Mr_Tea 0 Posted November 30, 2004 In 2005 IBM along with some Partners will release a new Chip Generation called "CELL" Chip. It will be used in TV`s ,PS 3 and from 2006 in Home Servers. The new Chip will have Prozessor Power nearly like a Super Computer. Share this post Link to post Share on other sites
FSPilot 0 Posted November 30, 2004 Sweet. One step closer to complete world domination. ... mwahaha Share this post Link to post Share on other sites
bn880 5 Posted November 30, 2004 there we go, they are simply going more into parallel processing, ever remember me saying how we are currently in the dark ages of computing? I think this will be a big step towards mostly parallel processors in the future. Our problems will not be solved through high frequencies and minutarization alone, you can bet on that. Share this post Link to post Share on other sites
Mr_Tea 0 Posted November 30, 2004 And one thing is for sure. AMD, INTEL, IBM and all the other Companys have much better stuff in their Lab`s as we can buy today. Share this post Link to post Share on other sites
feersum.endjinn 6 Posted November 30, 2004 Everyone has been going into more parallel computing for last few years now, former way of rising CPU speed and transistor density has already hit point of diminishing returns a while ago... Sun, AMD, Intel, they're all into multithreaded and multicore CPUs now. Share this post Link to post Share on other sites
theavonlady 2 Posted November 30, 2004 One step closer to complete world domination. The new Chip will have Prozessor Power The Germans - again?! Share this post Link to post Share on other sites
shinRaiden 0 Posted November 30, 2004 Sigh... how many young grasshoppers have already forgotten the lessons of the beowulf masters... He who wishes to cross the south bridge must answer me these questions three... What... is the computer? What... is the network? What... is the probability that we'll see 64-bit floating-point map data in OFP2 with at least micrometer grade controlable precision across multiple defined terraingrid levels? Quote[/b] ]1 cpu + 1 cpu != 2 cpu. How does the computer decide what goes to #1 and what goes to #n, or #n-1 for that matter? That is an operation, and one that can be ignored in uniprocesser situations. Perhaps a closer analogy is better. C4 is t3h k3wl f3r n00bs, but shaped charges and stategicly placed demolitions are much better. Researchers are experimenting with a deformable flight-surface mod of the F/A-18, basically in n00b terms the wings flap. How does this all work? Do they have an uber-rack of a bazillion Itaniums using quantum prediction streaming data via broadband laser? no, they use a 4-pack of 68040's. If you go down to your local PC junk store you can pull one out of an old Apple Mac IIfx for $1.95. The key is two parts: 1) Properly engineered OS and related operational limits. You don't have time in-flight to exclusively defrag a gigabyte swapfile. You need raw sensor input and flight-control output, and you need it now. If you want a robust x.400 directory services system and secure network security infrastructure, you can get a quad-proc Xeon with a couple gig of ram and Windows Server 2003, or Linux on a single P4 with ~768mb of ram, or Novell Netware 5 on a P3 500 with 256mb of ram from the "free" box in the alley. There are all sorts of multimedia wonders that the BeOS can do on half the hardware of Windows because of how it vectors data differently. 2) Bandwidth, baby... This is something all can understand. No pipes, hi ping, all = ownage. This is why hypertransport and AMD moving the memory controller on-cpu has such wonderful results. Even the server manufacturers have a hard time putting obscene piles of memory on the boards, you run into problems with mobo traces and distance latencies. When you're working with traces the size of what run between the CPU and your RAM sockets (DIP sockets for the old schoolers), timing becomes a critical issue. Next time you crack open your PC, look closely at the mobo traces. Notice how those on the "inside" of the track have extra squiggles to make them about as long as the "outside" ones. That and you can only make so many traces per layer, and increasing the unber of layers reduces reliability and increases cost. If you wanted ultimate l33t bandwidth between your devices, ideally you'd want a massive mesh interconnect. That gives you two problems though - optical interconnects at the chip level are not ready yet, and that's a lot of routing to do and add in each component. Cray did it, but Cray could do pretty much what ever they wanted to as the government was footing the bill. Ok, let's say you have a rack of Cell blades with multiple direct interconnects. What about your apps? What do you plan to use it for... to play a single-threaded copy of NetHack in ASCII console mode? No, you'll be using it for something more fun like linking functions of massively paralell-executable code, such as particulate analysis of volumetric data (Real clouds in OFP3?) and so on. Imho these dying gasps by the legacy brute-force method of binary engineering is rather over-rated. What is interesting is the amusing potential that binary paralell systems have in simulating the complexities needed to execute quantum logic. Some times I fear that our 1D linear obsession with binary logic has crippled society's ability to truely expand into utilizing 2D vector solutions, let alone the possibilities of 3D quantum logic. Sometimes feeble attempts to enrich data (such as XML et al) fly by in a haphazard fashion oblivious to the flaws inherent of trying to mash it into a 1D linear regression. As noble as these efforts are, they barely even hint at the depth the binary shackles have on logical expression. Ignore the hardware for now, think of the logic. Transcend the bliss. Enlighten through assimilation. Be Borg. Share this post Link to post Share on other sites
