[NaMo4184]

I think that Apple will be using "cell" with asteroid. This is because the "cell" can be used to process media and are to be extremely scalable(cheap). I think it would be fitting for apple to use cell to process audio into Logic and Garageband because this implementation is along the lines with other niche uses of cell. i.e. game systems, media work stations, and HDTV's. Apple will benefit from this implimentation of cell because of its vast market share in the music industry and that 0 changes to the operating system would need to be made.

What do you guys think. Is this possible. The only thing that makes me hesitate to think it could happen is that no other cell device has been released yet and i am not sure that apple would want to be the first to release a cell based device.

For even more extreem speculations.

I think that the cell based IBM media work stations will be licensing OS X or can only be used through a mac. This is because in the media world OS X is like Microsoft (with out the gag reflex). Apple has a staggering market share in this area (ain't no 2% ) Also for every mac sold IBM makes money. Its in IBMs interest for macs to cell more (yes.. the cell part was a pun).

It really is a beautiful relationship.

[ASZ993]

Apple has no intentions of using Cell. It is clearly a processor for media devices (e.g. HDTVs, game consoles such as PS3)

[NaMo4184]

Apple has no intentions of using Cell. It is clearly a processor for media devices (e.g. HDTVs, game consoles such as PS3)

I never said they would use it in computers.... if you read the post it said they will use it in asteroid... its not a computer. IT IS A MEDIA DEVICE (at least I assume so).

Sorry if I was being unclear

I am thinking this asteroid device will be used to feed music to your computer and you will be able to process things such as synthesisers and aac or apple lossless encoding on the fly

edit: If its not clear I am saying asteroid is a separate device that is attaches to your computer and to you instruments line out.

[iRobot]

It seems fairly reasonable to me that Cell could be used in a Breakout box like Asteroid, and its strength when used in a series would be a boon to anyone daisy-chaining such a device.

However, I don't think that Cell is ready yet, and I doubt that its necessary for such an application.

[NaMo4184]

Thank you for replying! this topic doesn't seem to be taking off.

Originally Posted by iRobot It seems fairly reasonable to me that Cell could be used in a Breakout box like Asteroid, and its strength when used in a series would be a boon to anyone daisy-chaining such a device. However, I don't think that Cell is ready yet, and I doubt that its necessary for such an application.

I love the idea of daisy chaining the device i didn't even think about that one. it could take advantage of the capabilities of firewire.

I am also concerned that cell wouldn't be ready for this but I also don't think asteroid is ready so who knows. As for no market for it. I think that isn't true.

What i envision asteroid to be is a small device that allows you to plug in several instruments into it and then be directly connected to one's computer computer. Of course this would have to be firewire since garageband and all the other iapps don't exist for windows. I think that allot of people in bands including my self would like a cheap and easy way to do this so that it can all be put into garage band and mixed. I am sure this can be done without cell but I think that using cell, especially if its as cheap as SIT proclaims, will add allot of value to asteroid. what if you could do on the fly dps processing and someone can be controlling it from the computer and you and this one device has enough power to apply processing to say 8 instruments concurrently and isn't too expensive.

Another reason is that garageband takes advantage of core audio so the only thing that might need to be upgraded would be core audio and that shouldn't be too difficult. core audio has been refined for a long time and it would be nice for garageband to exploit all that effort.

[baaron]

I can think of the perfect example for this. (And I know a couple of bands who would be interested.)

Take a PM G5, leapfrog a few of these things via FW800, and basically import the entire band into Garageband 2 in real-time. Then after the song is finished, they could do whatever they want with it, maybe even play around with it using some Jam Packs.

As they got more into using this setup, they'd probably wanna start recording in some sort of 5.1 surround using Logic Pro 7, and a few big monitors.

Problem is, I dunno about the feasibility of all this, especially in real-time.

[Enki]

Putting a Cell, as the Cell was announced, in a breakout box would be like putting a G5 in a toaster as a controller to get the perfect slice. Pure overkill. The processing should stay in the computer where it has access to the support resources it needs, leave the breakout as the I/O plug interface it is designed to be.

You don't need ANY processor to daisy-chain, just a FW controller.

[baaron]

Originally Posted by NaMo4184 I think that Apple will be using "cell" with asteroid. This is because the "cell" can be used to process media and are to be extremely scalable(cheap).

I think we're talking about a super-scaled down version of cell that would be somewhat superior than a FW controller, to facilitate multiple instruments at one time.

But then, Apple could just go ahead and use a 486 or whatever in the asteroid, just like it did in the Snow Base Station. No new technology = No R&D = Cheaper.

