"Dual Core" a big bore

The 420 is Yonah core aka stripped down Core Solo. I benched it (Lenovo N100) along with an Inspiron 710m for a review Im doing on my D420. The scores were very similar. Also you can live fine on a 4200rpm hard drive. For basic office productivity work my Dell D420 is better than my MX7515 due to the dual cores. Its just like a graphics card. The maximum frames may be alot better on say the MX7515 but the minimum frames is alot better on a dual core. While 1.2ghz its no where as fast as the 2.6ghz AMD64 4000+ but it never gets bogged down by anything. This is comparing a 1.2ghz Core Duo w/ 4200rpm vs a 2.6ghz AMD64 w/ 7200rpm. The Pentium M 2ghz is slower than the AMD64 2.6ghz (yes I benched it) while the 2.00ghz Core Duo is alot faster than the 1.2ghz Core Duo (I think this is a given).

p.s. When ppl usually suggest getting a faster hard drive instead of a faster CPU is in cases like 1.8ghz Penium M vs 2.00ghz Pentium M (Core Duo also). The price difference is around enough to get a 7200rpm. In this case it is much smarter to get a faster hard drive. However the difference between a 2ghz Pentium M and a 2ghz Core Duo is much much bigger.

 

Wow, mamaba thanks for those scores, I didn't realize how superior the core duo really was, I definately wanna pick up a core 2 duo now. You only proved that there is a 25% improvement on most processes that are done in the tests they did. That is also only the core duo and not the core 2 duo.

 

I'm so glad that I helped you work out what you "wanna" do. Clearly you do not currently own a decent Pentium M notebook, so maybe you should upgrade.

 

Mine is decent enough it's a Dell XPS Gen 2 with a pentium M 2.0Ghz, 1GB of RAM and so on.

 

Let's keep this argument under control in here - there are a lot of different opinions (facts too) floating around, and I would appreciate if everyone was respectful. If someone's wrong or you disagree, then point it out in a polite manner.

 

Just throwing this question in the air (I don't want to start a new thread for one small question); how well does Vista run on a system like the one in my signature?

 

Yes. The GPU is on the minimum requirements, so you might need to cut down a little on the aero glass, but otherwise you're very much fine. Granted, the HDD could be larger. You can check things comfortably running Microsoft's Vista Upgrade advisor.

 

You wont be able to run Aero with that since it is not DX9 compliant.

 

The Radeon 9000 uses DX9. Will it run Aero at an acceptable pace, or will I likely be forced to lower the aesthetics a little?

 

Well, time for me to poke my head back in, I guess... (Been busy for the last 2 days)

Looks like there's a lot to reply to...

Yes it does. The fact that it *requires* something special of programmers means that the hardware design isn't ideal. It meansthat what they're doing now isn't as good as solution as what they've been doing for the last 30 years.

Of course it is. But here's my point. The Core Solo architecture is *also* a vast improvement over the Pentium architectures in terms of work/cycle (IPC) and power consumption.

Sure, they made a dualcore version as well, which offers more raw performance, but the point is it only adds more performance for *some* tasks. That is why I say it's a second-rate solution. A *good* improvement is one that benefits any task at any time. And one that only works sometimes, and only if the programmer explicitly programs for it is far inferior

Wrong. You are not one to accuse Mamba of lacking technical knowledge at all, because *you* are the one who doesn't have the faintest clue about how CPU's work.

Do you think the Core Solo (the single-core version) is better than P4 because of miniturization? Or because they pump more power into the processor? That's obviously not true. There are 65nm P4's, and the P4 gets vastly more power thrown at it.
Minituarization is a gradual process that allows them to cram more transistors into the chip. That is *exactly* what dualcore chips make use of.
They are an admittal that we can't improve performance by improving our processor design. So instead we miniturize, and then duplicate what we've got already. And as you just said miniturization is not as good a solution as actually improving the design.

Core Solo, on the other hand, and Athlon 64 for that matter, are so powerful because they improved the CPU architecture itself, without relying on miniturization and duplicating your existing design because you can think of nothing better to do. Pentium 4 too, really, although their "improvements" turned out to have some faults.

