Fujitsu announce FX10 supercomputer capable of 23.2 petaflops

Not content with being top dog once this month, Fujitsu has today announced the PRIMEHPC FX10 supercomputer. Its theoretical computational performance tops out at 23.2 petaflops based on deploying 98,304 nodes across 1,024 racks.

At its heart the FX10 uses a SPARC64 IXfx processor running 16 cores at 1.848GHz. Memory capacity tops out at 64GB with a bandwidth of 85GB/s or 5GB/s between nodes. In its highest configuration the FX10 can be running a total of 6,291TB of memory. As you’d expect, the operating system is Linux based.

www.geek.com...

The thing that is amazing about electronics s is how fast they progress. IBM Blue Gene/L was the fastest supercomputer from late 2004 to mid 2008. After an upgrade in 2007 it has a peak performance of 596 TFLOPS. The FX10 has a performance 40 times higher than Blue Gene/L. In four years the performance went up by a factor of 40.

The processors in super computers are nothing special. They are generally the same processors found in high-end desktop computers, for example, the 2nd fastest supercomputer at the moment, Tianhe-IA, uses processors based on the Fermi architecture (found in graphics cards) and Nehalem archetecture (first generation Core i-series processors). The main difference is they have thousands of them linked up via a high speed connection and they support some extra features (error checking memory, double precision floats which graphics cards don't need). The processors in the FX10 use the SPARC processors designed for servers however.

dude! that is BA. if i win the lottery im buying one and putting it in my house

reply to post by C0bzz

can be running a total of 6,291TB of memory.

I read somewhere that to describe all the connections in human brain takes 50 000 TB, so memory-wise, we are not very far. If the computing power progress continues, how long until we can simulate a mind?

Cant wait for this to be available to the public at low cost so i can install BF3 lmao.

reply to post by Maslo


As far as I know, we will probably have the computing power to simulate the human mind in less than 50 years. However, simply having the computing power doesn't necessarily mean we can create something that is alive and sentient. How would you program it? As far as I know, nobody has a clue.

What games has it got?

reply to post by Maslo


Processing power in computer chips doubles every 6 months. With the invention of higher read/write speeds on actual memory devices such as hard drives (Google Revo Drives), your theoretically looking at less than 10 years for a super computer capable of computing at a human level. As well as new RAM & transistor designs.

Now our brains are the first "quantum" computer ever built, so you're not going to see super computer that can actually compete with humans without us telling it what to do, therefore making us the superior 'computer" for probably the next 100 years until we have organic AI.

~Keeper

reply to post by C0bzz

However, simply having the computing power doesn't necessarily mean we can create something that is alive and sentient. How would you program it?

You dont need to program it - you simply map the neural network of the human brain (human neurome) and simulate it in silico. Sentience is an emergent property of our brains neural network.

reply to post by tothetenthpower

Now our brains are the first "quantum" computer ever built

Citation needed. "Quantum mind" hypothesis has no evidence. As far as we know, human brain can be simulated completely classically, with sufficient accuracy for the sentience to emerge.

reply to post by tothetenthpower


I think Moore's law is a bit longer that every 6 months :

Moore slightly altered the formulation of the law over time, in retrospect bolstering the perceived accuracy of his law.[18] Most notably, in 1975, Moore altered his projection to a doubling every two years. Despite popular misconception, he is adamant that he did not predict a doubling "every 18 months". However, David House, an Intel colleague, had factored in the increasing performance of transistors to conclude that integrated circuits would double in performance every 18 months

reply to post by PhoenixOD


In 1975 he was right.

Today it's 6 months lol

If you actually pull up specifications for models over the course of 6 months, you'll see the diference between the last gen and the new gen is actually probably closer to double.

reply to post by Maslo


No citation required. Your brain is a quantum computer. It operates independendly and changes based on observations.

It can actually learn in it's environment, without being given the instruction to learn.

There are certainly different theories regarding this, as Quantum Physics cannot be stated as fact, only hypothesis. What you get when you solve the question, you get 10 more questions instead of 1 answer.

The quantum mind or quantum consciousness hypothesis proposes that classical mechanics cannot explain consciousness, while quantum mechanical phenomena, such as quantum entanglement and superposition, may play an important part in the brain's function, and could form the basis of an explanation of consciousness. There are several quite distinct quantum mind theories, and these are discussed in the sections below. This school of thought is rejected by the majority of the quantum physics community.

en.wikipedia.org...

Originally posted by Maslo
reply to post by C0bzz

 

However, simply having the computing power doesn't necessarily mean we can create something that is alive and sentient. How would you program it?

You dont need to program it - you simply map the neural network of the human brain (human neurome) and simulate it in silico. Sentience is an emergent property of our brains neural network.

That remains to be seen. I think there is a lot more to creating something that is sentient than just having the right wiring. It would not surprise me if the process of development from an egg to a complex brain is what is a major factor in creating something that is self aware. Not to mention all the input that is fed into the brain from the senses during this process.

I also think the need for things like food play a big part in becoming self aware.

reply to post by tothetenthpower

Processing power in computer chips doubles every 6 months.

?

Transistor density doubles approximately every 18 months, but doubling transistor density doesn't translate into double the performance. So in reality processing power doesn't even double every 18 months, at best it doubles every two years with highly scaleable architectures like those found in graphics cards; often at the expense of power consumption, which is why modern graphics cards are more power hungry than ever before. The architecture in CPUs in desktop processors (x86) doesn't scale that well so in reality the power of each x86 chip doubles over much longer time spans, like every 3 years or so.

reply to post by Maslo

You dont need to program it - you simply map the neural network of the human brain (human neurome) and simulate it in silico. Sentience is an emergent property of our brains neural network.

Yes, but how do we simulate it? If it were so "simple" we would of already of done it with a far less complex brain and/or non-real-time.

reply to post by C0bzz


I'd disagree partially.

With tri-gate transistors coming soon with the sandy and ivy bridge processors, along with a drop to 19nm architecture an soon to be 17 or 15, would certainly classify a doubling in efficiency, perhaps not "capacity" per say.

You make a good point about GPU, but that lends to my theory regarding the increase in clcok speeds.

reply to post by tothetenthpower

Your brain is a quantum computer.

From your wiki link:

This school of thought is rejected by the majority of the quantum physics community.

It operates independendly and changes based on observations. It can actually learn in it's environment, without being given the instruction to learn.

Which is not a proof of quantum nature of it. Classical neural networks and evolutionary algorithms can do such things.

There are certainly different theories regarding this, as Quantum Physics cannot be stated as fact, only hypothesis.

Quantum physics is a fact, quantum mind hypothesis is not. Not the other way around.

reply to post by PhoenixOD


If you map the neurome of already adult brain, it will all be there. You will essentially make a copy of the persons mind. No need to repeat the development process of it.

reply to post by C0bzz


wait 18 months until they come out with one twice as fast!

Moores Law!

reply to post by Maslo


Touché

Well IMO the brain is a quantum computer, to me it makes sense, even if we haven't proven it, or won't for some time.

Thanks for the info though

reply to post by Maslo

Maybe once we move past Quantum Computing by a "class" or three.