Thread: Upgrading to the new processors (more of a software ?)

Originally Posted by madmax2006

So basically, I'm just wondering - why can't 1 program be run through all of the cores, so it gets executed fast(er)?

Your program probably isn't multi-threaded.

Originally Posted by madmax2006

I was reading something on-line (so I have to take this with a grain of salt), and the guy was saying that moving to the new physical (4-core) logical (8-core), was like getting twice as many people pregnant to have a baby faster - it's just not going to work.

I don't know what the context of that statement was, but it seems plain wrong.

One can argue if an application isn't developed to take advantage of multicore processors the application isn't going to use more than 1 core for most of its operations. Many applications are developed just like this. However, the comparison the author of that argument uses doesn't apply, because the application is... monogamous.

If, on the other hand an application is developed to take advantage of multicore processors, it will employ multithreading. Assuming it is moderately well developed, the application will indeed run faster. There's no way around this argument simply because it has been proven over and over again for the past decades, from large scale enterprise applications to games and other consumer level applications. So, in this context, the application is indeed making babies faster. The problem with the author argument is that he crafted his analogy in a way that doesn't give room for this possibility (because fetuses don't grow faster if someone impregnate more people). You have to be more observant of the arguments some people use and shout Fallacy! when these type of tricks are used.

There are some intricacies to multithreading and multicore processors, however. An application may be multithreading but not take full advantage of the technology. But these are issues that fall outside the processor real (real!) ability to provide programmers with a multithreading platform to work with. You cannot blame the processor on a task that is the sole responsibility of the programmer. It would be sort of like blaming drunken car accidents on the cars, instead of the drivers.

One thing I did notice - is that when I build my next computer, I may as well just build one with the new hardware, all of the core 2 crap is hardly being bought. Not to mention I would like to play some games on the pc (free DLC).

There's no "core 2 crap" as you put it. The responsibility of taking advantage of the 2 cores is that of the programmer. Not of the processor. And not all applications need, neither all programmers want, to take advantage of multithreading at all times. Most of the times in fact, they rightfully say no to multithreading. Simply because it's not needed.

And just because an application takes the full 100% of a single core at some time, doesn't mean that application requires multithreading either, or that the costs involved in applying multithreading will outweigh any benefits for both the programmer and the user of the application.

"Running faster" is not always an observable feature that can justify any measures. How exactly do you observe a difference of .2 seconds when you click two buttons that display a window? From the point of view of the user of that application the .2 second delay in displaying the second window is never observed. This easily translates to the programmer as "don't bother with it". They can still bother, but what for exactly?

There may be circumstances that require a clinical eye into deciding whether or not multithreading will be necessary, like lengthy sequential operations that require the user to wait or the system to become unresponsive until they are finished. Put into context however, what most programmers come to realize is that a large number of these situations don't require, or cannot have, multithreading. Let the user wait because it's acceptable they do. On other situations multithreading is just obvious and is already established, like game development or weather prediction. These situations do not stop the user from waiting, they just move the threshold further high. But do not be mistaken, there are also a lot of applications that use multithreading and that offer no real advantage at all, where on many modern systems they would still fly on one core.

I don't know how many people keep up with hardware "stuff", because I used to - just in the way of knowing how to build them, not how the "stuff" worked.

I don't. Shortly after the Pentium I slowly started to disconnect myself from hardware news and keeping an eye on its developments. Kind lost interest in it and shifted instead to a more mundane and practical approach to computers and programming. As you keep learning about programming and accumulate knowledge, the "how things work" side of the question will get richer and richer. It will be your choice if "how things are and how they will become" is something you want to keep an eye out. Personally, I don't.

I echo all that which Mario F. stated.

In addition:

madmax2006, You seem to prefer clock speed/rate over cores. In reality, There is so much more that contributes to your program's running speed than the processor's clock rate. You might say, "I'll take the single core Pentium 4 with the 3 GHz clock rate over the quad core Core i7 with the 2 GHz clock rate, because I think it will run faster!", but in reality the Core i7 probably has a better bus speed, can handle working with faster memory, has a larger and faster cache, and has better optimizations in relation to pipelining and branch prediction...not to mention the 3 extra cores. (Please note that I did not actually look up processor specs for that example).

