1. Welcome to NikonCafe.com—a friendly Nikon camera & photography discussion forum!

    If you are thinking of buying a camera or need help with your photos, you will find our forum members full of advice! Click here to join for free!

Dual core processor and Nikon Capture

Discussion in 'General Technical Discussion' started by Kevin Scott, Dec 16, 2005.

  1. Is anyone using Nikon Capture on an Intel dual core processor? I just got my Christmas present early (computer) and it has this new technology. I've been holding off shooting raw until I felt my computer was more capable. I've only seen one comment elsewhere concerning Capture on a dual core and it wasn't exactly positive.

    Would like to take the plunge but would prefer to hear from others first. The machine came with 1gig of RAM but I'm sure I'll be upgrading that to 2gigs soon enough. I also got a 21" LCD flat panel so hopefully the palettes will finally be out of the way in PS!! :wink:
  2. Nice. I would get that bugger more ram for sure.

    I too wonder if NC or CS2 for that matter would take advantage of 2 CPUs?
  3. Steve S

    Steve S

    Feb 1, 2005
    SE Florida
    I've got a custom built Ron R puter that runs dual Athlons, and NC works fine.
  4. Just for clarification, it's a "dual-core" processor, not a "dual processor".

  5. I am running dual AMD Athlon 246 processors along with 4GB of Corsair RAM. Capture flies using this configuration.

    You are light on your RAM and should consider at least 2GB if not more.
  6. JeffKohn


    Apr 21, 2005
    Houston, TX
    Fast CPU certainly can't hurt, and I've not heard anything about NC being any buggier on an SMP system so you shouldn't have any problems. For PS and NC you definitely want fast disks and lots of RAM, though, since that's just as important as CPU (maybe more so).

    2GB of RAM is plenty unless you're running a 64-bit OS; on 32-bit Windows, applications can only address 2GB anyway because the other 2GB is reserved by the OS (and the OS doesn't need anywhere near 2GB for itself).

    Dualcore and dual-processors are for all intents the same (unlike hyperthreading which is basically "fake" SMP, dual-core really does have two processors).
  7. Actually there are not two processors in the system.. It takes the core of the CPU and adds another core to the same chip. Most software does not take advantage of this technology as of yet. You can get a copy of Microsoft Windows 64bit edition that does take advantage of this technology.

    All other programs should run fine on this system, but will not have full advantages of super speed of the system capabilities...

  8. jgrove


    Apr 13, 2005
    NC will run fine, but i am afraid it wont take advantage of the dual core technology, maybe V5 will.
  9. Ken-L

    Ken-L Guest

    There was a problem reported about dual-core and prior versions of NC..... but there was also a simple "fix"....can't remember what that was but a search should find it, and perhaps the newest version(s) of NC don't have that problem.
  10. Thanks for all the info, guys! Guess I should just bite the bullet and do it. Never know till you try.
  11. moffo


    Oct 20, 2005
    Central TX
    Gee, I've been running it on dual-core (3.2 GHz P4) all along, and never noticed an issue. I just looked at 4.40 and it seems to balance pretty well across the two (virtual) processors. Never looked at the earlier versions.
  12. That's encouraging to hear, Michael. I may try the 30-day trial first and see how it goes. How much RAM are you running with your setup?
  13. JeffKohn


    Apr 21, 2005
    Houston, TX
    Yeah they're on the same chip but you still have two independent execution cores. Of course Intel's P4 architecture isn't very well suited to SMP in the first place which is why the AMD dual-cores perform better. Software doesn't have to be written to take advantage of dual-core, any multithreaded software will benefit (and even single-threaded software will get some benefit because there are still OS compents and background tasks executing simultaneously).
  14. HHMMM Windows XP Pro out of the box supports two processors no need for 64 bit Windows for that, not sure if that works the same for dual core, I will ask someone and figure this out.

  15. moffo


    Oct 20, 2005
    Central TX
    1) I'm running 2 GB of memory (with 2-way interleave, I think.)
    2) You don't need XP pro to use both units in a multi-core processor.
    3) You certainly don't need 64 bit *anything*, 'cause they're still 32-bit processors.

