Regaining the #1 spot on the Top500 is merely a convenient side effect. The real reason for building these machines is that they are a key enabler for numerous science projects ranging from astrophysics to climate modeling to atomic phasefield simulation of crystal growth. This type of research can only be done on machines which offer Petabytes of RAM and Petaflops of performance. They cost hundreds of millions to build and operate. And if you can cut this cost down to a fraction by reusing Jaguar's existing housing, cooling and networking facilities, then this is financially a very clever move.
LLNL will receive their 20 PF machine dubbed Sequoia later this year. IBM's Blue Genes are known for their good ratio of CPU performance/network performance. This allows the MPI codes to scale well. The same is true for vanilla Cray XT5 and XE6 machines, but if upgraded with GPUs then each node receives a significant boost in computational power without increasing the network performance. This leaves the individual nodes bandwidth starved and makes it next to impossible to achieve peak performance in production code. The abysmal ratio of peak performance to actual production performance of China's Tianhe-1A tells the same story.
Maybe they'll achieve peak performance in Linpack, but for everything else Blue Gene/Q will be a much nicer system than Titan. Plus, on Blue Gene you don't have to deal with the heterogeneous system design, which already gave hell to coders on Roadrunner.
Johannes Schlumberger and others did some hacking on Mifare cards here in Germany. The University of Erlangen-Nuremberg uses them for wireless payments in their canteen and also for access control to sensitive areas. After notifying the manufacturer they didn't try to fix the problems, but threatened him with legal action -- even though it was a research project. As it says on Schlumberger's homepage: "Unfortunately I am not allowed to make my results public"
We're on an agreement on this: in all my examples the (from a technological point of view) better product did lose the market. The worse product won, either because it entered the market earlier or had the better marketing, or both.
It's really like the web: catalogs like Yahoo couldn't keep up with search engines, so why should you categorize your mails when there are search engines for them? Works well with mutt!
...of its own ridiculousness, they manage to pull off something even more stupid. These times the stuff you write on a postcard is better protected by German law than your private emails/data/whatever. It's like politician are afraid of the new technologies.:-/
Tilera's chips don't have FPUs, they're therefore no good for most compute intensive applications (e.g. scientific computing, simulations...)
We've seen on-chip networks, local scratch pads (read: caches) and FPUs on many previous chips. What sets this design apart is the combination of all three on a large scale. It's really like the IBM Cell BE, just not 1D, but 2D. Interesting.
Slashdot Mirror
Regaining the #1 spot on the Top500 is merely a convenient side effect. The real reason for building these machines is that they are a key enabler for numerous science projects ranging from astrophysics to climate modeling to atomic phasefield simulation of crystal growth. This type of research can only be done on machines which offer Petabytes of RAM and Petaflops of performance. They cost hundreds of millions to build and operate. And if you can cut this cost down to a fraction by reusing Jaguar's existing housing, cooling and networking facilities, then this is financially a very clever move.
LLNL will receive their 20 PF machine dubbed Sequoia later this year. IBM's Blue Genes are known for their good ratio of CPU performance/network performance. This allows the MPI codes to scale well. The same is true for vanilla Cray XT5 and XE6 machines, but if upgraded with GPUs then each node receives a significant boost in computational power without increasing the network performance. This leaves the individual nodes bandwidth starved and makes it next to impossible to achieve peak performance in production code. The abysmal ratio of peak performance to actual production performance of China's Tianhe-1A tells the same story.
Maybe they'll achieve peak performance in Linpack, but for everything else Blue Gene/Q will be a much nicer system than Titan. Plus, on Blue Gene you don't have to deal with the heterogeneous system design, which already gave hell to coders on Roadrunner.
BTW: GCGPU should be corrected to GPGPU.
Johannes Schlumberger and others did some hacking on Mifare cards here in Germany. The University of Erlangen-Nuremberg uses them for wireless payments in their canteen and also for access control to sensitive areas. After notifying the manufacturer they didn't try to fix the problems, but threatened him with legal action -- even though it was a research project. As it says on Schlumberger's homepage: "Unfortunately I am not allowed to make my results public"
We're on an agreement on this: in all my examples the (from a technological point of view) better product did lose the market. The worse product won, either because it entered the market earlier or had the better marketing, or both.
...or Windows vs. OS/2 or Betamax vs. VHS. To roll up a saturated market it's not enough to be /slightly/ better. You need something revolutionary.
Neodymium is called a rare earth, but actually it is no rarer than copper.
It's really like the web: catalogs like Yahoo couldn't keep up with search engines, so why should you categorize your mails when there are search engines for them? Works well with mutt!
...of its own ridiculousness, they manage to pull off something even more stupid. These times the stuff you write on a postcard is better protected by German law than your private emails/data/whatever. It's like politician are afraid of the new technologies. :-/
Tilera's chips don't have FPUs, they're therefore no good for most compute intensive applications (e.g. scientific computing, simulations...) We've seen on-chip networks, local scratch pads (read: caches) and FPUs on many previous chips. What sets this design apart is the combination of all three on a large scale. It's really like the IBM Cell BE, just not 1D, but 2D. Interesting.
RISC was always good at SMP. Back in the day, that was its biggest draw.
Virtually every CPU you get to buy today is a RISC design: even x86 chips are RISC designs. The front end translates the CISC into RISC micro ops.
With whom will it mate?
Maybe we could also add a start button to the console in Linux? At least in Linux 95.0?