Rosetta with Core 2 Duo

2»

Comments

  • Reply 21 of 37
    aplnubaplnub Posts: 2,605member
    Quote:

    Originally posted by Hiro





    That makes so little sense I feel dumber just reading it.




    Sucks you don't get it.



    I have two Core Duo's and they are never max'd out (at 100% use). Even in Photoshop the processors are never max'd out.
  • Reply 22 of 37
    slugheadslughead Posts: 1,169member
    Quote:

    Originally posted by adezero

    wouldn't 4 gigs of ram be overkill?



    OS X finds a way to use all the RAM you can cram in there.



    As far as I can tell, the ratio of virtual to real memory each process uses is based on how much RAM you have.. so if you have 4 gigs, each process will run faster than with 2 gigs.



    There are diminishing returns on adding more RAM, for sure, but technically speaking, adding RAM always increases performance on a Mac, no matter how much you add or how much you had before.
  • Reply 23 of 37
    slugheadslughead Posts: 1,169member
    Quote:

    Originally posted by aplnub

    Sucks you don't get it.



    I have two Core Duo's and they are never max'd out (at 100% use). Even in Photoshop the processors are never max'd out.




    I think that has more to do with rosetta's and photoshop's programming than it does with hardware..



    I think that's a response to what you said, but I can't be sure as I don't understand.



    I've yet to max out my Dual 2.5 ghz G5 at photoshop tasks.. In fact, the only process which has ever maxed out one of my processors was an infinite loop in a php script.



    I've TRIED to max it out using photoshop, but it never quite gets there.



    Even my DVD transcoder only gets to about 70% x 2.



    If you're saying that rosetta forces apps to be single-threaded, that seems unlikely... explain further, please.
  • Reply 24 of 37
    hirohiro Posts: 2,663member
    Quote:

    Originally posted by aplnub

    Sucks you don't get it.



    I have two Core Duo's and they are never max'd out (at 100% use). Even in Photoshop the processors are never max'd out.




    You are using a fully preemptive multiprogrammed operating system, and unless you went through and killed every daemon process on your box the OS will give those processes timeslices to do a little lite housekeeping. That alone is enough to keep you from maxing things out.



    If your photoshop files require streaminig RAM furiously and the filters do trivial transforms then you will wait on data, which unloads the CPU. Seeing as how that is standard for filters it is no wonder you don't "max-out" your CPUs.



    Maxing out both CPU's generally takes either bugs or purpose built software that takes into account the particular architecture of the CPU and does a high number of calculations on small amounts of data.



    Now since this is how it works and not some unintelligible bass-ackwards thought about making a CPU faster so it will max out more easily, I guess it doesn't suck that I don't get it.
  • Reply 25 of 37
    aplnubaplnub Posts: 2,605member
    Quote:

    Originally posted by Hiro

    You are using a fully preemptive multiprogrammed operating system, and unless you went through and killed every daemon process on your box the OS will give those processes timeslices to do a little lite housekeeping. That alone is enough to keep you from maxing things out.



    If your photoshop files require streaminig RAM furiously and the filters do trivial transforms then you will wait on data, which unloads the CPU. Seeing as how that is standard for filters it is no wonder you don't "max-out" your CPUs.



    Maxing out both CPU's generally takes either bugs or purpose built software that takes into account the particular architecture of the CPU and does a high number of calculations on small amounts of data.



    Now since this is how it works and not some unintelligible bass-ackwards thought about making a CPU faster so it will max out more easily, I guess it doesn't suck that I don't get it.




    For instance:



    When I open up InDesign on my:



    iMac G5-the proccessor is nearly maxed out for a few seconds

    Core Duo - the processors are not taxed over 20% or so for slightly longer period of time.



    When doing work in our 100+ page catalog and making some manipulations on my:



    iMac G5 - the processor works at 100% for a period of time

    Core Duo - the processors works at ~30% for a longer period of time.



    This tells me, in my opinion, Rosetta relies on more than just slamming a processor with data and getting a response out faster. I would say it works as fast as it can and in general terms, it won't go any faster minus work done on a per clock cycle basis.



    See my point?



    There is no need to be such a smart ass about it either. I was trying to be nice and helpful and you degrade people because they are obviously not on your mental level.
  • Reply 26 of 37
    hirohiro Posts: 2,663member
    Quote:

    Originally posted by aplnub

    For instance:



    When I open up InDesign on my:



    iMac G5-the proccessor is nearly maxed out for a few seconds

    Core Duo - the processors are not taxed over 20% or so for slightly longer period of time.



    When doing work in our 100+ page catalog and making some manipulations on my:



    iMac G5 - the processor works at 100% for a period of time

    Core Duo - the processors works at ~30% for a longer period of time.