    Want to see if you have multi-core running? Just turn on windows task manager (ctrl-alt-delete will get you there quickest), choose the performance tab, and see if "CPU usage history" gives you a single panel, or two panels side-by-side. I bet plenty of people are running multithreaded and don't even know it.
  16. JeffKohn


    Apr 21, 2005
    Houston, TX
    64-bit Windows isn't required to make use of dual-core CPU's, but it is required to get fully take advantage of more than 2GB of memory or 64-bit processors (Athlon 64, for instance).

    Two CPU panels in the performance tab of task manager doesn't mean you have dual-core, you'll see the same thing on a single-core box with hyperthreading enabled.
  17. Ken-L

    Ken-L Guest

    If memory serves me....the problems reported were only for the AMD dual-core processors, and there was an easy work-around fix. Maybe a future version of NC will take advantage of the new processor technology.
  18. JKirbs


    Sep 6, 2005
    analysing architectures

    I don't quite know what you mean by "Intel's P4 architecture isn't very well suited to SMP" and i think that statement is disingenuous, and misleading if you take apart what actually the overall architecture consists.

    My main objection (but it is more subtle in the end) lies in the fact that Intel memory controllers are not part of the processor die, and so not part of the processor architecture, but the supporting chipset architecture. This is the current limiter for Xeon SMP performance. You see a direct and consistent proportional correlation between available memory bandwidth (better for the AMD) and benchmark results that are memory intensive.

    On certain small set computations, the Xeon is awesome compared to the AMD. The aerly benchmarks for AMD Opterons skewed impressions for the average user because there was a preponderance of database benchmarks, where loading the whole data set into memory is facilitiated by 64 bit addressing, and the data sets were chosen to be very big, emphasising that Intel had no competing 64bit platform in the same market space.

    However most of what we're interested in here - processing big image files, is actually memory constrained, so the AMD wins that field for now.

    If you're talking about pipelines, they're actually a lot shorter now than they were, reducing associated problems.

    I may be nitpicking, but a little discussion goes a long way.

    "P4 architecture" is not SMP capable IF you take it as a product nomenclature i.e. Intel has not made SMP - capable variants of their _processors_ under the P4 name.

    But since the P4 _execution_ cores are the same as for the Xeon processors, [certain things are enabled on the Xeon which are not on the P4, such as extra ECC, but the die prints are only rarely different] the actual _processor_ architecture is obviously SMP capable.

    I just can't tell what you're referring to by "P4 architecture" and if you mean the pipeline architeture, netburst, SSE or any particular aspect of it. There is certainly an argument that deep pipeline stages depress SMP performance because of the increased branch prediction logic complexity (burdening the die transitsor count, cost, and engineering time) plus the obvious costs of a context switch. To those of you who are lost already, the original P4 design pushed instructions "down a staircase" where each step of the staircase could perform an operation. But when you needed to start a new job, all those steps had to be swept clear, leaving much of the processor doing nothing. But, importantly, when you fill that pipeline / staircase, you can fill it very fast if it's deep, which effectively increases clock rates. [this is a very rough analogy however]

    It's of note that for recent revisions, Intel have considerably cut the depth of their pipeline, specifically to improve SMP in multi-core, and that may go further in the future.

    The way you put it that AMD dual - cores have an advantage

    Integrating the memory controller onto the same die is a very good idea Intel have been too slow to implement. In that respect AMD _processor_ architecture is advantageous for SMP, because of reduced latency, but I nonetheless argue that actual controller to bus bandwidth numbers count for so much more of the performance diference as of now (and immediate future) that you cannot say one processor architecture has an inherent advantage over the other, when theoretical peak performences are not attanable due to lack of memory bandwidth.

    Memory bandwidth gets to be even more of a constraining issue in dual - core because presently each core can only talk to the outside world through the one single memory bus. This again gives AMD the jump, but because of the faster bus, not yet because of architecture. Worse, for Intel, their memory bandwidth is not direct between processors, and shared with RAM and PCI busses. That one really needs fixing, grrr!

    Core means execution core. Integer or Floating point. (SSE is really a instruction "firmware" to implement better algorithms for VLIW (Very Long Instruction Word) requests and is there because detailed instructions better describe many alorithms in code. Things like memory controllers are attachments, and can be delivered on the same piece of silicon, or not. It all comes down to space available and related cost. Look at the SUN Niagara processor. 6 or 8 cores, each with their own memory controller. But the downside - only one underpowered FPU, so many applications will be sorely hurt. But all that memory bandwidth (many many times what HyperTransport can deliver at present) allows some applications to fly.

