Could CELLs be used in PCs?

[Cryect]

cell in a pc would sure as hell be faster than what we have today. maybe we wouldnt be so cpu limited. its slower for general purpose code, but how fast do we need microsoft office to run.

For what we do today, it would be a hell of a lot slower for the average user who uses a total of one task at a time. A single APU would essentially being used which wasn't made very well with general computing in mind crippling performance even more...

Now on the otherhand for a heavy multitasker, who's doing MPEG2 to MPEG4 encoding at HDTV resolutions while playing music videos, putting together slides for a presentation and chatting on AIM though could definately use a Cell.

 

[ondaedg]

I work in software development and all I can say is a simpler user experience is definitely the way to go. Therefore, if cell can help create a more universal, secure, and simpler interface and link PC/appliance architecture, then let's do it! To all you naysayers ripping on the guy for trying to be innovative because you think you know more than he does, stuff it. You're just hindering progress. Besides, what harm can be done by suggesting it? Answer is: None.

 

[nelg]

A U$ 500.00 computer doesn't run games at almost CG graphics level. If you want that, you'll have to spend at least 3X the amount. And a lot of home PC buyers want to play games. IMO, games are the major reason that people buy expensive computers. Why else would someone pay 2K on a computer? To run excel??

You just can't compare the power of the U$500.00 PC with a PS3/XB2. Unless PC graphics solutions have their price greatly, greatly reduced, cheap PC's will never be in the same ball park as consoles.

Wow, imagine if Microsoft charged a licence fee for every piece of software written for Windows. They could then give Windows away for free. They could extend this concept even further and also give away the hardware, even a super computer. So instead of having up-front fee’s for hardware and software which represent their value, you have a system that hides this. Users who invest in a lot of software would in fact be paying more for their hardware (through hidden fixed fees) than those who purchase very little. Hey wait. Something like this already exist. Its called a console.

Seriously, the revenue model of consoles is in conflict with the use of the same components in other products that do not have continuing revenue streams.

 

[DudeMiester]

I for one am waiting to buy a CELL workstation/PC or whatever form it comes it. If people buy it, the software will come. Plus since the XBox 2 is a PowerPC based platform, then they will have to start distributing en masse as PowerPC compatible version of Windows, which will probably be able to run on Macs. Not surprising that they were using modified G4s for XBox2 development platforms. So if you combine XBox2 & CELL & PS3 & Macs & Windows & Linux & More, by golly that's a hell of a lot of software all using the PowerPC architecture and all thus fairly portable between the variously platforms.

I think software won't be an issue. Even if it is, I don't care, I'm getting one anyways for the cool factor! lol.

 

[hovz]

cell in a pc would sure as hell be faster than what we have today. maybe we wouldnt be so cpu limited. its slower for general purpose code, but how fast do we need microsoft office to run.

For what we do today, it would be a hell of a lot slower for the average user who uses a total of one task at a time. A single APU would essentially being used which wasn't made very well with general computing in mind crippling performance even more...

Now on the otherhand for a heavy multitasker, who's doing MPEG2 to MPEG4 encoding at HDTV resolutions while playing music videos, putting together slides for a presentation and chatting on AIM though could definately use a Cell.

considering a pentium at less than 1 ghz is fast enough for what most people do on the computer i doubt the cell would be agonizingly slow for general purpose computing. then when u realize that it has over 10x the gflop performance of the fastest athlon64 things get exciting.

 

[Killer-Kris]

cell in a pc would sure as hell be faster than what we have today. maybe we wouldnt be so cpu limited. its slower for general purpose code, but how fast do we need microsoft office to run.

For what we do today, it would be a hell of a lot slower for the average user who uses a total of one task at a time. A single APU would essentially being used which wasn't made very well with general computing in mind crippling performance even more...

Now on the otherhand for a heavy multitasker, who's doing MPEG2 to MPEG4 encoding at HDTV resolutions while playing music videos, putting together slides for a presentation and chatting on AIM though could definately use a Cell.

considering a pentium at less than 1 ghz is fast enough for what most people do on the computer i doubt the cell would be agonizingly slow for general purpose computing. then when u realize that it has over 10x the gflop performance of the fastest athlon64 things get exciting.

Thing is that Cryect is right. A single APU could probably be considered the equivalent of a really high clocked 486 or something similar. It get's its great power from having lots of these. But you can only harness this power if you have a massively parallel program, or lots of programs running. You're average desktop user will only be able to tap into a tiny fraction of its potential. So it may not be agonizingly slow, but it will probably be pretty close to that point.

Edit: Note: From here on down is STRICTLY conjecture based on some limited facts...
Also from what I've seen they also don't intend to include much cache into each APU, and as such if you attempt to run a heavy duty thread (ie. your average app) performance will probably be quite abysmal. On the other hand, if you can break your massively parallel program down into very small simple chunks it won't suffer quite as much.

 

[_xxx_]

To make your note shorter, you'd need an optimized compiler. Which is the first thing that happens when a new CPU appears...


There would be absolutely no problem making a PC with a CPU made of dried crap, with an appropriate OS if someone should really, really want to.

 

[Killer-Kris]

To make your note shorter, you'd need an optimized compiler.

Problem is that there is only so much a compiler can do. It likely would be able to take care of the parallel program no problem, but there's not much if anything it can do about the serial programs.

 

[Ragemare]

Well I'm willing to bet you will be able to run Linux and various other open source OS's/programs on a Cell based computer.

As for the speed of programs not specifically written for the Cell, that all depends on how well the compiler can parralise the algorythms in the program and/or stream them through the APU's on it's own.

Regardless of whether or not the APU's are used, the Cell won't be that slow because of the PU which is supposedly a bit similiar to a G5 core.

On the subject of consoles destroying the PC and PC like architectures; I very much doubt it. I for one wouldn't want to be not able to upgrade my computer from the huge choice of components that the PC market offers.

People like me will support the PC market (or something like it) so long as we live. If all the computer iliterates bugger off to the console market it won't bother me too much, as long as there's enough people buying PC hardware to make companys like Nvidia and ATi create it, I'll be happy.

The idea of being forced to use a console for everything horrifys alot of people on these boards, as you can tell by the hostility to any such idea.

 

[PC-Engine]

Exactly!! I absolutely love the variety of computers and computer components out in the market.

 

[Killer-Kris]

The idea of being forced to use a console for everything horrifys alot of people on these boards, as you can tell by the hostility to any such idea.

I can only speak for myself, but personally I love the idea of something like a cell based PC. Only problem is when you start thinking about the practicality of it. Many users on the B3D boards would likely more than love, and have a use for a cell-like PC. It's just that it's incredibly impractical for the rest of the "computing" world where your standard CPU design is the best way to go.

And in all actuality I'm not even entirely sold on the cell/niagra way of thinking about CPU design. They're likely to be incredibly poor performers on all but the most parallel of code. The second you attempt to run the other 99.99% of the software out there, performance will tank and tank hard. And if that weren't bad enough there is not a whole lot anyone can do about most of those programs either.

In all reality, the likely best future CPU path will be something more along the lines of multicore PMs and A64s. They're small, flexible, well understood by coders and compiler writers, and far more likely to begin to approach their theoretical performance.