Truthfully...the processor's clock rate is really only one of many factors that combine to form the overall performance of a processor...and that's true for the amount of cores the processor has as well.

Originally Posted by C_ntua

Finally, performance is one factor for a "good" CPU. Energy spent is another one. Heat generated can be another one as well if you plan to overclock, which would be easier in a low clock rate multi-core CPU (I believe) than a high clock rate single core one.

If a high clock rate is the aim of the overclock (As it often is), then a single-core is the best way to start out. If overall higher performance is the aim, then the question becomes whether or not the target app is multithreaded, overclocking doesn't change much really.

All overclocking records in place have been set with single-core processors or multi-core processors with only one active core. Good results with absolute clock rate or relative clock rate (relative to the stock clock) are not suitable on multi-core CPUs.

Edit: Each CPU core is very different when it comes to overclocking capabilities. In a quad-core for example, one of the cores might not be able to reach more than 3.3 Ghz, while the other three would almost do 4.0, this limits the entire CPU to 3.3Ghz. The best way to go is to get the best possible single core. Multi-cores are simply more unstable when overclocking due to the difference in the cores.

Multi-cores are usually more overclockable than single cores because they really are underclocked already to fit the TDP.

I had thought about starting a thread asking how many cores are necessary before it becomes superflous for pretty much everybody, including gamers/power-users. I suppose the question is not entirely off-topic in this thread.

'Twas reading about Intel/AMD releasing 6-12 core CPUs in the near future. Is that overkill? When will it end?

I'll be sticking with my P4 single-core for at least another 2-4 years.

Of course it's not overkill. And it will never end. Not in the near future, anyway.
More demanding calculations, games and programs require more beefy processors. Think 8 cores is overkill? Think again. Video encoding can bring even the newest, best and fastest processors to their knees.
That said, the number of cores you need is proportional to what you want to be doing.
However, I would recommend everyone to at least use dual core. It's more ergonomic and makes your system more responsive. Dump that crappy P4 and get a real processor made in this century.

Thinking about the P4 makes me think about the Atom and the Atom makes me puke.

It seems like processors have reached their top speed at about 3GHz, so the only thing left for them to do is add more cores. I've got a quad-core and my total CPU usage rarely goes past 10-15%. Although I don't do any gaming or other really CPU intensive things.

I do want a multi-core machine, but I can't find a reason to get one.

This P4 exceeds my needs. I just did the most extreme thing ever on this machine: ran antivirus, played a DVD in the optical drive, and surfed the net/did office work all at the same time. There were one or two moments where the DVD sound was briefly jerky, but otherwise, it all ran to my satisfaction.

I had recently upgraded the CPU from a Celeron, and with the Celeron, when I was running the AV, the computer virtually became useless for everything else. Even internet browsing was almost impossible with just the AVG running.

But damn, I do drool over the multi-cores.

I do want a multi-core machine, but I can't find a reason to get one.

Hmm... that's an odd statement. There's no reason to upgrade if there is no reason to upgrade. But you have a reason. You want to upgrade.

And then there's the issue of you finding no reason. I mean, you can say that all you do, you have been doing fine on your single-core. And I can understand that. One doesn't need much to program in C++. And even less if under Linux. But you can't find any reason?

- For one, you can fool around with multi-threading and hyper-threading and see the results on the correct platform. Since thread programming is an area that you will eventually grow an interest on, a single-core CPU is at a serious disadvantage for testing purposes.

- Virtualization is an excellent way to test certain type of software. Networking software or portable software, for instance. Current virtual machine clients take full advantage of multi-core processors, not to mention hardware virtualization that is not supported by any single core CPU that I know of. Trying to run a VM with a modern operating system on a single-core is already a complicated task. Much more when one is expected to debug software from there.

Those are two examples. Not to mention, you know, prices. You can get a complete dual-core system (with no monitor) for... what? $400?

Originally Posted by cpjust

It seems like processors have reached their top speed at about 3GHz, so the only thing left for them to do is add more cores. I've got a quad-core and my total CPU usage rarely goes past 10-15%. Although I don't do any gaming or other really CPU intensive things.

For now. The problem is that the processor gets too hot inside where cooling cannot reach if they bump up the speeds.
But when we get cube-like processors where cooling technology intertwines it like a maze, you can be sure they will ramp up the speeds.
But for now, it's the core race.