BF2S Forums What's so good about the i7?

GC_PaNzerFIN wrote:

rdx-fx wrote:

i7 is neat, cool, and fast.

The future of computing, however, is in the architecture of the current Graphics chips.
That's where the really cool developments in hardware are taking place.

Monolithic core CPUs with a few cores, the innovation is in the management of their complexity.
In contrast, the GPU in your video card has the real innovation in solving shit-tons of complex mathematics simultaneously, rapidly, and in bulk quantity.

you know why?

CPU needs to do massive number of different kinds of calculations. GPUs can be optimized to do certain calculations faster. Besides, graphics rendering just goes well with massive multithreading. Making software for lets say, 100 core CPU, would be VERY tricky.

The modern GPUs have instruction sets more complex than necessary even for a general purpose CPU.
The GPU is optimized for array math, but it is more than capable of the general CPU operations (move/jump/logic/math).

Making a CPU do GPU stuff, however, is truly painful.
(Learned assembly on the then state-of-the-art 6502 and 68000 processors - to give you my frame of reference.)

In short,
Modern GPUs could take over the job of a CPU.  The reverse.. not so much.

Last edited by rdx-fx (2008-12-18 14:53:56)

GR34 wrote:

So what your saying is that if NVIDIA or ATi made a CPu out of one of there 280 GFX cards or 4870's it could do a better job then a cpu?

I think he means if you swapped a CPU for a GPU, and a GPU for a CPU, then the GPU would be better suited than the CPU.


If you get that.

cospengle wrote:

rdx-fx wrote:

(metric assloads, not american assloads, BTW)

You mean 'arseloads'?

I agree with your point, but you have to realise (as I'm sure you do), that changing to a different architecture isn't as simple as programmers changing the way they do things. If you program in a different architecture then everyone who uses your program will have to buy that architecture before they buy your program. As we've seen with Windows, people don't like to change. Vista has been out for ages and most people still use XP, even though they don't need to buy a whole new computer to run it (except for 64-bit). So to write a program for an architecture that no one has isn't really going to make you much money.

It's not the architecture that's the issue.  It's not the users.  It's the programmers.

If Windows and Office ran on the new architecture, the users wouldn't really give a damn what was under the hood.





And Vista, too, is a case of user-land not wanting to change.  Not because Vista is new and different - but because Vista doesn't do much better than XP (from a user perspective), and does a great many things more poorly than XP.

A trick to getting a new architecture in place;
Make your new platform do everything that the old platform did, using the exact same userland interface and skills - then add a few bells and whistles, and do it much much more efficiently/faster/cheaper than the old architecture.

Case in point:  Every generation of video cards.  Does what the old cards did, adds a few features to get the press drooling and writing, and does it all better/faster/cheaper than the previous generation.

Another case:  MS Office and Adobe CS.  Not too hard to bounce between OS-X and XP if you use those two suites alot.

Freezer7Pro wrote:

rdx-fx wrote:

cospengle wrote:

You mean 'arseloads'?

I agree with your point, but you have to realise (as I'm sure you do), that changing to a different architecture isn't as simple as programmers changing the way they do things. If you program in a different architecture then everyone who uses your program will have to buy that architecture before they buy your program. As we've seen with Windows, people don't like to change. Vista has been out for ages and most people still use XP, even though they don't need to buy a whole new computer to run it (except for 64-bit). So to write a program for an architecture that no one has isn't really going to make you much money.

It's not the architecture that's the issue.  It's not the users.  It's the programmers.

If Windows and Office ran on the new architecture, the users wouldn't really give a damn what was under the hood.





And Vista, too, is a case of user-land not wanting to change.  Not because Vista is new and different - but because Vista doesn't do much better than XP (from a user perspective), and does a great many things more poorly than XP.

A trick to getting a new architecture in place;
Make your new platform do everything that the old platform did, using the exact same userland interface and skills - then add a few bells and whistles, and do it much much more efficiently/faster/cheaper than the old architecture.

Case in point:  Every generation of video cards.  Does what the old cards did, adds a few features to get the press drooling and writing, and does it all better/faster/cheaper than the previous generation.

Another case:  MS Office and Adobe CS.  Not too hard to bounce between OS-X and XP if you use those two suites alot.

The problem is that we've been doing just that for 20 years now, and we've more or less reached the top of what x86 can perform.

I'm not really sure how this argument started, but you're both kinda right and you're both kinda wrong.

The problem is not the hardware architecture. It's the software architecture.

Up until now, though Windows implements multi-threading it has always done so using a design that is most suitable for single-core processors (as that was all that was available when those parts of windows were first developed), now, as multi-cpu hardware became common, microsoft added layers to support that, and then an extension of that capability that allowed it to support hyper-threading and then, most recently, it was extended to make use of multi-core cpus.

But, right at its core, it's based on code that was designed to run on single core cpus.

And that's just windows.

Likewise, apps have always been designed to run efficiently on single core cpus.

And games are even worse. Until very very recently all games ran in single threads.

Multi-core x86 based hardware still actually has a lot of life in it.

Writing code to run efficiently on multi-core cpus isn't simply a case of re-compiling the source code.

You have to actually write the code differently.

But programmers will catch up and all the user really cares about is the UI, but what's under that UI needs reworking before the true power of multicore x86-based cpus will become apparent.

GC_PaNzerFIN wrote:

Windows 7 and DX11 oh yeah... that's when we'll see some epic multi-threading...

If Windows 7 actually is a whole new system, and not an add-on to the same, old NT5 kernel as all MS OSes have been for the last decade.

GC_PaNzerFIN wrote:

Freezer7Pro wrote:

GC_PaNzerFIN wrote:

Windows 7 and DX11 oh yeah... that's when we'll see some epic multi-threading...

If Windows 7 actually is a whole new system, and not an add-on to the same, old NT5 kernel as all MS OSes have been for the last decade.

DX11 specs gives wide support for multithreading tbh mean FINALLY games will LOVE quad core CPUs.

You're an optimist.