Three-way CPU slugfest: AMD vs Apple vs Intel

“AMD, Intel or Apple: it doesn

30 Comments

  1. The writer and editor were both smoking crack when they wrote that article. And how in the heck are you going to report your assessment of the numbers when you don’t even show the reader the numbers so that he/she may come to his/her own conclusion.

    Fodder.

  2. This is what I e-mailed to the editor of that article…

    Did you know that…

    Despite the fact the AMD is 64bit, it is only an extension of the 32bit architecture. Therefore, when running x86 applications, it is running in a highly optimised way. The main advantages of a 64bit chip is it’s ability to crunch large datasets. Therefore, a 16GB 64bit server running a 64bit OS will perform better than the same with a 32bit chip running a 32bit OS (note : 32bit chips are usually limited to 4GB memory limit per process). The only main advantage of 64bits chips are their ability to do things in 64bit chunks (i.e. multiply 2 64bit numbers in one operation as opposed to it taking twice as long on a 32bit chip).

    The G5 on the other hand is woefully under-optimised. It will run standard 32bit apps as fast (and faster due to high clock speeds over the older G4s) as the previous generation G4s. To give you an example, Apple re-compiled and optimised the MPEG “Compressor” application (which comes with FCP) which resulted in a 55% boost in speed. This is NOT the same as saying they made it 64bit – it is not 64bit – still the same 32bit code but optimised by G5 aware compilers.

    To provide another example, you state that the AMD chips are better at scientific math than the G5. This may seem so on the surface in your testing; but for the same reasons given above, it is not the whole truth. Basically the G5 has two FPU units and 2 vector units which are designed to process data simultaneously. Unfortunately, the current compilers (gcc) can not optimise code to do this just yet. IBM are about to release their XL PowerPC native compiler – which I believe Apple will incorporate into their IDE – which should be able to do this. Basically as they further refine the PPC compilers, the faster the Macintosh G5’s will run the applications (and thats without adding anything to the machine). Time will bear this out, mark my words.

    Lastly, if the AMD is better than the G5 at scientific computatio, then why has a certain University managed to build the 3rd fastest super-computing cluster on earth using only 1100 Apple G5 Macintoshes for around $5million – and yes there are AMD based SC’s in the list also but further down.

    I’m not saying the G5 is better than AMD or Intel; just trying to bring some pertinent points to balance up the information you have provided.

    Regards
    Andrew Sheridan

  3. From apcmag.com Article:

    “Apple is also cooking up a 64-bit release of Mac OS X. Both of these operating systems are expected in the first quarter of 2004.”

    Apple is… this is news to me.
    The ONLY (viable) OS is Microsoft XP that will be running a 64bit system.
    Apple is again as allways behind.

    m.

  4. I would have concluded that AMD are good for games, G5 are good for science and Pentium 4 is a kitchen utility that you can fry eggs on. Seriously, the world’s cheapest super-computer is built on PowerMacs and besides that, Unix applications can be launched on any OS X capable Mac. It seems like Intel is heading to irrelevance with a product that offers very little over previous versions of the Pentium processor.

  5. {quote]
    “AMD�s . . . and can cope with over a terabyte of 400MHz DDR SDRAM.”
    It can? Poor thing, talk about willpower. I’m guessing this is another deception.. I’m pretty sure the g5 can address over a terabyte of ram (that’s a bit more than 8gb, right?), too, it’s just a matter of having a system board that has enough dimm slots. Of course, they don’t say that about the AMD.. that’s a lot of slots to fit in one computer.. Gack, well I can ramble on for a lot longer about that, but I think I’ve made my point.

    {yet another quote} “Apple is again as allways behind.”
    I sure hope you’re joking there, bub, because..well, I’ll just count to ten and wait for other people to address that inaccuracy.. eh….. heh.

  6. i mentioned this before a few months back… but intel will become a dinosaur. i am hoping that IBM embraces the next insulation breakthrough technology – manufactured diamonds. intel has made an all-out commitment to silicon and will not look at anything else. but then again, intel was never a true innovator, IBM has always been.

    mark my words, we will see diamond technology in chipests by 2015 for the mass market… possbily as early as 2011. there was an excellent article about in wired last year and i have seen declassified specification papers about from the Navy. i believe MIT has looked at this as well.

    my mouth is watering at the prospect of a diamond-insulted G10 running at 1THz.

  7. oh…. and sputnik – your just a communist blow hard.

    XP? if you are serious, you are one of the biggest idiots out there. M$ will not have 64bit support for more thean 4 years – and that is being generous to the evil empire.

    Apple is 64bit now. OS X will be soon. and in 18 months, all new apps will be 64bit native. by the time M$ supports 64bit natively, Apple will be at 128bit.

    i feel so sorry for you, little person.

  8. ‘Lastly, if the AMD is better than the G5 at scientific computatio, then why has a certain University managed to build the 3rd fastest super-computing cluster on earth using only 1100 Apple G5 Macintoshes for around $5million – and yes there are AMD based SC’s in the list also but further down.’

    Hate to rain on your parade, but the 2004 ranking will paint a very different picture I think. Live by the sword, die by the sword. When Apple drops off the top ten, it will seem like a bigger deal than it is, simply because the initial success was so hyped up.

    None of the AMD or Intel machine had as many nodes (2200) as the VT system. 2200 Opterons will be faster than 2200 G5 at the same clock speed.

