Centrino vs Athlon Mobile 3200 - Page 2

You can say the same about the A64 (except the FX) in the anand benchmarks have 512K L2 cache, inherent biased some of the tests in favor of the 2MB L2 of the PM. The in real life performance, the cache will be less of a factor (no one is going to run a business application with 100 percent CPU usage for 10 minutes straight, except on high end workstations). That is why I linked you to the bench of A64 4000, which has 1MB L2, which evens it out a little. The 533 MHz bench mark compared to the old 400 FSB Dothan only showed << 10 percent improvement. The greatest problem for the sonoma is not bandwidth, it's latency. It's still using the same 4 year old quad-pumping FSB, while the A64 is using the asych serial fully duplex Hypertransport + separate direct interface for memory controller.



I never claimed that A64 is 115% faster than PM, I said it is about 115% the performance of a PM at comparable clockrate. For a majority of business applications, a P-III will suffice. I don't think average offic applications are going to really differentiate the top performers from the pack.

And A64 right now does not have Hyperthreading, which is irrelevant for most tests except encoding and rendering anyways. A64's are faster because of superiro FPUs, larger internal buffers, better decoding units, larger and larger number of registers, larger branch target buffer, and basically a superior form of superscalar architecture than the PM. PM is for the most part based on the P-III architecture, while a venerable CPU, is approaching the limits of its performance, which is why in 06 the Yonah will need to go dual core to improve upon the present performance significantly.



I don't know if his GPU is underclocked (It's hard to believe why HP would do this, if they do , they should call it the 9550, which is the underclocked version of 9600) But many of the symptoms are clearly CPU related, such as characters not responding to the dynamics of the building order (you have to have played this game to understand: that directing the construction of buildings and facilities is the main way to influence the AI of the characters), and simply the response time of the input devices. All of these are clearly CPU related, not graphics subsystem.

I fully agree with better batter life and cooler notebook, etc, as I said in the earlier posts.

I misread your assertion of the 115%. Sorry.

I have to agree with your info about AMD here. Yes P-M will only go so much further until Yonah arrives. Apparently Yonah wont be much power hungry than Dothan. If that is true that will be good news for mobile computing.

Vendors may sometimes underclock the GPU to boost battery performance etc. Your point on the CPU/GPU aspects of the game are very pertinent and informative. I am not much of a gamer actually.

All in all it was nice discussion - started by my misreading your comment about the 115%

Thanks

Thanks for your understanding as well.

I do agree that PMs are superb CPUs for mobile platforms, especially for their thermal characteristics, which A64 or any other line of CPU has come close. Maybe Turion will give sonoma a run for its money, but that's the future.

I didn't mean that your points and the article you linked are irrelevant. They are indeed well taken, and I agree with with you on most of them. I just think that PM has its limitations, especially compared to some of the top end A64s.

It's been a good discussion, and look forward to chatting with you again krishd, in the future here in the forum.

HardBall

Here are a few reviews that I think displays the differences in performance very well:

http://techreport.com/reviews/2005q1...f/index.x?pg=5

http://www.gamepc.com/labs/view_cont...id=gmso&page=4

http://anandtech.com/cpuchipsets/sho...spx?i=2129&p=7

They will come worldwide available on April 18 according to this:
http://www.digitimes.com/systems/a20050205A2003.html

It's really that simple:

Business traveler -- always PM, no need to think about it (and always IBM of course)

Gamers -- Never PM, no need to think about it (unless you like watching jerky frames and defunct AI)

Students -- Unless you have the dough, no need to spend money on either PM or A64, a Sempron laptop will do you as much good for taking notes and writing papers.

Unless you are getting the Acer Ferrari 3400. Then you get both raw power "AND" a light notebook..hehe, not sure about battery life..but Snorre could swear by it.

I just read the hothardware.com comparo of the PM vs the Athlon 64...the Athlon 64 was a desktop chip using dual-channel memory? Wouldn't it have been a more reasonable comparison to have used a single-channel S754 chipset mobo with a DTR Athlon-64 3200+ at 2.0 ghz so that the PM wouldn't have to be overclocked? I believe the DTR 3200+ also has a 1 MB cache as well?

i am not sure why they keep comparing desktop cpus to mobile, do a mobile to mobile test and leave desktop cpu's out of it. Its like comparing a ferrari and a kia two cars specifically made for different reasons, as is with the mobile and desktop cpu.
It will be great when amd releases the Turion so we can have some wicked, I would like a test just mobile chips, intel and amd, with as much as the same specs as possible .
when i see an FX processor in a mobile test benchmark i dont bother reading it, why dont they add some server chips in there as well,

example like the test xbitlabs did
the ferrari 3400 and the asus amd notebook, where it did better.
now those kinda of test i'd rather read

Then how about if we disable half of the 2MB cache of PM so that it would be matched with the amount of L2 on the S754 A64. The cache was added on the PM to mask the performance of the memory BUS, so if you even the memory bandwidth, you have to get rid of the excess cache to make it fair.