What I think you might have missed is the huge explosion in CPU performance over the last 20 years. That is not due to miniturization. Of course that helped (by allowing more transistors which can be used to implement more advanced components), but the main improvement is simply that todays' CPU's are just designed vastly better than 20 years ago. CPU pipelining and out of order execution are the two main contributors to this, but under the hood there have been vast improvements every single generation since, say, the 286. I challenge you to Google for some block diagrams showing the designs of the 286 compared to a Core Duo. And then try to tell me that the only difference is "smaller transistors and more power".

Now look at Core Solo vs Core Duo.
There, the difference is *exactly* smaller transistors and more power.

Been there, done that. Singlethreaded is actually the recommended way to listen for GUI input while handling network data. It is *the standard* way of doing it, at least in C/C++. I know Java wants you to use a thread per network connection, and that scales like crap. It works ok for <5 connections, but after that, it becomes a joke. Even Java programs use a single thread for this when scalablilty is required. And more software is written in C/C++ than Java anyway.
You just use select() to monitor network data (on linux it can also monitor the keyboard, but otherwise it can easily be made to exit when specific external events occur, such as GUI input)

Other than that, I don't know where to start.
Multithreaded software is not the next logical step in programming.
It's already being used (so it's not a "next step" at all), but it's being used where it makes sense, given our programming model that isn't suited for this at all. But even so, for some tasks, multithreading makes things a lot easier. But not all tasks.
However, if we are to take advantage of multicore CPU's, there are so many better ways to do it. Just writing software using imperative languages and a small number of separate threads, each of which are limited to one core is not going to work for long. What when we switch to quad core CPU's? Do software developers have to rewrite their software again? "Octocore? We only made our program use four threads, so now we have to rewrite it again? Sixteen-core? Hell, we can't do that. Sorry."
It doesn't scale at all. It's a horrible hack.
For the last 30 years we've had much better programming paradigms that *could* scale to use multiple CPU's almost trivially. Google does it. They have written their own functional language because those can be parallelized almost trivially. Then they bought a few thousand obsolete PC's, and made them run their distributed programs. *That* is scalability and multi-core processing. They can use thousands of cores without breaking a sweat.
Researches have used another paradigm which I guess you could call "process-oriented" instead of object-oriented programming. And it allows them to think in terms of tasks. A task is just a bunch of work I need done. I don't care exactly when it's done, and I don't care *where* it is executed. So just grab a handy core, and tell it to do my task and send me the result. That too allows us to scale gracefully to any number of cores.
But "multithreading"? Anyone who's done it knows how crippled it is. It's ok for "simulating" parallel execution on a singlecore system. And it more or less lets us take advantage of dualcore and maybe, if we stretch ourselves, quad-core as well. But that's more or less as far as it goes.

I think I already covered the hardware side. Yes, when all superior solutions have been exhausted, multicore *is* a next logical step. But that doesn't make it any more elegant. And it does require some major changes to how we develop software. And the idea of multithreading is not going to cut it.

Nope, it really isn't. Because if you had seperate threads for network and GUI, you have to be careful as hell to avoid race conditions and deadlocks and other threading "issues". It might seem simple to begin with. But then you can spend the next eternity debugging subtle errors that come and go as they please. But the singlethreaded version is just as easy to implement, and it works from day one.

Uh, no?
Intel hasn't thought that.
Intel has not made a chip that made it harder for programmers to implement it.
Multithreading is not "really becoming popular and neccesary". Sometimes it's more convenient to use multithreading, but it is not widely used at all, and if it is neccesary, it is only because Intel could not come up with further core-level improvements. So talk about putting the carriage before the horse. All this is prompted by intel (and other manufacturers) running out of ideas, and switching to multicore chips. It's not the other way around. It's not Intel deciding "Well, we could make cores that are twice as fast, but people seem to want multiple cores instead".

No and no. It is an indication that there is no *need* for the last 95%. You say Winamp is not written to "take advantage of the hardware yet"?
I say it is written exactly to take advantage of the hardware. It takes what it needs, and nothing more.
You sound like it would be a good thing if Winamp used 100% CPU time.
No clue where you got "hardware incompetency" or "compromised design" from though. I am simply pointing out that for most tasks, users don't *need* multiple cores. They hardly even need *one* core.

What idle time? What blocking?

