Can Cell Processor make it into PC's?

skilzygw

Newcomer
Just wondering from your knowledge of the cell chip. Could it replace the x86 processor in the future?

I mean throw out the monopoly intel thing. Purely from a technical standpoint, could it run everything a intel processor could run if coded for? Would it be better?

etc..
thanks for the info. Love tech talk.
 
If you aproach the problem as triying to replace a Pentium with a Cell, the answer is obviously not.

If you think of M$ making a migration towards the powerpc-cell architecture, well... mmm... Do'h, this is not going to happen :rolleyes:
 
The CELL is an awesome processor for what it is intended to do. But it cannot really compete with a x86 processor for desktop apps because it was not designed with the goal for general computing tasks as its primary tasks.

Three reasons CELL wont overtake x86:

1. Massive x86 software platform. A CELL PC would not only need to offer a significantly better computing experience--so much so Intel could not close the gap--it would need to offer a steady stream of great applications and updates. So far there is no real platform to support mainstream software on CELL. It takes years to build a quality software and developer base, which leads to...

2. While CELL is building develpper support to create this software Intel will be moving to more and more advance multicore systems. Intel has had CELL like ideas on their roadmap for a while. They eventually plan to put dozens of processors and co-processors on their chips. So while CELL may be a great idea, Intel can move in on the same idea. Re-engineer the Pentium x86 series to have some Floating Point intensive units that perform a couple hundred GFLOPs and you get x86 compatibility with massive FP performance. I tend to think Intel can make this happen faster than CELL can develop a significant software base.

3. The Wintel "monopoly". CELL is up against two massive companies that are not only financially stable but also continue to invest billions yearly into research and development. They have the know how and marketing power to move the industry.

CELL is a great chip. It is going to make a GREAT console CPU. It also boads VERY well for the PS4 (which will be CELL based) because it will create a smooth transistion. CELL was built around the concept of expanding power through more CPUs, so CELL will serve this purpose well. Sony will be able to save a lot of money on RD with PS4 because the chip will have already been refined and in mass development. This will give Sony a huge financial edge, and most likely put them yet again in a dominant position in performance. CELL may also find its way into supercomputers that need a lot of FP performance.

I just don't see it making significant headway into the x86 market.
 
Hey, ya never can say never, eh? Let's just see how Cell's "virtualization" functionality comes into play in the future. Who can really say where a PPC Windows may or may not make a reprisal appearance in history?... ;) ...or maybe it ends up being a pure Linux setup. It may not do "everything" well that you find in the x86 world of codebase, but it may not need to. It needs only to do most stuff "fair" (because no one "needs" MS Office to run 200 fps), and the really time/compute intensive really well. That's where it was designed to deliver the big punches on stuff that brings x86-ware to a crawl. The "power" of the hardware is focused on the types of stuff that really need it, rather than just indiscriminantly on anything and everything (even menial stuff that people will never notice if it is running faster or not). That's where it may or may not raise some eyebrows when it comes to adapting new hardware styles. It's a big gamble, yes, but anything "good" starts out as a gamble.
 
If anything, i could see a vastly modified version of Cell in future Macs. Not gonna happen, but it's not as impossible as Cell taking over x86.
 
Vaan said:
If you think of M$ making a migration towards the powerpc-cell architecture, well... mmm... Do'h, this is not going to happen :rolleyes:

Something like this is happening right now.. It's called the xbox 360 (the premium version with HDD will allegedly do web surfing, IM and media playing, and it were supposed to be a third, PC-like version ?!?)
 
Yes, I think we can finally dispense with the notion of how "impossible" it is for MS to migrate to an entirely different ISA that isn't x86 simply by the XB2 being what it is. They certainly have given x86 a long, deserved run, but it is by no means cosmically bound to it (though it may seem so if you are a young'un to the business).
 
Isn't Cell based on RISC architecture? I think it is. If i am right then i'd se it wont be seen in ay PC for desktop usage anytime soon.
It might be used in some workstations though where it's potential can be used.
 
mustrum said:
Isn't Cell based on RISC architecture? I think it is. If i am right then i'd se it wont be seen in ay PC for desktop usage anytime soon.
It might be used in some workstations though where it's potential can be used.

These days, the line between RISC and CISC is so blurred it hardly makes a difference.
 
So how would you convert Windoze to work with a Cell CPU?
Could it work just by M$ writing a new kernel or some other intermediate layer?
Or would every PC program (/every one that bypasses the win kernel) need to be at least recompiled if not rewritten to match the different architecture?

The former would seem possible but the latter won't happen.
But if Sony were to provide a userfriendly OS with browser, office suite, email etc then there will at least be a bunch of Cell based media centre PCs kicking around & would presumably garner a market for new non game software.
 
arrrse said:
So how would you convert Windoze to work with a Cell CPU?
Could it work just by M$ writing a new kernel or some other intermediate layer?
Or would every PC program (/every one that bypasses the win kernel) need to be at least recompiled if not rewritten to match the different architecture?

The former would seem possible but the latter won't happen.
But if Sony were to provide a userfriendly OS with browser, office suite, email etc then there will at least be a bunch of Cell based media centre PCs kicking around & would presumably garner a market for new non game software.

