When everything is 64 bit...

Posted:
in Future Apple Hardware edited January 2014
When every program is made for 64 bit computing including OSX, what kind of improvements in speed are we talking about?



5? 10? 15? 20?



25% change in overall speed?





Realistically, i know that it's not "double" the performance of a 32 bit system. But all these benchmarks just make me wonder what will be the true value of the Powermacs once they are really ready for action.
«13

Comments

  • Reply 1 of 59
    chuckerchucker Posts: 5,089member
    Um, like, none? Unless you're talking about database applications etc.
  • Reply 2 of 59
    Not just database apps, There are some other ones where you will see quite a big increase but very few apps that the average user, uses. Safari for instance would see little to no speed increase from 32 to 64bit memory addressing.
  • Reply 3 of 59
    programmerprogrammer Posts: 3,455member
    We should actually be talking about how much of a speed loss most apps would see in going to 64-bit.
  • Reply 4 of 59
    so that's it!!!



    we're here? and 64 bit only means it can adress more ram??





    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?
  • Reply 5 of 59
    mr. memr. me Posts: 3,221member
    Quote:

    Originally posted by macnn sux

    so that's it!!!



    we're here? and 64 bit only means it can adress more ram??





    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?




    64-bit addressing has immediate benefits for every user of newer computers. A 32-bit computer can directly address only 4 GB. As hard disks grow exponentially in capacity, vendors of 32-bit OSes have resorted to "hacks" to address them. The last published modification of MacOS 9 was such a "hack." A 64-bit computer can directly address the sectors on large hard disks without resorting to hacks.
  • Reply 6 of 59
    socratessocrates Posts: 261member
    It's not just about memory addressing. 64 bit computers deal in 64 bit NUMBERS, i.e. ones that are 4 billion times as big as 32 bit numbers.



    This has implications for any apps that do a lot of number crunching. Programs that previosuly had to do math in several stages in order to avoid loss of precision can now do it in one. Apps that previously had to use slow floating point math can now use fast integer math instead in many cases. Apps that have to move large chunks of data around can now do it more quickly because they can send it in bigger lumps.



    Photoshop will be faster for example, as will any well-written graphics or audio apps. The effect on games will be negligeable however since they do most of their processing with the custom graphics hardware which is already 64 or 128 bit (or more) for most operations (as in "Rage 128").



    I cannot see any reason why any apps would become slower - 64 bit numbers are not harder to deal with than 32 bit ones (unless you are using 32 bit hardware).



    Socrates
  • Reply 7 of 59
    programmerprogrammer Posts: 3,455member
    Quote:

    Originally posted by Socrates

    It's not just about memory addressing. 64 bit computers deal in 64 bit NUMBERS, i.e. ones that are 4 billion times as big as 32 bit numbers.





    This will only help programs that deal with large amounts of integers >2 billion or < -2 billion. In my experience most software deals with integers that are fine as 32-bits, or they use floating point. Some will use 64-bit integers but so rarely that any speed improvement from 64-bit hardware is insignificant. Sure there is the occasional piece of software that will benefit noticably, but it is rare -- very rare. Naturally those will be among the first converted to 64-bit.



    The reason I say a 64-bit program will get slower is because in 64-bit mode all of the pointers double in size, which takes more memory. More memory means more cache misses, and a cache miss is the slowest thing a process does -- even with the 970's superfast FSB. Waiting for data coming from memory takes >100 cycles according to Apple's documentation on the G5.
  • Reply 8 of 59
    kiwi-in-dckiwi-in-dc Posts: 102member
    Quote:

    Originally posted by macnn sux

    so that's it!!!



    we're here? and 64 bit only means it can adress more ram??





    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?




    Repeating myself somewhat from another thread - perseveration is a sign of frontal lobe brain damage I know, but please try to ignore that that that that



    In the end, for most people doing general PC stuff - document editing, spreadsheets, web surfing, games etc. 64-bit is pretty much meaningless IMO.



    Where it becomes useful and important is where you are dealing with very large datasets i.e. > 4GB limit imposed by 32-bit addressing.