But then again, we're merely speculating as to future uses of the cell processor in Apple's hardware. It's not gonna be for a long time, considering Apple & IBM are already investing R&D dollars into the Power 5 and the Power 6 architectures....

Maybe it could be the C7, instead of the G7....

[Kestrel]

Originally Posted by Enki Putting a Cell, as the Cell was announced, in a breakout box would be like putting a G5 in a toaster as a controller to get the perfect slice. Pure overkill. The processing should stay in the computer where it has access to the support resources it needs, leave the breakout as the I/O plug interface it is designed to be. You don't need ANY processor to daisy-chain, just a FW controller.

A processor has to be involved somewhere before the computer, because there is analog->digital conversion going on in the breakout box. They're suggesting that a cell processor would have the power necessary to, say compress the audio streams at the breakout box. Taking that load off the host computer would allow more breakout boxes to be used and would (possibly) reduce the FW bandwidth needed for each box. This means more instruments inputs, more real-time effects, and perhaps higher-quality recording could be done simultaneously.

Also, a FW controller has a small processor in it. A controller is a system that includes processor, memory, and usually some kind of storage.

[Enki]

DAC's are dirt cheap even for good ones, you dont need a general purpose computing processor to do the things you guys want. There are existing, way cheaper, ASICS that do those things for a fraction of a fraction of the cost involved with supporting anything like a Cell, x86, Gx or such. Playing with the semantics of what a processor "is" is like a past president playing semantics with his semen. It just doesn't work well and you are still left with a sticky stain.

[jouster]

Originally Posted by NaMo4184 ...because this implementation is along the lines with other niche uses of cell. i.e. game systems....

You consider that the PS 3 will be a niche use?


Heck of a niche.....

[Kestrel]

Originally Posted by Enki DAC's are dirt cheap even for good ones, you dont need a general purpose computing processor to do the things you guys want. There are existing, way cheaper, ASICS that do those things for a fraction of a fraction of the cost involved with supporting anything like a Cell, x86, Gx or such. Playing with the semantics of what a processor "is" is like a past president playing semantics with his semen. It just doesn't work well and you are still left with a sticky stain.

If by "dirt cheap" you mean $10-30 for a high-performance ADC ASIC, then yes, that's about what they cost in quantities of 1000+. Many of those are also single-channel units, so you would need several for something like Asteroid. You can check this out at http://focus.ti.com/paramsearch/docs...null&pageId=82 for ADCs and http://focus.ti.com/paramsearch/docs...f amilyId=392 for DACs.

A single Cell processor would be comparable to a high-end embedded ARM processor, which would be completely appropriate for something like Asteroid that involves heavy-throughput encoding and compression (assuming it does). Remember that this breakout box would likely support at least two channels/instruments.

Straw man fallacy aside, you seem to have missed my point that all firewire devices have a processor in them. The fact that Cell is not a system-on-chip like the controllers that many firewire devices use does not mean that it would not be appropriate for a firewire device.

[Enki]

First, Asteroid will not fit a high performance description, it's not a high performance price point. So the costs of your ASICs just dropped by about a factor of 6-8, then buy in lots of 10K and the prices should quarter-to-half again. That can put reasonable performance at a few dollars apiece. That's not straw, that's real world dirt cheap.

Second, Cell is not the savior for every widget that comes along.

Third, iPods use ARM7 processors. You saying that family of processor is on par with a Cell? I really hope not. I also haven't heard anyont talk of putting any ARM CPUs into workstations or parallel supercomputers, but yes to those accounts with Cell. That argument that Cell is cheap low end broke as well.

YES YES YES you can program a Cell to control Firewire. But why would you want to?

Fourth, I didn't miss anything about the firewire controller. I poked fun at your trying to equate the processing demands of a FW controller with needing something as powerful as a Cell to do the Astreroid justice. Your proposal to use a Cell isn't a 800 pound gorilla, it's like dropping a blue whale on an ant as far as overkill. And that's a damn expensive replacement for a Oxford911 controller + a $30 handful of ASICs. Apple doesn't do that in the desktops when a G5 could handle that controllers processing load. And you think it make sense to do the same thing with a processor that has similar cost, power and motherboard logistical support requirements? Possibility does not equate to making sense.

At Cells die sizes they should initially go for over $100 externally, you think that's going to allow a $149 breakout box? No. Those economics are also why Sony and Toshiba are building their own fabs for Cell. Then they get them at cost which then makes it economical to put them into their equipment.