As it stands now, Cell is a processor that is made to compute a massive (MASSIVE) amount of more or less complicated math in the quickest period of time. It does that VERY well, and not much else, as it doesn't need to, being in a console.

These days, Intel and AMD processors made for PCs are much more suited for Windows than Cell.

So, until we get a vastly modified version of Cell with all the FLOPS power AND all the other PC-centric features desktop processors have, Cell will not be used for PCs.

Imagine, that would look like an X-core POWERPC at G5 level, PLUS all the massive SIMD units (SPUs) that Cell has. Very big chip. And when that happens, a G5 core will already be old.
 
arrrse said:
So how would you convert Windoze to work with a Cell CPU?
Could it work just by M$ writing a new kernel or some other intermediate layer?
Or would every PC program (/every one that bypasses the win kernel) need to be at least recompiled if not rewritten to match the different architecture?

no need for a new kernel, and the intermediate layer has been here since the beginning of windows NT : it's the HAL (hardware abstraction layer).
I guess it would run if you recompile, but if it isn't optimised (some vectorized code to feed the SPEs) it would suck.
The PowerPC core (with SMT) in the Cell is NOT a G5 but much simpler (in order execution), it might well be slower than a typical PC CPU.
 
london-boy said:
Imagine, that would look like an X-core POWERPC at G5 level, PLUS all the massive SIMD units (SPUs) that Cell has. Very big chip. And when that happens, a G5 core will already be old.

this is what plan AMD and Intel I think. not PPC but likely x86-64?
it's what I want, an AMD "cellthlon" on a PC still having a IBM compatible BIOS, with PS/2, serial port and floppy still there, able to run MS/DOS if I want to :D (along with pirated windows 2007, ReactOS, linux, all at the same time through native virtualization)
 
Well, MS is really shooting towards having Windows exist on arbitrary hardware platforms. That's part of the point behind having Longhorn entirely sitting on a CLR backend. As long as there's a CLR JIT for CELL, at the very least, you can run just about anything Windows on it.

The hard part is getting things to perform even halfway decent on there. Especially if you're talking about a potential gaming PC. If games are written on CLR, chances are that developers are not thinking about threading to hell and back and breaking everything down to tiny apulets. CELL as we know it is a little too demanding of a specific development model. The world of general apps is definitely not ready for that. It's way too naive to think that the CLR backend will handle it all for you.

Now upper-end workstations and stuff I can see in the near future. CAD boxes, render workstations, farmed tasks, etc.
 
I think the biggest limitation with regards to CELL being used as a "desktop" processor is whether or not the vector units on it can be easily accessed in a multi-user environment. Certainly it's fast enought o do things like word processing, and for mutlimedia apps it'd be absolutely fantastic.

Apple would probably be the most likely candidate to use it, but IBM is making the Cell workstation, so perhaps we'll see some form of linux running on it.

Nite_Hawk
 
From what i've read, it seems like Cell is going to obliterate any AMD or Intel offerings.

Could someone who knows this kind of stuff tell me whether Cell is really an AMD/Intel killer, or if AMD/Intel are still overall better, faster chips?
 
Intel17 said:
From what i've read, it seems like Cell is going to obliterate any AMD or Intel offerings.

Could someone who knows this kind of stuff tell me whether Cell is really an AMD/Intel killer, or if AMD/Intel are still overall better, faster chips?

Once again, in pure floating point math, Cell kills pretty much everyone. But running Windows, using Word and Excel, floating point math matters as much as hippos.
It really depends on what application you're running. Cell is made for a console, Intel and AMD chips are made for desktop. It also explains why MS is going for neither of them for Xbox2, preferring a PPC architecture, much like what's inside Cell.
 
london-boy said:
Intel17 said:
From what i've read, it seems like Cell is going to obliterate any AMD or Intel offerings.

Could someone who knows this kind of stuff tell me whether Cell is really an AMD/Intel killer, or if AMD/Intel are still overall better, faster chips?

Once again, in pure floating point math, Cell kills pretty much everyone. But running Windows, using Word and Excel, floating point math matters as much as hippos.
It really depends on what application you're running. Cell is made for a console, Intel and AMD chips are made for desktop. It also explains why MS is going for neither of them for Xbox2, preferring a PPC architecture, much like what's inside Cell.

london-boy,

How would AMD/Intel v.s. Cell look in terms of gaming applications?
 
Intel17 said:
london-boy,

How would AMD/Intel v.s. Cell look in terms of gaming applications?

With current games engines, Cell is likely to perform significantly slower than Intel/AMD. But for future games it's anyone's bet. For Cell to take the lead, games will have to introduce tons and tons of interactive physics objects. For multi-core Athlons/Pentiums to keep the lead, games are going to need to use a bunch of AIs.

Keep in mind those two examples are just that examples, I'm sure there are plenty of things developers could introduce/use more heavily that influence performance in the same way for the same reason.

So it all comes down to floating point math or branching.
 
There's also OS overhead on the PPE portion of the processor.

The SPEs are all scheduled by software running on the PPE, which is at least in these early examples rather spare.

The PS3 will avoid a lot of the crud associated with the PC software platform, so that may not be as large an issue for the console.

Considering the bloat associated with fully-featured operating systems, it may not be that great for the PC market for a generation or two.

We also don't know how well it does on tasks that don't suit the SPEs very well.
 
Back
Top