Will the Cell Processor make its way to pcs?

[Smartazz]

It seems as though Toshiba plans to use the Cell Processor for many applications, is there any plan to bring it to personal computers? Thanks.

 

[andy04]

Its the future, its going to make its way into everything...

I feel eventually you will have only one Processing unit which will do everything in your home... and that will be cell too...

 

[DaveSimmons]

Zero chance except maybe in some oddball linux PC that sells 1,000 units to the 100 million a year sales of x86 PCs.

x86 has 100% of the desktop CPU market now that Apple has switched, and there is no incentive to switch away as long as intel and AMD keep increasing performance.

General-purpose cores are much easier to write code for than specialized ones in a single application, and running multiple applications on multiple cores takes no extra work at all.

 

[Roguestar]

Doubt it, the Cell is too specialised. It's like a giant multi-core GPU from what I understand. The way GPUs can run highly-threaded largely FPU-based code these days (such as the F@H client) reinforces that in my mind.

 

[Matthias99]

I doubt it will replace x86 CPUs any time soon, as Dave mentioned above. Multi-core x86-64 is where it's at in desktop computing for the foreseeable future.

I could see some descendant/variant of Cell being used to replace things like sound/video/physics cards... using it as a parallel general-purpose FPU for stuff that's too intensive for a CPU. That was what Sony was talking about a few years back as the architecture of the PS3.

However, you'll note that the PS3 ended up going with a discrete GPU, and maybe also discrete sound? So the technology's not quite there yet.

 

[her34]

cpu's may become more cell like in the future. ati's fusion chip is like a mini-cell

 

[Seekermeister]

DaveSimmons,

Zero chance except maybe in some oddball linux PC that sells 1,000 units to the 100 million a year sales of x86 PCs.

It may still be a small number, but it would actually be 360,000 out of 100 million.

 

[DaveSimmons]

Originally posted by: Seekermeister
DaveSimmons,

Zero chance except maybe in some oddball linux PC that sells 1,000 units to the 100 million a year sales of x86 PCs.

It may still be a small number, but it would actually be 360,000 out of 100 million.

It's true a few PS3s will be used to run linux, but I wasn't speaking about the PS3 with that comment. 🙂

I mentioned linux because Windows and OS X will almost certainly never be ported to run on cell, while linux already has been.

 

[Rubycon]

Blackberry Ruby in 2008.

 

[bamacre]

Intel has already said they will be able to offer an 80-core to consumers in 5 years.

No way in hell Cell will be on PC's.

 

[GetMedieval]

I think its pretty much been confirmed that the Cell was coming to PCIE for development tasks.

 

[TanisHalfElven]

it already has. i don't have a link but a company is already making card with cell procs on them that can plug into a computer slot

 

[UMfanatic]

orginally cell processors were going to replace CPUs because of there cooling efficiency, because as things became more powerful the heat levels rose too, but most of that talk has been squandered

 

[Smartazz]

Originally posted by: GetMedieval
I think its pretty much been confirmed that the Cell was coming to PCIE for development tasks.

Yeah, it has been confirmed, but only for linux as of now, it may come to windows eventually though. Doesn't intel have plans for an 80 core consumer level cpu in 2010? I'll try to look that up.

 

[Oyeve]

Isnt there some laser optical processor thingy coming in the next decade or so? I know the old transitor tech on current processors are about to hit a limit soon and the frequencies will soon interfere with other components. There is something out there in the works for sure.

 

[Matthias99]

Originally posted by: Oyeve
Isnt there some laser optical processor thingy coming in the next decade or so? I know the old transitor tech on current processors are about to hit a limit soon and the frequencies will soon interfere with other components. There is something out there in the works for sure.

There are companies working on optical processing... but it's still in the lab, and a LONG ways from commercialization. Silicon's not going anywhere for a while. Although yes, Moore's Law is supposed to top out somewhere in the next 10-20 years, because you won't be able to make conventional silicon MOSFETs any smaller/faster without running into quantum mechanics issues.

You can also make higher-speed chips with more exotic materials (such as gallium arsenide) -- but they are expensive, and ill-suited for mass production. They're mostly used for high-speed communications equipment that has to run at 10+Ghz.

 

[Smartazz]

Originally posted by: Matthias99

Originally posted by: Oyeve
Isnt there some laser optical processor thingy coming in the next decade or so? I know the old transitor tech on current processors are about to hit a limit soon and the frequencies will soon interfere with other components. There is something out there in the works for sure.

There are companies working on optical processing... but it's still in the lab, and a LONG ways from commercialization. Silicon's not going anywhere for a while. Although yes, Moore's Law is supposed to top out somewhere in the next 10-20 years, because you won't be able to make conventional silicon MOSFETs any smaller/faster without running into quantum mechanics issues.

You can also make higher-speed chips with more exotic materials (such as gallium arsenide) -- but they are expensive, and ill-suited for mass production. They're mostly used for high-speed communications equipment that has to run at 10+Ghz.

hmm, that stuff is toxic and carcinogenic. Interesting info though. 4 years back my teacher told me that laser optic processing was on its way and that the capabilities would crush our current silicone processors. Silicone still has a lot of life left in it though.

 

[40sTheme]

Originally posted by: bamacre
Intel has already said they will be able to offer an 80-core to consumers in 5 years.

No way in hell Cell will be on PC's.

But I have to ask: What would be the point? Every new number of cores and every new number of threads the processors have, software developers have to revamp the way they think. They have to reconsider the entire utilization of the cores. If we're at 4 cores now and don't have support, 80 cores in 5 years would be worthless. Dual core CPUs aren't even being utilized in many applications.

 

[Smartazz]

Originally posted by: 40sTheme

Originally posted by: bamacre
Intel has already said they will be able to offer an 80-core to consumers in 5 years.

No way in hell Cell will be on PC's.

But I have to ask: What would be the point? Every new number of cores and every new number of threads the processors have, software developers have to revamp the way they think. They have to reconsider the entire utilization of the cores. If we're at 4 cores now and don't have support, 80 cores in 5 years would be worthless. Dual core CPUs aren't even being utilized in many applications.

Yeah, but dual core and quad core processors multitask so well, 80 cores would be overkill though, but we don't know where technology is headed anyway. I'll probobly switch to a quad core amd processor in a couple years, dual core is way better than single core in almost everything.

 

[cheesehead]

Originally posted by: Smartazz

Originally posted by: Matthias99

You can also make higher-speed chips with more exotic materials (such as gallium arsenide) -- but they are expensive, and ill-suited for mass production.

hmm, that stuff is toxic and carcinogenic.

You have to consider, though, that the portion of a CPU wafer that forms a modern processors weighs under a gram. Gallium arsenide was at one point considered as a material for use in LEDs.

I think that the Cell would make a much better alternative to an AEGIA card - just write the software. You could use the various cores for various sub-sets of gaming (use one for networking, one for physics, one for smoke effect mapping, et cetera) and the cost would not be unreasonable.

Also, is'nt BeOS optimized for use with lots and lots of CPU cores? The Cell would make for a killer audio-video editing workstation.

 

[songokussm]

i was reading an article about a user's experiance with linux and the ps2. he stated it it required nearly 10x the programing code and time to write the same app for a cell chip as it did for a x86 chip. i know nothing about programming so jus tmy 0.02.