11-21-2010, 08:21 AM

Intel: 1000-core processor possible
Intel: 1000-core processor possible | Processors | Macworld

1000 cores? Geee. Then each app and each process could have it's own dedicated core. I think would be pretty sweet.

11-21-2010, 08:35 AM

Not impressed a bit. I would rather take a 10 fast processors than 1000 lame speed processors. Whatever happened to developing fast, single processors anyways? The Mac Pro is a great example of speed vs. multi core advantages.

11-21-2010, 09:30 AM

I think even 10 fast processors is not the solution. The information from the processors can only currently travel so fast to the screen and the hard drive and other places. So a kind of bottle neck in a way.

And I think of it this way. Sure this is a very simplified way of explaining things but you'll all get the point.

Say you have one chip with one core that can do or think about 10 processes a second. If that chip has 100 processes to think of or do. That means each of the 100 different processes gets a processes on the chip to work 1/100 or one every 10 seconds.

But say you take those 100 processes and spread them over a chip with 50 cores but each core can only do say 2 processes a second. Here you can spread out the load so each core gets 2 processes to think about. And that gives you each core doing 1 process a second.

Assuming all the processes are running at the same time, and time taken to communicate between the cores is factored out, though the single core chip is 5x faster then each of the individual cores on the 50 core core chip, the 50 core chip gets work done faster.

I know there is probably a lot of inaccuracies with my little story there. But my main point is there. Lots of "lame as you call them" cores might just be better then one beefy fast one. We just don't know at that silly number of cores intel is proposing. For me it's a wait and see game.

And secondly this 1000 number is at the moment just a theoretical limit. We might never see in the real world a 1000 core chip. But knowing your maximum theoretical limits is good. So you can use the data to work out a good balance between core speed and core number. And other things too. So yeah I won't ever be expecting a 1000 core chip but this info will surely make better chips in the future.

11-22-2010, 08:10 PM

Come on, 8th, you know you want your 1000 core GPU powering a holodeck! Just not an intel gpu!

11-23-2010, 02:40 AM

Originally Posted by DarkestRitual

Come on, 8th, you know you want your 1000 core GPU powering a holodeck! Just not an intel gpu!

I like people to be real and not just a hologram. Even though on the holodeck they'd feel and seem real. And I was thinking more like a robot butler. One programmed to do gardening too. I hate gardening.

Or make an electronic babelfish so I can understand every language on the Earth.

11-23-2010, 08:08 PM

Oh man a babelfish would be sweet. idrinor, I'm not sure what you're smoking, but the Mac Pro is a great example of high clock speed and high core count application. The super multicore chips are designed for low power consumption server functions, but would have applications across the board and help develop future higher power processors. When you're trying to compare a design that is aimed at the low power server market to a high power production market, that's just not meant to be.