Intel Demos System Based on 48-Core Processor

[nebun]

this is a joke

 

[amnotanoobie]

[citation][nom]Komma[/nom]While these advances in multicore CPU research are very interesting, I don't understand the significance of them. We've also seen earlier tech demos of 80 cores or more already. But what place is there for such devices? [/citation]

You get to pack more processing power in a smaller footprint.

For simplicity's sake, let's say a company has 200 ATX-cased server pc's that serve for different functions (virtualization, db, web servers, etc). Then Intel comes along with a processor that could replace 4 of those boxes with 1. So if the said company upgrades, the 200 boxes could be replaced with 50 boxes. Less cooling requirements, less hardware to look after, and less power consumption. To a degree, it would be a cost saving measure for the company (unless if Intel ups the price though).


[citation][nom]Hargak[/nom]intel is planning ahead. left turn would be going with a new architecture. x86 has been around forever, if they changed it adoption would be rough and other would have to follow suit. so, more x86 cores is easier. someday regular software will take advantage of mass core technology.[/citation]

+1. Right turn is Itanium, and we know how well that went.

 

[ta152h]

[citation][nom]Hargak[/nom]90+% of the people here do not understand cloud computing or they wouldn't be commenting on gaming, programs not requiring more than 1 or two cores, This is a new type of processing technology like gpu parallel processing. which html5 and adobe and many others are taking advantage of. This will also force programmers to utilize multi-processor bandwidth versus singular which seems to be slow coming. Or quad core processors would be much more useful for ALL software. when you hit a wall with cpu speed turn right or left. they turned right, and are stacking the cores. intel is planning ahead. left turn would be going with a new architecture. x86 has been around forever, if they changed it adoption would be rough and other would have to follow suit. so, more x86 cores is easier. someday regular software will take advantage of mass core technology.[/citation]

x86 is an instruction set, not an architecture. The instruction set does have some effect on architecture, but, you only need to look at a K6, Phenom, Pentium 4, and Core i7 to see how dramatically the architecture can vary even with the same instruction set.

When you consider how many people use x86, and then consider it's more expensive to create, uses more power, and runs slower than the RISC core the decoders are attached to, it's got to be costing billions of dollars a year.

It's a pity Intel went with unproven technology like VLIW instead of RISC. If they had not, we probably wouldn't be plagued by x86 today. The scary part is, Intel has prostituted itself to the extent they are spreading the Black Plague to phones, and even GPUs. They went from hating it, to loving it? I guess someone has to.

 

[zaznet]

[citation][nom]dogman_1234[/nom]People who say the quadoctem core CPU is for home use or gaming/intenet/photoshop, really need to face reality. This CPU is a Server CPU for company like banks and high end retailers who get heavy traffic.[/citation]

It's actually NOT well suited for high traffic. It's suited for small chunks of data that require a lot of processing like decryption or genetic sequencing. You don't need this kind of CPU power for providing web pages or databases where network and data (hard drive) IO is more crucial.

 

[palladin9479]

SPARC's have been doing this type of thing for years. Not this high of density, but I suspect that is just 24 specially designed atom cores connected together. Looking at the power requirements and clock speeds combined with the cache amounts, its pretty obvious each core isn't particularly powerful. A modern i7 would run rings around this in a home system or other low thread count environment. But this thing would smoke most any current setup in a high thread count number crunching environment.

Depending on how they built the I/O bus to the main memory / rest-of-the-system (is there a single I/O bus to the CPU or multiple system connections) would determine the efficiency on server farm virtualization.

I relate this more to a SUN SPARC T2 style design, just with a CISC core vs a RISC setup. I'm still partial to RISC myself, just don't want Intel going about it cause they tend to be Apple like with their patents and business deals.

 

[figgus]

[citation][nom]CoryInJapan[/nom]{ehanger} "But Can it play Crysis?"But Can it Play Crysis 2? lol.Im sorry,I just had too.[/citation]

But can it play Google Pac-Man?

 

[twbg4cq]

I don't know about people that play games, but for the people out there that create games, something like this is absolutely awesome. Software like ZBrush would make use of each and every one of those cores, giving digital sculptors even more power for creating polygonal character and environment models.
In fact I think Zbrush supports something like 1000 processing threads or close to it.