Huh? So we don't need to worry about writing multithreaded code now? That's nice, when did that happen? Is that a Vista feature?
That responsibility is on the programmers, and the OS can't just take it over.
Of course, the programmer doesn't have to say "I want this function to be executed on core 1, and this to be executed on core 2. When core 2 is idle, it should switch to this function, and then ...."
But the programmer still has to make sure that there are at all times enough available threads to fill the available cores.

Exactly. That is what Adobe claimed in their article, and it is often true. As you say, you are a heavy user. You use tons of applications at the same time. You multitask like mad. And you have no need for more than one twentieth of your CPU. That suggests that we should start making quarter-core systems, rather than quad.
And yes, I/O is often the fundamental bottleneck. And adding more cores is not going to improve on that, at all...

Nah, if anything, it lowers your overall system performance (but not so much it's noticeable)
When you set cpu affinity, you take a lot of freedom away from the OS. You force it to put this thread on that core, regardless of what would otherwise make sense to do. The OS can execute programs most efficiently when it can freely pick the cores that make sense at the time. The game in particular is suddenly restricted from running on the other core, which may actually slow it down a little bit.
But it doesn't make any noticeable difference. Use cpu affinity when it's required to make the game work properly.

There's a big difference between knowing the limitations of something, and being against it.
I suspect Mamba is with me on this. We have both used multicore systems, we have both written multithreaded software. We are not against it, we just know it's not the perfect solution, and it's not elegant compared to everything else that has been done in PC hardware over the last 30 years.

Hmm, google for some of the following:
CSP (Communicating Sequential Processes) is what number-crunching researchers are in love with. It's what I described above as "process-oriented programming". It's also quite a bit similar to what the Cell cpu suggests.
Functional programming languages are gaining ground as well (and actually, functional ideas are starting to penetrate even into C++).
Google use a homemade functional language. They have the nice property that they are "pure". There are no sideeffects. If I call a function, I can be sure that *nothing* else changes, it just returns some data to me. And it doesn't need to read anything other than the arguments I send to it either. That means I no longer have to worry about which core it's executed on. Or even which CPU. Or which PC. Because it doesn't have to access anything else from my local memory, and it doesn't try to write anything to its local memory either.

CSP is still very much a researchers toy, with limited practical usage. Functional programming has been used a good deal (I could name a few console games that are written with one of these languages), but it's still something that only computer scientists really know about (roughly speaking). It isn't used widely in the industry, and it is being held back by lack of industry support

Nah, the program isn't really responsible for picking which core to run on. That's what CPU affinity does (and yes, a program *can* specify that itself, but if anything, that is bad coding, because it *forces* the OS to put it on one single core, which means no matter what, it can never take advantage of a multicore system. Prime95 does exactly as programs are supposed to. It allows the OS to pick the core that makes sense to use. (The cache thing is a drawback here, but it's unavoidable. If we use more than one core, we're going to have some cache coherency problems slightly slowing us down. But it shouldn't account for more than 2% performance at most. Probably not too critical)

Actually, it's not too bad. It does its job about as well as the Linux one. They have different priorities and design goals (the Windows one is tweaked to fit desktop users who run *one* main application, and then a bunch of lesser stuff in the background. Linux is tweaked for overall throughput on servers, which means that the "main" application gets slightly less CPU time, but all the stuff in the background is executed more often and more reliably. Neither are bad decisions, it just depends on the usage pattern you're targetting.

I fail to see the relevance. True, you can't easily buy older processors, and you can no longer buy many singlecore processors at all, and yes, dualcore processors are pretty cheap by now. And so what? How on earth does this relate to what computing needs the average user has? Or to whether or not multithreading/core is a good solution.

But a Jess said, what about Celerons? They don't cost the $300 you spent on a dualcore, and they offer enough performance for most people.

 

Yes, what he said!

 

Radeon 9000 does no support DX9. I dont think even 9250 is fully compliant. Anyway Aero is like a hit or miss. You either run it or you cant. Im currently running Aero on a GMA950 and it runs it really well. If your card got two things: 1) DX9 compatibility and I believe Shader Model 2 support, 2) WDDM driver you should be able to run Aero with no problems.

Its a little wierd with the Aero Glass part since I havent had a case where I can run Aero but not Aero glass. There were problems back when I installed Vista in beta but not anymore in RTM (driver issues).

