I'm trying to clear up a misconception and I do not believe I am doing a disservice. If others believe this then I'll shutup. These are not blanket statements and if you had a background in what I am talking about you would understand that chip makers are simply trying to get you to buy more chips through clever marketing. Marketers are not engineers. I do believe credentials serve a purpose here because I've worked on these types of chips and can tell you the thought process behind them.
Proof is in the pudding and not some theorectical design discussion. I have done the measurements, did the tests and they conclusively in my case prove that I am benefitting. In that regard you cannot speak for me that i am not benefitting and am falling into some marketing hype. You are the one hyping here that it has no benefit but sorry you cannot make that statement without knowing a) what I am running and b) what my measurements, adjustments and subsequent test results are. Without measuring you are talking out of your nose. I am done Sherlock
One thing to consider for those investing in a new PC and thinking of using Microsoft Vista (aka Longhorn) is that it seems it will have some fairly hefty hardware requirements. Partly I think because the GUI stuff will be vector graphics based (rather than bitmap) MS are suggesting at least 256 Mbytes graphics card and 1Gb main memory I think. For multi monitor setups who knows what the situation will be like but the idea of several high end graphics cards heating up the room doesn't sound particularly appealing. Dual core will probably help but how much so is at the moment, unknowable. I'll be sticking with Linux so for me the issue is academic.
If so, few folks running multimons will upgrade. The half-a-gig memory cards cost ~$700 and there are only two models that I can think of. Either that's incorrect, or Longhorn is a ways away, or MS is nuts. Or all three. Man, to run dual mons, you'd have to upgrade to 2 gig RAM and a $700 video card (cuz I don't think there are any 256 meg PCI video cards, but I do recall reading something about some mainboard with dual AGP slots). According to Microsoft, any "PC that meets current designed for Windows XP logo requirements, has a mainstream or performance class CPU, 512MB of RAM, and a discreet graphics subsystem that will support the new Longhorn Display Driver Model (LDDM) will run [Windows Vista] very well."
Hi All, Time to break in. I essentially tend to agree with Mike's view. (1) Multiprocessing operating system. Popular OS's for years all support multiprocessing (don't know about MAC). Having a multiprocessing computer and an OS supporting it DOESN'T MEAN that your OS is going to distribute your load as you dream it will. In my experience IT DOES NOT. I abandoned since about 1999 my P3 XEON multiprocessing machines for this reason. I have been carefully watching hyperthreading: same story. (2) Load balancing by virtue of the operating system is very effective in server machines. Why? Because you basically run mainly one and the same program: your OS. That's all. The OS itself is designed to distribute its load over many processors as required. (3) Workstation environments are much different. Of course you run your OS but if you do any serious computing, the cpu load of the OS itself is minor. Given the enormous diversity of programs run by users, OS writers cannot really do much in making sure a given user with a given application mix will neatly load up a few processors: IT WILL NOT. (4) Of course, some applications may have been specifically designed to exploit multiprocessors. This is sometimes the case for high end professional applications. In house designed software can be developed with processor load balancing in mind. This makes programming quite a bit more difficult. (5) The fact that most of today's software makes use of threads doesn't mean that these threads will indeed be magically dispatched to an idle processor. MOST LIKELY THEY WILL NOT. You touch here upon the basic distinction between multiprogramming and multiprocessing. One doesn't imply the other. (6) It is safe to say that anything proposed about this is going to turn out much different than what people dream about it. As is known for many, many years, the only way to tell is TEST: benchmark. I did a lot in the past and always came out much ahead by sticking to powerful single processors, except in server environments. (7) Some things can be learned from large scale scientific multiprocessing based applications. These computational processes consist often of highly parallel evolving operations and software design fully takes advantage of this. This is hardly the case of a workstation loaded up with a garden variety of applications. X2/Dual Core? It's basically the answer of cpu manufacturers having run into the wall imposed by physics itself. These days, they can only put two of the same under the hood to double power. Can any change be expected in all this? I dunno. Let's wait for some benchmarking - still better, do your own and come back after you did. It's a new ball game altogether. You want more power? In truth, your favorite cpu manufacturer can't do the old trick anymore. You'll have to make it work for yourself. nononsense
>>Let's wait for some benchmarking - still better, do your own and come back after you did. It's a new ball game altogether.>> - Anyone ran some bechmarks for the X2 CPUs?
How about this one for starters: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2397&p=27 Anandtech and Tom's hardware are two of the most widely trusted tinkerer's web sites on the internet IMHO. PS: Moderator, I'm not sure if I'm allowed to post direct links, but if not, please delete it and I apologize.
Mike805 threw some serious light on this subject. I too, in a previous life was near being an EE. I did 4th gen Database applications (to an extreme) usually never seen in corporate transactions, and we moved almost every faultering client to multi-core (for public consumption: dual-core) processor machines. However, these only produced minimal improvements in the multiple operating system scenario until we configured all the configurable options both in the OS, the backplane and the Database applications to take advantage of however many processors were present. In some cases we built machines with 4-plex(s) and gave them over to the database to dominate. He is more than correct in pointing out the confusion and differences between multithread, multiprocessor / multicore and hyperthreading. For most persons, these more advanced machines, even the reference to the dual processor xeon, will show significant improvement over single cores as the basic motherboard (backplace on servers) will to some extent load balance on their own and multi-pipeline (thread) processess or more correctly "demand upon the server" to whichever pipeline or processor core is available and polling for work. Spend the money, especially if you're making it and stay on top. If you're mildly profitable then make do until you can splurge! (i.e. conserve capital) I tried all these scenarios on a CPQ AMD 64 3700 with 1.5Gb and brought the machine to many a halt and slowdown. It never overheated though. There are significant differences between the backplane on the laptops than the desktops. There is one (possibly more, but just one) vendor that fabricates their own small footprint desktop MB in a laptop diecast. Yeah, its a luggable though.
Guys, I upgraded to the 3500+ Venice core 939 AMD with a gigabyte K8NS Ultra MB and a gig or RAM running in dual channel mode. I'll give you an update after I've installed the OS and can run a few tests with my existing set up. I wil look at upgrading the chip if esignal ever optimizes their application for multithreading. Runningbear