    Another nitpick might be that AMDs memory architecture is actually specified by the HyperTransport consortium, which is independant of AMD, who are a founding member.So you can't _strictly_ talk about "AMD arhitecture" [sic] when it comes to memory capabilities :)  I had this all explained to me once by the founder of that consortium, before which to be frank, i was a bit lost as to what was specifically so good about AMD boxes.

    So, the real reason why AMD do better in multi-threaded SMP benchmarks and applications is they simply have much more memory bandwidth available to keep two processors talking on the same job. This chit chat between processors during a job that takes advantage of two processors, is a very real bottleneck.

    Also, with dual-core processors, currently both cores share the same memory bus to talk to other processors. This restricts the performance of the second core, unless you specifically send that core a job where all the required data can load in a single thread and be stored in cache memory. Notice the 2*2MB caches Intel uses on its dual core Xeons. That's expensive cache. It's partly there because they're still stuck with relatively slow memory busses. The cache size is also related to the deep Intel pipelines, because to keep those 'lines filled without too many wait states, you need to grab lots of data really fast. Cache takes up lots of space, generates lots of heat (more transistors used proportionally in memory than in a computational unit - take a feel of a recent laptop any day, and i assure you one of the biggest hotspots is where your RAM is located), and this is likely a good reason, in addition to memory bus redesign needs, why Intel does not yet integrate memory controllers.

    Finally: "Software doesn't have to be written to take advantage of dual-core, any multithreaded software will benefit".

    Not in the immediate future for your desktop, anyway. This will quite possibly change sooner than you or the authors think.

    It's the OS kernel thread scheduler that distributes the work over the processors. However, for massively multiprocessor or many core applications, you may find you need to talk more intelligently about the nature of the work to assist the OS kernel scheduler. SUN requires a recompile of code for their Niagara boxes, but though i've not gone so deep into their documentation yet, there are a lot of kernel hooks available to help manage and balance code execution across cores.

    Practically, you are quite correct, but if the trend towards multicore / bigger SMP accelerates, some rethinking will be necessary. My apologies if this is arcane, i just spent a few weeks reading up about the SUN Niagara architecture for my company (to see if we want some of that) and at 8 cores, with 32 visible virtual CPUs it's actually very important to know what your code is doing before you assume it can be speeded by a Niagara box. Get it wrong, and performence actually slows to treacle.

    There, that should have bored everyone already! But forgive me, if you can - there's so much confusion around over dual core and AMD vs Intel, that i gave a shot at illuminating some of the dimmer recesses :) 


    - kirbs
  19. MontyDog


    Jan 30, 2005
    #1064 - You have an error in your SQL syntax;
  20. JKirbs


    Sep 6, 2005
    But i do believe you need XP Pro to run more than one actual processor, dual - core or otherwise.

    If you enable Intel HyperThreading in BIOS, you'll see your apparent cores double again in the performance manager. So a dual dual - core Intel box will appear to have 8 processors!

    Dell messed up their PSUs when i was about to order a dual dual- core Xeon. Raving idiots. They ship a 550W PSU as standard, and when i was placing the order i mentally worked out a nearly 600W drain from major components. A day later, they withdrew the systems, so I have a dual single core Xeon on my desk (and a very welcome, ahem, additional, discount negotiated on account of their cockup!).

    I'm totally convinced that learning the performance vaguaries of all these architetures is much more expensive in time than the cost of the very highest end box i specified for my company. That's something to bear in mind if you're a photog working on your own or plan to own less than a *lot* of machines - it may be cheaper and even advantageous for you to simply buy the highest end gear you can afford, and not waste any time investigating or reading long posts from me! :) 

    Happy Christmas Everyone!

    - kirbs

    p.s. tip to everyone, never, ever call Dell and say you're a private user. Always give a company name, and tell them you're a heavy user in the graphics / DCC business, dependant on your machine for work. I have never failed to get excellent discounts from their internet pricing, even on individual box orders. And no we're not by any means a big company, just a tiny shop that needs fast fast kit. Once that's factored in I cannot build a comparable myself for cheaper (especially so if a high end _professionally certified_ graphics card is included because those always retail at 3 times the OEM price) Consumer reps for Dell cannot negotiate with you. If you buy at the end of the quarter, the business sales staff and the company actively want to make quotas - that's by far the best time to buy.
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.