    This tells me, in my opinion, Rosetta relies on more than just slamming a processor with data and getting a response out faster. I would say it works as fast as it can and in general terms, it won't go any faster minus work done on a per clock cycle basis.



    See my point?



    There is no need to be such a smart ass about it either. I was trying to be nice and helpful and you degrade people because they are obviously not on your mental level.




    A) I wasn't degrading YOU, I was making a commentary on your statement. Two entirely different things. When you aren't sure what you are talking about ask a question, lot's of folks will help. But talking out your arse sideways will promptly generate a pithy reply or two.



    And no I don't see your point, I see badly misinterpreted data. See my points below.



    B) You are trying to compare a single G5 against a Core Duo. A little math courtesy of Apples marketing hype show why none of this is the least bit odd. Apple says the Core Duo is ~4x faster than a G5. If your G5 was at 100% your Core Duo should be at 25% per processor. Sure this is a tad over simplified, but close enough to what you are seeing, especially if Apple marketing was stating best case. The problem isn't the chip is slow, a large part of the problem is it's a heckuva lot faster than you give it credit for.



    C) You are also seeing behavior that has EVERY appearance of being either bandwidth limited, or memory limited if you don't have at least a Gig in that box [2GB is more like what you need if you are working on catalogs that big]. If you are RAM limited you are feeling the slowdown due to excessive paging to get your data, Rosetta is famous for using large quantities of RAM as a way to cache it's translated instructions. Overall this technique is a performance win unless you don't have enough RAM in the box.
  • Reply 27 of 37
    aplnubaplnub Posts: 2,605member
    I have maxed ram in all my machines. 2 GB in everything.



    I am not saying the Core Duo's are slow, I am saying that Rosetta is more than chip power, at least from my observations.





    Quote:

    Originally posted by Hiro

    A) I wasn't degrading YOU, I was making a commentary on your statement. Two entirely different things. When you aren't sure what you are talking about ask a question, lot's of folks will help. But talking out your arse sideways will promptly generate a pithy reply or two.



    And no I don't see your point, I see badly misinterpreted data. See my points below.



    B) You are trying to compare a single G5 against a Core Duo. A little math courtesy of Apples marketing hype show why none of this is the least bit odd. Apple says the Core Duo is ~4x faster than a G5. If your G5 was at 100% your Core Duo should be at 25% per processor. Sure this is a tad over simplified, but close enough to what you are seeing, especially if Apple marketing was stating best case. The problem isn't the chip is slow, a large part of the problem is it's a heckuva lot faster than you give it credit for.



    C) You are also seeing behavior that has EVERY appearance of being either bandwidth limited, or memory limited if you don't have at least a Gig in that box [2GB is more like what you need if you are working on catalogs that big]. If you are RAM limited you are feeling the slowdown due to excessive paging to get your data, Rosetta is famous for using large quantities of RAM as a way to cache it's translated instructions. Overall this technique is a performance win unless you don't have enough RAM in the box.




  • Reply 28 of 37
    hirohiro Posts: 2,663member
    Quote:

    Originally posted by aplnub

    I have maxed ram in all my machines. 2 GB in everything.



    I am not saying the Core Duo's are slow, I am saying that Rosetta is more than chip power, at least from my observations.




    Good on the RAM. Again though I have no clue what your second statement is supposed to mean. Is it Rosetta reduces total box performance compared to native performance? Well...yeah. But this exactly what His Steveness told the world about high performance requirement apps under Rosetta compared to late G5 offerings. No news here. If you mean something else I'm totally stumped.
  • Reply 29 of 37
    1337_5l4xx0r1337_5l4xx0r Posts: 1,558member
    Hiro, if reading his first five posts doesn't clarify things for you, I doubt him explaining it again will help. What he says is pretty clear. Rosetta is not heavily threaded, and doesn't max out his Core Duo. Having more processors will not help, but he posits that having a faster processor, in GHZ speed, might.



    Let's move on.
  • Reply 30 of 37
    aplnubaplnub Posts: 2,605member
    Thanks.







    Quote:

    Originally posted by 1337_5L4Xx0R

    Hiro, if reading his first five posts doesn't clarify things for you, I doubt him explaining it again will help. What he says is pretty clear. Rosetta is not heavily threaded, and doesn't max out his Core Duo. Having more processors will not help, but he posits that having a faster processor, in GHZ speed, might.



    Let's move on.




  • Reply 31 of 37
    hirohiro Posts: 2,663member
    Quote:

    Originally posted by 1337_5L4Xx0R

    Hiro, if reading his first five posts doesn't clarify things for you, I doubt him explaining it again will help. What he says is pretty clear. Rosetta is not heavily threaded, and doesn't max out his Core Duo. Having more processors will not help, but he posits that having a faster processor, in GHZ speed, might.



