I've not had a chance to read it, but it looks interesting and I'll be reading it tonight.
http://www.xbitlabs.com/articles/cpu/print/netburst-2.html
Mirno
http://www.xbitlabs.com/articles/cpu/print/netburst-2.html
Mirno
Seems very interesting... I'll read it tomorrow I think.
This article is extremly interesting. One year of reverse engineering.
And definitely this increase a lot the power consumption too. Man... I'm gona chage to AMD ;)
It seems that we definitely should reconsider and rethink all the programming paradigms, because not the "brute force" processing power is the problem, but the memory access and memory organization. Eliminate the "IFs" inside the loops, and try to access memory in linear manner to use the cache as much as possible.
Maybe it's time to switch to a such kind of computer, it's not x86 related, but I think it worth to read:
New Scientist Breaking News: Self-wiring supercomputer is cool and compact
At the same time, replay causes a number of operations to be executed multiple times, which will eat up processor resources.
And definitely this increase a lot the power consumption too. Man... I'm gona chage to AMD ;)
It seems that we definitely should reconsider and rethink all the programming paradigms, because not the "brute force" processing power is the problem, but the memory access and memory organization. Eliminate the "IFs" inside the loops, and try to access memory in linear manner to use the cache as much as possible.
Maybe it's time to switch to a such kind of computer, it's not x86 related, but I think it worth to read:
New Scientist Breaking News: Self-wiring supercomputer is cool and compact
Maybe it's time to switch to a such kind of computer, it's not x86 related, but I think it worth to read:
New Scientist Breaking News: Self-wiring supercomputer is cool and compact
It's not a surprise for people who already worked on embedded system/FPGA before. Anyway, this idea already been discussed indepth the Computer Engineering Handbook. But, still, the article is worth to read
It's not a surprise for people who already worked on embedded system/FPGA before. Anyway, this idea already been discussed indepth the Computer Engineering Handbook.
Well you're right, the ideea is not a surprize. But don't forget: one thing is the ideea, other is the implementation. Actually in that article not the ideea is the point, but actually the IMPLEMENTATION is the important thing. Somebody had the bravery to put "that" ideea into reality, that's the point.
Jules Verne wrote about the travelling to moon, in very detailed manner, but only after 100 years did the mankind achieve this. But the ideea was very old.
Those people who worked on embedded FPGA systems knows that this reconfigurable cluster is a hell lot complicated work (otherwise somebody would have done it already), which will run for several years. Not to speak about the programming part:
will develop software tools to enable programmers to create code for FPGA chips more easily.
Once the 64-node machine is built, the designers will try to transfer several existing supercomputer programs onto the new hardware using these tools
Once the 64-node machine is built, the designers will try to transfer several existing supercomputer programs onto the new hardware using these tools
This will be a major brakethrough in the FPGA based Hardware/Software CoDesign, because you can forget the classical Neumann paradigm also the structral and programming style. You can see that the whole thing is an experiment they don't know actually what to expect.
That's why I'm enthusiastic about this, because somebody (who has both financial and technical background) is bothering to do it.