    So, where do we see these very large datasets? When I started using 64-bit processors (Alpha and PA-RISC), it was in a scientific environment, and we used those 64-bits happily to memory map very large datasets and then allow the VM systems to deal with paging them in and out efficiently.



    That need still exists today as in genomic and protemic research, fluid dynamics, very large databases - a data warehouse running on 64-bit architectures performs much, much better than one using 32-bits; that's what all the major server processors are 64-bit, and why Intel/MedicoreSoft are investing so much in Itanium and 64-bit windoze. They have to in order to really get into the enterprise server market.



    There is however another area where we deal with very large datasets: Video.



    64-bits allows you to memory map a video file, even one at very high resolution, like for instance all the pieces of LOTR that needed to be composited using Shake.



    It makes writing the code for doing this much easier, it makes the whole process of dealing with these large files much more efficient because is done as a fundamental part of the OS via the VM system which is highly tuned to be efficient.



    Isn't it interesting that the applications Apple has been focusing on for some time will mostly benefit from 64-bit addressing...



    A Final Cut Pro, Shake, Logic Audio even iMovie using 64-bits will be simpler to write and maintain, and respond faster. That means better adoption of their apps, lower cost to produce and maintain, better margins/lower prices etc.



    That, apart from speed, is why the G5 is important to Apple. It allows them to really go after the video market with both 64-bit desktops and 64-bit XServes. A rack full of G5 XServes with a couple'a XRAID arrays makes a pretty compelling solution in the film and video market. Hmmm, a single rack with all the compositing and storage power for a feature length movie...



    The Mad Kiwi
  • Reply 9 of 59
    bungebunge Posts: 7,329member
    Quote:

    Originally posted by kiwi-in-dc

    There is however another area where we deal with very large datasets: Video.



    And it could help when dealing with 12-bit per channel (rather than 8-bit) video as well.
  • Reply 10 of 59
    programmerprogrammer Posts: 3,455member
    Quote:

    Originally posted by bunge

    And it could help when dealing with 12-bit per channel (rather than 8-bit) video as well.



    More bits per pixel just increases the memory footprint & bandwidth requirements. Processing will still be done using AltiVec which can handle up to 32-bits per channel just fine.
  • Reply 11 of 59
    yevgenyyevgeny Posts: 1,148member
    Quote:

    Originally posted by macnn sux

    so that's it!!!



    we're here? and 64 bit only means it can adress more ram??





    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?




    Having 64 bit be present means that it is easier to program applications that require lots of memory, not "more efficient". You can get some speed increase if you were using 64 bit integers on a 32bit machine and then started on working on a 64 bit platform.



    ...Speaking as someone who is writing some code to access more than 4GB of Ram on intel x86... it would be nice for 64 bit to be everywhere. In the meantime, oh well, I will employ any of a number of well known ways of accessing lots of memory on a machine with less than lots of memory.
  • Reply 12 of 59
    Quote:

    Originally posted by macnn sux



    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?




    I hear if you use this super-secret carbuerator in your car, it'll automagically get 100 mpg.
  • Reply 13 of 59
    programmerprogrammer Posts: 3,455member
    Quote:

    Originally posted by macnn sux

    so that's it!!!



    we're here? and 64 bit only means it can adress more ram??





    Pretty much -- that's why you hear so many people saying that most users don't need 64-bit hardware. It doesn't give you much besides the ability to use far more RAM than most consumer apps currently need. How many apps currently use anything close to 4 GB? Only a handful. A noteable handful to be sure, but only a handful. And virtually none are consumer apps.



    Quote:



    you mean developers/programmers can't program/code their apps to be more efficent on a 64bit machine??!?





    Here you have to be more careful in what you are saying. Programs are optimized for particular processors so there is plenty of optimization that can be done to get the most out of the PowerPC 970, but it is the same optimization work regardless of whether the software is a 32-bit or 64-bit app.



    In a few cases there is particular advantage to be in 64-bit, but by-and-large given the same app in 32-bit and 64-bit versions running on exactly the same G5, the 32-bit version should run a little faster. The difference will be very minor, however.
  • Reply 14 of 59
    okay guys. i am starting to get it better.





    so i guess Office and Quark will probably not change from top end G4 to the new G5s....