    1 G4 is enough for me though.

  9. ‘XP? if you are serious, you are one of the biggest idiots out there. M$ will not have 64bit support for more thean 4 years – and that is being generous to the evil empire.’

    64bit XP can be downloaded on bit torrent today and installed on your Athlon64. It should be out by fall.

    If you are thinking of Windows for Itanium, then:
    http://www.microsoft.com/windowsserver2003/64bit/default.mspx

    I know a lot of people running Server 2003 instead of XP because it is a slightly newer kernel.

    I am a Mac head, I just thought I would set the record straight.

  10. Sputnik: “Apple is again as allways behind.”

    First of all, its spelt “always”. Measure things how you will, I (and so many more) will take Mac OS X over _anything_ running Windows, regardless of which is faster. One last time… Apple is almost always in front, being a true innovator, while everyone else follows with poor implementations of Apple’s ideas.

    Hope your Win XP day is not too frustrating…

    Gordon

  11. I haven’t read the article, but it sounds like Overington is trying to stuff AMD and Apple into their niches, while giving Intel the big gaming market. Since some of you say there are no numbers in the article, I won’t waste my time reading it.

  12. ah, thank you perry… did not see that article listed before. i stand corrected. at least i am not afraid to admit when i had a misconception.

    but the question is this, when will M$ have 64bit support for the desktop? servers is one thing, and it makes sense. 64bit server/enterprise software architectures have been around for over 15 years and M$ was not the first to get there. but what about the consumer side? does M$ not feel i need 64bit software architecture to run video-editing software (final cut pro) or 3D graphics generation (lightwave 3D and maya) or music generation (logic) or general compositing (shake)? of and i guess photoshop, illustrator, indesign and quark will NEVER be 64bit native. and i have not even started with games yet… i am pretty sure Halo will suck running in 64bit mode.

  13. Directed toward Perry :

    Perry, the 3rd fastest supercomputing cluster was ranked as such in October 2003. Of course things move on – not disputing that. However, you neglected to mention that Stevey boy promised us a 3Gig in summer this year; and depending on how fast IBM can ramp up production, the speed bump increases will be fairly frequent (3 times a year?). You also say that the 2.2Gig AMD is faster than the 2.2Gig G5. The main point I was making in my response is that the Mac applications are NOT optimised whereas the AMD ones are. As stated on Ars, the PowerPC architecture (not talking about Apple here) is much better than Intel’s x86 period. It’s only now Apple has IBM’s considerable wealth of experience in chip design and money to boot that the PowerPC architecture can be brought screaming into the 21st century. Whilst the Intel/AMD high speed trains were screaming down the track, good old Motorola decided to keep the PowerPC chips powered by stream locomotives.

    I’ll also say this, Intel Itanium is also very powerful but with 20+ years of x86 running this planet, even Intel are finding it hard. And like the G5, the Itanium is not very optimised at all. For example the FPU performance of the Itanium is incredible.

  14. and if the power architecture is so bad for gaming, then why does IBM develop chips for nintendo gamecube, sony playstation2, and now the 2nd generation of MS xbox? and lest we forget all of nvidia chips for their graphics cards. and Intel is superior for gaming?

    i know i know, none of the chips mentioned before are true PPC chips, but they ARE all IBM developed… i sense a convergence in 5 years.

    i want what they are smoking!!!

  15. “None of the AMD or Intel machine had as many nodes (2200) as the VT system. 2200 Opterons will be faster than 2200 G5 at the same clock speed.”

    have to correct you there …go to the URL below and look at #6 on the list (it’s the first Opteron system) …looks like it has 2,816 2ghz Opterons in it ….also, you should take into consideration the Rpeak numbers …the VA Tech cluster has an Rpeak of over 17 teraflops, the Opteron system is barely over 11 teraflops

    http://www.top500.org/list/2003/11/

  16. Zwei,

    I forgot to mention that. Thanks.

    Isn’t it incredible that the posters who are berating the writer of that article make the same simple mistakes – don’t research before publishing articles or comments!!!!

    Perry is clearly barking. Probably and AMD guys I guess ” width=”19″ height=”19″ alt=”wink” style=”border:0;” />)

  17. And talking of not standing still the VSC is infact at this moment being revamped with the latest Xserve + G5 chips which will improve its power by an order of magnitude over the present set up.

  18. I still love how on the top 500 list, they dont show that the terrascale beast is running OS X and NOT linux like all the other clusters. Even the xeon cluster (has quite a few more CPUs than the Big Mac) is running linux and not some crappy brand of a M$ OS. Sorry to the guy preaching about the AMD 64 and the Opteron, but with more tweaking, and the xserves switching with the G5 towers, those systems will not compare. Nice try on the FUD though.

  19. I just love the line under the VT cluster on the top500.org website which reads, “Self Made”

    I think that says a lot. It should say “self made in 3 months”. The Top500.org should put up the data of how long each one took to build as well as OS and cost.

    I wonder how much of a boost the VT cluster will get when they move to the Xserve. Once they go to ECC memory, I wonder how much faster it would be.

    BTW, does anyone know if Xgrid can be incorporated into a cluster? Or what the peak Terflops of the seti/folding@home projects have hit?

Reader Feedback

This site uses Akismet to reduce spam. Learn how your comment data is processed.