Intel gave a statement about Cell!

[McFly]

Some competitors, however, are skeptical that Cell will find much of a home outside of video games. One of the big problems with Cell, said Justin Rattner, an Intel fellow, is that the processing units aren't identical, a situation that increases complexity and the opportunity for bugs.

"You've got this asymmetry," Rattner said. "It's like having two kinds of motors under the hood. We are very reluctant to adopt architectures like this because they take compatibility and throw it out the window."

http://news.com.com/PlayStation+3+chip+has+split+personality/2100-1043_3-5566340.html

But are the APU's not all the same? He can't seriously talk about the asymetry between the PowerPC core and the APU's or does he?

Fredi

 

[Titanio]

I think he's talking about the asymmetry between the Power core, and the APUs, yeah.. The APUs are all the same, afaik.

 

[DeanoC]

Well yes, SPUs are very specialised (256Kb mem with DMA), so you can't run normal code on them. Therefore every task has to be targeted before hand at which processor class to aim for.

ASMP is a lot harder to program for than SMP (and that not easy). This has been known since the 1970's, will the extra pain be worth the gain? Thats a question only the future will answer.

 

[Npl]

But are the APU's not all the same? He can't seriously talk about the asymetry between the PowerPC core and the APU's or does he?

Fredi

Thats what he meant - worst argument posible, I mean just imagine you would take a x86 Processor and a specialised Processor like a GPU and use it in the same Machine?

Backwards-Compatibility is a big issue though.

 

[McFly]

So I guess he don't likes PowerPC/Altivec as well.

Fredi

 

[Fodder]

And Intel would never try to force a new architecture with pathetic back-compatibility on the industry. :?

 

[McFly]

http://www.siliconvalley.com/mld/siliconvalley/10840038.htm (The whole article is interesting as well)

Cell's designers said they are running a variety of operating systems on the processor at their lab in Austin, Texas. But they would not say whether Microsoft's Windows is one of them. In fact, they only confirmed running Linux, the open source environment.

OS X maybe?

Fredi

 

[j^aws]

http://www.siliconvalley.com/mld/siliconvalley/10840038.htm (The whole article is interesting as well)

Cell's designers said they are running a variety of operating systems on the processor at their lab in Austin, Texas. But they would not say whether Microsoft's Windows is one of them. In fact, they only confirmed running Linux, the open source environment.

OS X maybe?

Fredi

Apple sources don't sound convinced on CELL yet for OSX,

...
One area of wide speculation is whether Apple might become a partner in the Cell alliance in the future. Apple is already the largest customer for the PowerPC chip, and it would be simple for the company to take advantage of the Cell design. Several people familiar with Apple's strategy, however, said that the computer maker had yet to be convinced that the Cell technology could provide a significant performance advantage.

NYTIMES

But at least they've thought about it!

 

[McFly]

Yeah, I think Windows is a given if they want to be a real threat to Intel, but Apple will sure show their interest if they see the chip performing well.

Fredi

 

[j^aws]

Yeah, I think Windows is a given if they want to be a real threat to Intel, but Apple will sure show their interest if they see the chip performing well.

Fredi

Actually, it's prolly easier to port OSX to Xenon than CELL!

 

[McFly]

Actually, it's prolly easier to port OSX to Xenon than CELL!

Thing is ... I don't care that much about Xenon.

Fredi

 

[Alejux]

I imagine that the problem with using CELL in some general purpose machine, is that most of today's software are not made in a way that would take advantage of an architecture like CELL. In fact, I bet the performance of these programs would suck in it.

 

[nAo]

I imagine that the problem with using CELL in some general purpose machine, is that most of today's software are not made in a way that would take advantage of an architecture like CELL. In fact, I bet the performance of these programs would suck in it.

Most of the software we use today would probably run very fine on the CELL PU..
Most of the sw that can't just run fine on the PU part would fit very nicely with the CELL architecture as a whole..
I don't see a CELL CPU efficiently run a database or a web server but it would be a great replacement for my desktop CPU

ciao,
Marco

 

[PC-Engine]

I don't see a CELL CPU efficiently run a database or a web server but it would be a great replacement for my desktop CPU

You still using a 8088?

 

[nAo]

You still using a 8088?

No, a P4 (Northwood) 3.4 Ghz. And it doesn't run standard desktop application much faster than my other older PC (K7 1.2 ghz..).

 

[Wunderchu]

Until actual products come to market we can only speculate on the possible uses of the Cell technology, but it is a large leap of faith to assume that its designers will be able to engage the world-wide community of software developers to create a role in the general marketplace that is similar to the vast impact of the PC.

[source: http://www.anandtech.com/news/shownews.aspx?i=23790 (direct link: http://www.trustedreviews.com/article.aspx?art=1025 )]

 

[psurge]

I could see Cell being fantastic for databases as long as you aren't dealing with very large fields.

 

[a688]

I could see Cell being fantastic for databases as long as you aren't dealing with very large fields.

Databases are very random access and dynamic. I don't think cell would do a good job. Faster storage would provide a bigger increase in performance. PS: cell doesn't do everything (or anything at this point) faster than other processors.

 

[psurge]

They are also extremely IO bound and (from what I can tell) generally organize data into pages that store data at greater than row/field granularity, making a software page cache mapped to SPEs via DMA feasible. IMO a number of queries could benefit from being split or pipelined across SPEs. I agree that it isn't an optimal design for a DBMS system, but it does have a lot of bandwidth (in/off chip) and parallel resources available.

Also, I'm aware that Cell is not the second coming. In fact, as far as "new" architectures go, I find it dissapointing (especially compared to some of the ideas in academic designs). On the other hand, its not clear to me that Cell will be slower than existing monolithic designs (targetted by current code) just because a particular task happens to be very dynamic. I can see efficiency going down the tubes in such cases, but there are so many ALUs available that it could still come out ahead (just IMO). It will be very interesting to read comments from developers working on cross platform titles.

 

[Ty]

You still using a 8088?

No, a P4 (Northwood) 3.4 Ghz. And it doesn't run standard desktop application much faster than my other older PC (K7 1.2 ghz..).

I believe he was saying that only if you had a 8088 CPU would you want to replace your CPU with a cell one.