    But my Photoshop 8, 9 and the ability to run multiple apps and keep them open all the time is what the G5 is getting me.





    i guess i can live with that.
  • Reply 15 of 59
    Quote:

    Originally posted by macnn sux

    so i guess Office and Quark will probably not change from top end G4 to the new G5s....



    Oh yes.. The G5 should be a bit Snappy? compared to the G4 at the same Mhz number.



    But uhmm... Maybe Quark wont get any advantage.. But thats just Quarks way... (Mhz * Ram)^2 = Dayliy Errors!
  • Reply 16 of 59
    the swanthe swan Posts: 82member
    Quote:

    64-bit addressing has immediate benefits for every user of newer computers. A 32-bit computer can directly address only 4 GB. As hard disks grow exponentially in capacity, vendors of 32-bit OSes have resorted to "hacks" to address them. The last published modification of MacOS 9 was such a "hack." A 64-bit computer can directly address the sectors on large hard disks without resorting to hacks.



    Not really sure what you're talking about here. Although I can't speak for Windows XP, MacOS 7 could use a 200GB disk. Unlike the windows world MacOS uses dynamic block sizes, Windows machines have a fixed block size, which is why they ran into a size wall on their hard drives. As has been alluded to as well, 32-bit computers can process 64-bit numbers, you just have do some extra programming. And this programming isn't all that hard, did it freshman year in college. Also, if you've got a 32-bit filesystem, its going to act the same way no matter what processor you run it on. I believe most systems now are 32-bit, certainly HFS+ is.



    Justin
  • Reply 17 of 59
    mr. memr. me Posts: 3,221member
    Quote:

    Originally posted by The Swan

    Not really sure what you're talking about here. Although I can't speak for Windows XP, MacOS 7 could use a 200GB disk....



    Justin




    Not without partitioning, it can't.
  • Reply 18 of 59
    programmerprogrammer Posts: 3,455member
    Quote:

    Originally posted by Mr. Me

    Not without partitioning, it can't.



    This is an IDE limitation. SCSI and S-ATA don't have this limitation. Earlier OS versions may have other limitations, but as of MacOS 8 or 9 I don't think there are any roadblocks to large volumes. MacOS X ought not have any limits on drive size (well, at least in the sub-terabyte range).



    The Swan's comment that 64-bit memory addressing has nothing to do with disk access is correct, however. The two are completely independent and a 32-bit machine is just as capable of accessing huge volumes as a 64-bit machine. The increased address size of a 64-bit machine is relevent only to the processor's access to memory in each application's address space.



    (There is a technique called "memory mapped files" which confuses this issue, but it is not a required technique)
  • Reply 19 of 59
    the swanthe swan Posts: 82member
    Sorry, MacOS 7 could use a 200GB SCSI volume, it would be limited to a 137GB (or is it 128?) IDE volume.



    Programner, I think when MacOS 9 came out I remember Steve on stage saying that MacOS could then handle disks as large as 8 exabytes.



    Justin
  • Reply 20 of 59
    tidristidris Posts: 214member
    Quote:

    Originally posted by Socrates

    Programs that previosuly had to do math in several stages in order to avoid loss of precision can now do it in one. Apps that previously had to use slow floating point math can now use fast integer math instead in many cases.



    The G5 is faster with floating point operations than with integer operations, so I would expect to see more use of floating point operations, not less.



    Quote:

    Originally posted by Socrates

    Photoshop will be faster for example, as will any well-written graphics or audio apps.



    The main reason Photoshop and multimedia apps are so fast on G4 and G5 machines is their extensive use of AltiVec. I don't see the 64-bit integer capability in the G5 changing that equation.



    Quote:

    Originally posted by Socrates

    I cannot see any reason why any apps would become slower - 64 bit numbers are not harder to deal with than 32 bit ones (unless you are using 32 bit hardware).



    The 64-bit integers and address pointers use twice as much memory as their 32-bit counterparts. Moving 10000 64-bit integers around takes twice as long as moving 10000 32-bit integers.
Sign In or Register to comment.