 

the radeon 9000 is based of the 8500, wich is a directx 8.1 card and does not suppoet shader model 2.0, so no, it will not run areo

I wouldnt run areo on a laptop anyway, its eye candy only, and it will increase your systems heat and lower battery life, it looks neat, but its not very practical in a laptop as your 3d hardware will be engaged nonstop, draining power and generating heat, how much heat and power depends on the card, but basically the bottom rung cards will be effected the most, as they will be pushed closer to their limits from the get go

 

So, because I write long posts, I'm a jerk?

As for my code? You only need to ask, if you want to see some examples. Actually, it isn't bad at all. As you might have guessed, I'm thorough, which actually isn't a bad thing when you're coding.

I guess you could say that yes, I write awfully long posts. I do that because I don't want any half solutions. If I participate in a discussion I want to make sure my side is represented properly. That each "opposing" claim is examined and considered, and that each of my claims are supported by more than just "Because I say so".

For the rest of your post, I find it interesting that you yourself say you don't know what we're talking about, and still you're able to judge which ones of us are right, and who are just arguing because they can't stand when people disagree. And you also feel qualified to judge people's intelligence and education level.

Fascinating...

I'm also curious, *did* you read my long meandering post? If you did, it's a bit silly to ask who can stand to read it. And if you didn't, it only makes it more impressive how you can judge me without reading my post.

Cheers,

 

software always progresses more slowly than hardware, thats how its always been and always will be, software is written to the lowest common denominator, as that gives developers the largest consumer base. look at video cards, dx8 cards were into the refresh of the 2nd generation before there was any real application that took advantage of them, all you had were benchmark programs and a few poorly designed quickly thrown together games, single core is dead, if your inthe market for a new pc, you shouldnt even consider it in my opinion, and even if you dont use applications that take advantage of it, the multitasking advantages make it more than worth it, the ability to encode a large video file, run a virus scan,playback video files, run your browser, and workin photoshop all at the same time with no slowdown or lag is well worth it to me, its something I could never do on any of my single core systems, and I do it now everyday on my A64x2 desktop and my core2 duo laptop.

If your single core setup works fine for you right now, thats fine, keep it, but if your in the market for a new machine, going single core makes about as much sense as getting an agp based system

 

Well, since you stand by your original post, and felt obliged to make a post *only* to say that, perhaps you would care to elaborate on your accusations?

Where were I guilty of "flaming and hostility"?
I think that if you really want to accuse me of that, you ought to back it up with some kind of evidence. it shouldn't be difficult because I'm such a big jerk, and I write so long posts, so there should be plenty of material to pick from.

Therefore, give me an example of where I flamed and were too hostile.
Otherwise, I suggest you retract your accusation, because then it is nothing more than a personal attack, or a flame.

Also, since you like discussing the theory of logic and argumentation so much, perhaps you would care to explain how the following are in any way relevant to the debate?
- Me being a jerk
- Me having insecurity issues
- The length of my posts
- The quality of my code
- Mamba's use of fancy words
- Which thesaurus or dictionary he uses
- The most accurate representation of his intelligence

I'm also curious how you "can tell" that Mamba doesn't know anything. Because despite your mastery of logic and debate theory, you completely forgot to provide any kind of reasoning behind this impressive discovery.

And if you can't elaborate on these points, your post is nothing more than a flame. Needless, unfounded personal attacks with no other purpose than to insult.
If there is a purpose to it, then perhaps you'd better elaborate a bit.

I'm impressed. You encode large video files *while* virus scanning *while* playing video files *while browsing the web *while* working in photoshop?
And you do that every single day?

Yes, you definitely need a dualcore system, no doubt about that. But not everyone does all of the above every day...

 

OK, this discussion is devolving rather quickly. Please avoid making personal insults as they do nothing to further the thread's value. As you can see, a number of inappropriate remarks have had to be removed.

If you cannot stay on-topic, please refrain from posting.

Thank you.

 

more folks will take advatage of multitasking when they know they can, I got my little bro a laptop with a CD 1.6ghz, and he intially still used it the same way he did with his p4 based desktop, once he finally relized that he could run multiple cpu intesnive programs without a hitch, he does so routinely as well