Intel: 80-Core Processor Prototype

Intel has issued a press release announcing that its researchers have developed a prototype 80-core chip; Intel says that the real deal would take from five to eight years to hit the market. More cores at what cost? Not much, it seems. The chip-maker is claiming the 80-core device executes teraflop performance while dissipating less than 100 W of power (less than many of today's dual-core chips). The company's engineers will present a paper on the technical details of the 80-core processor at this week's (Sunday through Thursday) Integrated Solid State Circuits Conference in San Francisco. Stay tuned. Check the front page as more information about the CPU becomes available.

Intel's Sunday press noted that teraflops performance could make possible a wide range of new applications: "For example, artificial intelligence, instant video communications, photo-realistic games, multimedia data mining and real-time speech recognition--once deemed as science fiction in 'Star Trek' shows--could become everyday realities."

News source: InformationWeek

Report a problem with article
Previous Story

ATI's "R600" Graphics Cards' Image Transpires

Next Story

NeowinCAST Gaming - Episode #5

28 Comments

Commenting is disabled on this article.

hmm. 100W? with eighty cores? Sounds so good to be real

edit

The main attraction of the router and the network layout of the chip in general is that the PE can be replaced by anything, in general is that the PE can be replaced by anything, including an x86 core or a special purpose core (e.g. DSP or hardware encryption engine)

sounds a lil scary

80 cores? Thats a pointless goal. What good is having 80 cores in one processor if developers cannot get 2 or 4 cores to work at full capacity?

The last time I checked (and tried), parallel processing was a tough obstacle to overcome. Did we finally get dual and quad processing to live up to its potential? So much so that now we're being bottle-necked by lack of cores? No, we didn't. Good job Intel, spend your time developing hardware that we neither need, want, or can take advantage of.

spacer said,
80 cores? Thats a pointless goal. What good is having 80 cores in one processor if developers cannot get 2 or 4 cores to work at full capacity?

The last time I checked (and tried), parallel processing was a tough obstacle to overcome. Did we finally get dual and quad processing to live up to its potential? So much so that now we're being bottle-necked by lack of cores? No, we didn't. Good job Intel, spend your time developing hardware that we neither need, want, or can take advantage of.

Yes here here, let's just hold back progress and just stay where we are because it's good enough.

Maybe we should go back in time and tell cavemen to stop trying to make fire work since raw meat actually has more nutrients than cooked meat.

I mean come on, who really cares about stopping the spread of disease through heating the food, we're losing nutrients, vital nutrients.

So yeah, advancement is never important.

What's important to realize here is that the 80 cores in the processor doesn't have to be all x86-64 general processing cores. In fact, if you make these cores x86 cores, you probably can't fit 80 of them on the same chip. Say that you have a 12 core processor, you have make 6 of them x86 general processing cores, and the other can be graphics processing, media acceleration, security acceleration.. etc cores.

The most impressive fact about this chip is that each core can run at different speeds. So you can have your x86-64 cores running faster than your graphics processing cores.

"For example, artificial intelligence, instant video communications, photo-realistic games, multimedia data mining and real-time speech recognition--once deemed as science fiction in 'Star Trek' shows--could become everyday realities.”

Hehe... :p

I thought we already had real time speech recognition? AFAIK, it's not the CPU power that's the limitation when it comes to recognizing voice, but good enough algorithms, at least right now. You don't really see anything but recognition software that requires more than a fraction of the available power. I think the same holds true for "artificial intelligence" -- for example the AI "chatterbots" of today use anything but heaps of CPU, and is a quite abysmal performance still. Theories and understanding of AI holds of back still, not really CPU, or otherwise we'd at least have powerful AI systems that struggles within the bounds of processing power. We also already have "instant video communications"...

Jugalator said,

Hehe... :p

I thought we already had real time speech recognition? AFAIK, it's not the CPU power that's the limitation when it comes to recognizing voice, but good enough algorithms, at least right now. You don't really see anything but recognition software that requires more than a fraction of the available power. I think the same holds true for "artificial intelligence" -- for example the AI "chatterbots" of today use anything but heaps of CPU, and is a quite abysmal performance still. Theories and understanding of AI holds of back still, not really CPU, or otherwise we'd at least have powerful AI systems that struggles within the bounds of processing power. We also already have "instant video communications"...


i think they mean some AI that is actually intellegent, not something that has a certain amount replies to certain words.

CheeseFart said,

i think they mean some AI that is actually intellegent, not something that has a certain amount replies to certain words.

that was his point, and i agree with him. We just don't have enough understanding of intelligence to make intelligent systems. That has nothing to do with the processing power.

Sounds nice, but alas not much use when hardrives are as slow as they are right now and solid state drives don't give that much performance increase.

It's probably a CELL-like architecture, where they aren't really 80 fully-functional CPU cores by the usual standard. That probably means they'll be extremely efficient at a very specific set of tasks. Hopefully Intel's project will have more practical uses than the misguided CELL, though.

majortom1981 said,
well why cant this be real. the cell processor has 8 so 80 cores should be doable.
No the cell has 7 and they arent "cores" they are stream processors which are nothing like cores.

80 cores, but they are weaker as I remember, doesn't support a lot of technology, but still such many cores makes performance a way greater than nowadays CPUs...

This reminds me of what Reagan said about star wars back in the 80's. it caused such a stir that the russians spent about 75% out of every rubble on defense to make russia bankrupt.

I think Intel is playing games.