Desktop Pentium 4 in Speed
Spy Handler writes "On Wednesday during the launch of its new Sonoma Centrino Mobile, Intel put on a demonstration running a video game on a laptop. It matched the performance of a high-end Pentium 4 desktop running the same game, declared Intel. The contenders were a laptop sporting a 2.13 GHz Pentium M processor, 1GB RAM, and the Alviso chipset versus a desktop with a 3.6 GHz Pentium 4 with hyperthreading, 1GB RAM, and the Grantsdale chipset. Is this a testament to how far the Pentium Mobile architecture has come, or a sad comment on the clockspeed-pushing design of the Pentium 4?"
Is this a testament to how far the Pentium Mobile architecture has come, or a sad comment on the clockspeed-pushing design of the Pentium 4?
I think it's a testament to the fact that whatever game they were running doesn't bottleneck at the CPU. Most video games are not CPU-limited beyond a GHz or two.
Centrino is the Pentium-M with a wi-fi chipset. I think the article was meant to highlight the pentium-M's performance versus the pentium *4* performance
Centrino is fine and dandy, but I still want a little bit more speed than that. I want a Pentium M, but I also want to know how it really is vs that desktop. I would really like to have the fastest thing for a new laptop, as judged by experienced people. So Intel, come on and test all your chips against that 'fast' desktop!
FYI, Centrino is the same thing as Pentium M. Centrino just means you bought a laptop with a Pentium M processor and Intel's wireless chipset. bundled together. Why, oh why does Intel insist on giving everything some crappy marketing name and confusing customers?
"When the president does it, that means it's not illegal." - Richard M. Nixon
They don't really mention any of these factors about the laptop. What good is having good performance if it weighs 10 pounds and has a battery life of an hour?
Monstar L
That's part of the beauty of the Pentium M: most of the notebooks based on it are in the 6lbs or less range, and I've yet to see one with a battery life of less than four hours.
As I had mentioned in a previous comment, the front side bus speed is the biggest limiting factor on Pentium M processors. The day we see an 800MHz FSB Pentium M is the day the direct MHz comparisons will apply (i.e. 1.8GHz P-M vs. 1.8GHz A64). Even the Tom's Hardware Guide review of the new Sonoma chipset for P-M shows fairly marginal gains and proves the FSB is the limitation, PLUS they do the stupid thing here and put in DDR-2 which does little for performance but increases system costs.
It's a shame all the benchmarks disagree. Have a look at Benchmark
This benchmark also shows that a Pentium M 2.3 (Yes it is overclocked) is as fast as a AMD Athlon64 FX-53 (2.4 GHz)
It isn't just an issue of clock speed, but also CPU core voltage. Throttling the clock speed from 2Ghz to 400Mhz can also be combined with a lower CPU voltage. So you are not only running the CPU slower, but also supplying it with a lower voltage.
I don't know about P4, but on AMD systems this allows for the CPU fan to completely stop when the CPU is idle (on desktop systems).
Here is a link to a benchmark that show that intel might be right.
This benchmark shows that a Pentium M 2.3 (Yes it is overclocked) is as fast as a AMD Athlon64 FX-53 (2.4 GHz) in many games
Also you can generally turn off HT in the bios, if you're really interested in seeing what effect it has.
I tried it with 3DMark2003. Turning off HT makes a very small increase in performance.
The P-M chipsets etc. do not support SMP. Maybe later.
Bob Hyatt is a very smart man, so I'd usually take his word as fact. That the P3 beats the P4 on all integer tasks, however, is just wrong.
The Pentium 4 has much, much better integer performance than the Pentium 3. The P4's dual ALUs actually function at 2x the clockspeed of the chip, meaning it has a theoretical peak which is simply ridiculous compared to the P3's. This was because Intel thought most users used stuff like word processors and power point, which comprise mainly integer operations.
What Intel forgot is that these things were already fast enough on old processors. The issues with performance on the P4 are caused by its long pipeline: it becomes very slow dealing with "if" instructions. That Mr. Hyatt dislikes the P4 makes a lot of sense: Crafty is a very branch heavy program (thought its bit-board move generation and evaluation contains few branches which are pretty easy to predict, the negascout search is one of the worst applications imaginable for a processor where mispredicts are deadly). It is far more likely that this is the culprit than the P4's fantastic integer performance.
If you remember your history, the Pentium-M is based off of the Pentium-III. You know, the same P3 that kicked the crap out of the P4 when it came out, despite the fact that the P3 was topping out around 1.0Ghz and the P4 started at 1.4Ghz.
Unless you mean that the the Pentium M has no hyperthreading, which I suppose is a valid point, as hyperthreading is about all the P4 has going for it right now.