sijarvis

Distinguished
Sep 1, 2011
5
0
18,510
This is my first post on this site, so please be easy on me. I have already done a search for more information but found nothing on the forums that helped.

I read in 2006 that Intel had developed an 80 core CPU and were looking to release it commercially 5 years on. It's now 2011, so I was wondering if anyone had some up to date information about it.

I've also read that Intel believe they can get 100's of cores onto a single chip, so why stop at 80?

I know everyone is waiting for the i9 6 core chip to arrive but is that not going to be short-lived? superceded by this megacore monster.

So again, if anyone has some good solid information on this I would be most grateful. It might end up as computech masturbation but it's a topic i'd like to know more about before I update my computer. After all, I don't want to miss out on my own bluegene home equivalent by a matter of weeks/months.

 

sijarvis

Distinguished
Sep 1, 2011
5
0
18,510



Excellent....just what I was looking for.....now all they have to do is release an 8 socket cpu motherboard to match!
 

sijarvis

Distinguished
Sep 1, 2011
5
0
18,510


The real question is when are home computers going to be so powerful that I can run my heaviest of applications in real time full quality without having to render?

 
Just like IPC per mm in cpus, its only a matter of time before we see this in gpus as well, besides your OP, and what these types of fused chips can/will bring, which could even turn the industry
 

sijarvis

Distinguished
Sep 1, 2011
5
0
18,510


When do you think these Knights Ferry chips will be in the hands of the average consumer?
 


I would assume because of the fact that AMD is going with MIMD for its HD7K series instead of VLIW?



No. But there is Knights Corner which is set for a 2012/2013 release and will be a 22nm die shrink with 50 cores and quad HT. That might possibly be out to consumers.
 
Yes, the slow merging of gpu and cpu will eventually create these super chips
Currently, gpus are running into a memory wall, while cpu doesnt contain enough fixed function hardware as yet
Combining the two, with some sort of memory pool, and a large one, with possibly on die, or close to on die, this is whatll be the end result, and it should be what the OP and many here want
The on die solution pertaining to memory is both running into too slow and not nearly enough
Widening the bus will help some, but innercommunication between cpu and gpu will help, but we dont know how much yet, as this will improve both sides, again, by how much is important, as it will determine how much more, or less, we will need regarding memory and its speeds, or how wide we need to go, which as in everything, is cost and size