[Kestrel]

Originally Posted by Enki First, Asteroid will not fit a high performance description, it's not a high performance price point. So the costs of your ASICs just dropped by about a factor of 6-8, then buy in lots of 10K and the prices should quarter-to-half again. That can put reasonable performance at a few dollars apiece. That's not straw, that's real world dirt cheap.

I didn't realize the expected price point for Asteroid was so low - this makes sense. The straw man I was talking about was the business about the president's semantics, as though my argument was less valid because I corrected you on a word's meaning.

Originally Posted by Enki Second, Cell is not the savior for every widget that comes along.

Obviously. In this case, it may indeed be higher-powered than necessary. But you seem to think it's absurd to even consider the possibilty that it could be used, and I'm not with you there.

Originally Posted by Enki Third, iPods use ARM7 processors. You saying that family of processor is on par with a Cell? I really hope not. I also haven't heard anyont talk of putting any ARM CPUs into workstations or parallel supercomputers, but yes to those accounts with Cell. That argument that Cell is cheap low end broke as well.

Yes, I'm saying the application of a lone chip is similar. People aren't putting ARM processors in high-end parallel processing environments because they aren't specifically designed to work in parallel. The iPod uses low-end processors, but Intel's XScale line currently goes to 600+ MHz. While there is probably a substantial gap between ARM performance and Cell's general-purpose performance, it's not the lightyears you're making it out to be. A single cell processor is not going to blow away a single processor of any other type except in some easily-vectorized applications.

Originally Posted by Enki Fourth, I didn't miss anything about the firewire controller. I poked fun at your trying to equate the processing demands of a FW controller with needing something as powerful as a Cell to do the Astreroid justice. Your proposal to use a Cell isn't a 800 pound gorilla, it's like dropping a blue whale on an ant as far as overkill. And that's a damn expensive replacement for a Oxford911 controller + a $30 handful of ASICs. Apple doesn't do that in the desktops when a G5 could handle that controllers processing load. And you think it make sense to do the same thing with a processor that has similar cost, power and motherboard logistical support requirements? Possibility does not equate to making sense.

Again, I think you're overestimating Cell's performance. It's not going to replace even a current G5 on its own, and it makes sense to do a lot of processing at the breakout box instead of the CPU. Also, logistics for one processor doing only one kind of I/O operation with limited memory and little-to-no storage would be much simpler than an entire motherboard.

Originally Posted by Enki At Cells die sizes they should initially go for over $100 externally, you think that's going to allow a $149 breakout box? No. Those economics are also why Sony and Toshiba are building their own fabs for Cell. Then they get them at cost which then makes it economical to put them into their equipment.

There is no way a four-processor PS3 is going to cost $400 for the processors alone. Obviously these chips will have to be cheaper than that. Supposing the processor costs $50 when production is really underway, there is still room for $50 of support hardware and a third of the product's cost is still profit.

[NaMo4184]

Your proposal to use a Cell isn't a 800 pound gorilla, it's like dropping a blue whale on an ant as far as overkill.

[Enki]

Originally Posted by Kestrel Yes, I'm saying the application of a lone chip is similar. People aren't putting ARM processors in high-end parallel processing environments because they aren't specifically designed to work in parallel. The iPod uses low-end processors, but Intel's XScale line currently goes to 600+ MHz. While there is probably a substantial gap between ARM performance and Cell's general-purpose performance, it's not the lightyears you're making it out to be. A single cell processor is not going to blow away a single processor of any other type except in some easily-vectorized applications. Again, I think you're overestimating Cell's performance. It's not going to replace even a current G5 on its own, and it makes sense to do a lot of processing at the breakout box instead of the CPU. Also, logistics for one processor doing only one kind of I/O operation with limited memory and little-to-no storage would be much simpler than an entire motherboard.

On intelligently scheduled code a 4Ghz PPE core should be roughly equivalent to a 2Ghz 970 +/-, then add 8 high throughput SPEs and a Cell is going to utterly thrash any 970 at things Cells are good at. Yes, the blue whale analogy is still valid.

Originally Posted by Enki At Cells die sizes they should initially go for over $100 externally, you think that's going to allow a $149 breakout box? No. Those economics are also why Sony and Toshiba are building their own fabs for Cell. Then they get them at cost which then makes it economical to put them into their equipment. There is no way a four-processor PS3 is going to cost $400 for the processors alone. Obviously these chips will have to be cheaper than that. Supposing the processor costs $50 when production is really underway, there is still room for $50 of support hardware and a third of the product's cost is still profit.

Like I already said, that's why Sony & Toshiba are building their own cell fabs, so they won't have to pay those prices.