The CORE I7 OR The CELL Processor????

[TheOtis]

It states that the number if transistors per square inch on an IC double every two years. I think he said 18 months at first, but then changed it to two years. Theirs a lot more to it then that, but that's the just of it.

And to sparkmonkey- I'm pretty sure it says they double every two years thereby effectively doubling performance every two years, power consumption going down a certain amount, and cost going down a certain amount. Don't directly quote me, it's what my physics teacher(used to work for Intel) told me a couple years ago.

I think the only reason Moore's law is still around is thanks to multiple cores. I know getting to 32nm was harder then they had thought, and I know they will or already are running into physical boundries. If i remember right the distance between the gate and base on a transistor is only 3 atoms thick at 45nm or 32nm. Again don't quote me directly on this, I'm trying to remember when I learned two years ago.

 

[SparkMonkeyHellion]

The "3 atoms" figure sounded way off to me, because I know for a fact that electrons tend to jump well before then, according to wiki:

Consequently, the smallest atom is helium with a radius of 32 pm, while one of the largest is caesium at 225 pm.

Those are picometers, a full 3 orders of magnitude smaller than nanometers. I think that if we had components which were really 3 atoms thick, that would classify as nanotechnology if not quantum computers!

Anyways, it looks like you're right about the transistors thing. There should be a finite limit to silicon transistor technology, not to say that silicon won't get replaced by something else soon enough.

@cell vs cores - I can't remember if it was already mentioned, but the biggest setback with cell technology is that you have to program for it. This is new and difficult as far as I understand it. If it is too difficult to learn to make efficient, then it really doesn't matter how potent the processor is.

"A poor gun in the hands of a good shooter is more dangerous than a good gun in the hands of a sh***y shooter"

 

[TheOtis]

The "3 atoms" figure sounded way off to me, because I know for a fact that electrons tend to jump well before then, according to wiki:



Those are picometers, a full 3 orders of magnitude smaller than nanometers. I think that if we had components which were really 3 atoms thick, that would classify as nanotechnology if not quantum computers!

Anyways, it looks like you're right about the transistors thing. There should be a finite limit to silicon transistor technology, not to say that silicon won't get replaced by something else"

Ha, I figured it was. I want to figure that out now.

You think Silicon will get replaced by something else? I'm no physics guy or scientist, but what other elements act like SiO2?

 

[zedman3d]

I lol'd when i saw the thread title.

For some reason i have a mental link between the performance of the PS3's cell and something like the original Phenom's....
And yes, moore's law is to do with the doubling of the amount of transistors.

 

[uneeb83]

I agree with sparkmonky...
But thats just the point. If programmers start making programs which utalize the full power of the cell processor then the cell tecnology can come up really fast and out perform the core processors

But this however needs a great revolution on the sofware development part. this is unlikely to happen but considering that the core processor tecnology is running into physical bounds, sooner or later we would need to shift to cell tecnology....

 

[SparkMonkeyHellion]

You think Silicon will get replaced by something else? I'm no physics guy or scientist, but what other elements act like SiO2?

Well there's the hydrogen array processor. Since hydrogen have a dipole, they can act like little magnets, except they are "polar" they are bent a little bit. This makes them like little levers. If you put them in an array, they "wiggle" and each wiggle will affect it's neighbors in a predictable way. So once you've got a few trillion H atoms in a grid, you have a modern processor running at several THz - powered by ambient heat.

Sadly, I cannot find the original article I saw this on.

@uneeb - I heard on it's launch that the PS3 has about 9x (or more) the raw processing power of top-end PC's due to the cell processor. This benchmark was also compared by how fast a PS3 can run F@H compared to a typical PC.

 

[Poizen22]

Well there's the hydrogen array processor. Since hydrogen have a dipole, they can act like little magnets, except they are "polar" they are bent a little bit. This makes them like little levers. If you put them in an array, they "wiggle" and each wiggle will affect it's neighbors in a predictable way. So once you've got a few trillion H atoms in a grid, you have a modern processor running at several THz - powered by ambient heat.

Sadly, I cannot find the original article I saw this on.

@uneeb - I heard on it's launch that the PS3 has about 9x (or more) the raw processing power of top-end PC's due to the cell processor. This benchmark was also compared by how fast a PS3 can run F@H compared to a typical PC.

magic words typical pc. typical pc is equivelant to a p4 performing system. would say the world iso utaded enough for it to be a p3 on average but id say its a p4. at that it wouldnt be a high performence 4 either maybe just 2ghz.

 

[C0RR0SIVE]

IMO the cell has more processing capabilities, hence the reason people at schools are networking them together to become mini super computers and to process huge chunks of data...

 

[SparkMonkeyHellion]

Still, i can imagine the headache of writing the machine code to say "okay, send this task over core V and this task over to core IX"

It would be great if they had a sort of "RAID controller" for multiple cores, where the application and OS layer only sees 1 processor, but you've got a piece of hardware capable of divvying up the threads appropriately, just like how a RAID controller can keep track of a whole file system, and your OS only sees one drive, even though it could be 7.

So maybe instead of RAID controllers we should have RAIC controllers??? "Redundant Array of Inexpensive Cores"

 

[C0RR0SIVE]

True, but that sort of function should be, in it self, programmed into the CPU, and can be turned on/off by the software that is running on it.