    Let's move on.




    OK. Your post makes sense, but the idea behind it doesn't hold with how threading and multiple cores work. It also not what he said, although it may be what he meant as evidenced by his thanks above. What I won't just leave alone though are posts that show lack of understanding on how the underlying engineering and science work and thus make baseless statements and conclusions.



    Threads are a "good thing", but they are not magically going to make every program faster "no matter what". Two cores can only run two threads simultaneously. Period. Ever. [SMT kinda updates this but then we are talking four logical cores on 2 physical ones, which just scales the math by two proportionately.] More threads on a constant number of cores will only help if there is some sort of blocking behavior going on, too many threads that never routinely block for intelligent reasons will actually slow down an application if the thread to processor ratio is out of whack.



    I have posted a couple times the most likely reasons his Core Duo iMac is not getting "maxed out" and it has nothing to do with threads and everything to do with memory and type of processing. Making a CPU faster would actually make it LESS LIKELY he would be able to "max out" his processors because of the real limitations. He would still get done faster, but be disappointed it seems.



    And finally doing any beating around the bush to say a faster processor will run something faster is just way too much effort and wasted words. Faster IS faster! No extra misdiagnosed issues need apply.
  • Reply 32 of 37
    Hiro, you lost me when you said his processors are "too fast". Also when you said he should expect %25 CPU usage because it's 4x faster.



    Nobody buys a faster Mac so they can have their CPUs perform the same tasks at the same speed, but use less CPU time. That doesn't make any sense. If your CPU is 4X faster, you want it to complete tasks nearly 4X faster (obviously not the full 4x, since there is overhead, and bandwidth limitations, etc).



    So if a CPU is "4X faster," and instead of running Photoshop faster, it just runs it at the same speed, but uses only 25% of CPU time, then something else is limiting the speed. The problem is not that the CPU is "too fast," it's most likely due to bandwidth starvation. Although with 2 GB RAM, that shouldn't be a problem.
  • Reply 33 of 37
    gene cleangene clean Posts: 3,481member
    Quote:

    Originally posted by Junkyard Dawg

    So if a CPU is "4X faster," and instead of running Photoshop faster, it just runs it at the same speed, but uses only 25% of CPU time, then something else is limiting the speed. The problem is not that the CPU is "too fast," it's most likely due to bandwidth starvation. Although with 2 GB RAM, that shouldn't be a problem.



    You have a point, but look at it this way: if it only uses 25% of your available CPU time, then the other 75% is free to do other things, like encode a movie perhaps, or decode one. This is moot when you only want to run Photoshop, but it's handy when you want to run Photoshop and do something else that hits the CPU hard.



    It's confusing, and the advantage is there only for those that multitask a lot between CPU hog programs, but it's there.
  • Reply 34 of 37
    hirohiro Posts: 2,663member
    Dawg, he didn't say it took the same amount of time, he was just pissed he couldn't max the processors out as shown I suppose in the Activity Monitor. His stuff did get done faster unless he was comparing a dual G5 2.5 to a iMac Core Duo.
  • Reply 35 of 37
    aplnubaplnub Posts: 2,605member
    Quote:

    Originally posted by Hiro

    Dawg, he didn't say it took the same amount of time, he was just pissed he couldn't max the processors out as shown I suppose in the Activity Monitor. His stuff did get done faster unless he was comparing a dual G5 2.5 to a iMac Core Duo.



    Somehow, we still can't understand each other.



  • Reply 36 of 37
    hirohiro Posts: 2,663member
    You're right. Your comparison was to an iMac G5, not a DPG5 so what part are you missing now? Your iMac Core Duo beats an iMacG5 in just about everything short of a couple purely Altivec filters, even in apps running under Rosetta. Unless you can give actual times, not your guesses based on the processor usage there's nothing else to go on. I've yet to see a reliable actual benchmark by anyone that contradicts what I am saying. For your one computer to magically buck the trend just doesn't make the reasonable cut.



    I can see situations where you would like it to be faster, that's quite reasonable and who wouldn't. But most of the rest doesn't fit. I'm not saying you are BS'ing anyone or making things up, I just believe your knowledge of what is actually going on has some significant gaps, and that skews your subjective observations and conclusions.



    I'm not even asking you to drink any bitter Kool Aide, just pull out the stopwatch on non-Altivec heavy tasks and I think you will surprise yourself. That will make it easier to wait on the laggards over at Adobe.
  • Reply 37 of 37
    kukitokukito Posts: 113member
    Here are some Conroe Mac benchmarks that, if you believe them, would be very good news for Photoshop users. If Core and Rosetta can do this for Photoshop there wouldn't really be a need for an MS Office UB.
Sign In or Register to comment.