bn880 5 Posted November 30, 2004 Everyone has been going into more parallel computing for last few years now, former way of rising CPU speed and transistor density has already hit point of diminishing returns a while ago... Sun, AMD, Intel, they're all into multithreaded and multicore CPUs now. Actually the movement is _extremely_ slow upto now, only done when absolutely necessary. While I have ides for totally different processor types which could be implemented given a new wave of programming methods could be developed. What I will tell you is this; we can do a lot more with what we have today, we do not need a brakethrough in technology to break some serious limits, we need a brakethrough in design, hardware and software... Oh yeah, life is great. (thought I'd throw that in for good measure) Share this post Link to post Share on other sites
shinRaiden 0 Posted November 30, 2004 Quote[/b] ]we need a brakethrough in design, hardware and software... No, we already have too many broken designs, shackled hardware, and fuzzy (fizzy? ...) software. Quote[/b] ]...only done when absolutely necessary... You're right there though in that stuff like Deathrack runs much better on a 286, and is unplayable on a 486. (too fast) I wish there was a suitable DE-accelerator so that the good old games would still be playable. Quote[/b] ]we can do a lot more with what we have today Layered interfaces with space-pucks. Seriously, the transcendent logic of it all... Better yet, get tennis elbow from reaching out and 'touching' your computer using a modded PowerGlove. Quote[/b] ]given a new wave of programming methods could be developed All compilers inevitablley lead back to assembly, which needs not be nearly as binary in logic as the data it manipulates. But to see the sea one must first stop looking through a straw. Share this post Link to post Share on other sites
feersum.endjinn 6 Posted November 30, 2004 This all reminds me about hype when PS2 came out - 3 separate specialized vector processors, running in parallel, capable of processing huge amounts of data. Well, we all know how great PS2 games look nowadays As for quantum computing, fundamental logic operations are still there, quantum computing only allows solving some specific problems faster than traditional computers. Things are still either true or false, quantum bits are 0 or 1, and quantum computing only tells you which state of quamtum bits is most probably the correct answer. Share this post Link to post Share on other sites
Colossus 2 Posted November 30, 2004 All we need now is a super RAM chip and mega good cable modem, and there will be no sutch thing as ping. Share this post Link to post Share on other sites
Ex-RoNiN 0 Posted November 30, 2004 Or, maybe, the publishers stop bugging developers with their deadlines and the developers can then produce more efficient code Share this post Link to post Share on other sites
bn880 5 Posted November 30, 2004 Quote[/b] ]we need a brakethrough in design, hardware and software... No, we already have too many broken designs, shackled hardware, and fuzzy (fizzy? ...) software. Quote[/b] ]...only done when absolutely necessary... You're right there though in that stuff like Deathrack runs much better on a 286, and is unplayable on a 486. (too fast) I wish there was a suitable DE-accelerator so that the good old games would still be playable. Quote[/b] ]we can do a lot more with what we have today Layered interfaces with space-pucks. Seriously, the transcendent logic of it all... Better yet, get tennis elbow from reaching out and 'touching' your computer using a modded PowerGlove. Quote[/b] ]given a new wave of programming methods could be developed All compilers inevitablley lead back to assembly, which needs not be nearly as binary in logic as the data it manipulates. But to see the sea one must first stop looking through a straw. Haha, okay there shinRaiden... take a deep breath. I don't know what you are going on about besides bandwidth, I'm serious about what I say, and I do have a background in processor design.... Share this post Link to post Share on other sites
Mr_Tea 0 Posted December 1, 2004 One step closer to complete world domination. The new Chip will have Prozessor Power The Germans - again?! The Germans - again? Again - what? Dominating the World? rofl Year 1989 1991 1994/ 95 1999 2006 2010 Troops 495.000 487.000 370.000 335.000 285.000 250.000 Share this post Link to post Share on other sites
cozza 24 Posted December 1, 2004 the german's will make robots and Nuke LA and a terminator will be sent back in time to kill the leader of the resistance in LA (Ralph Wiggum ) but I was sent back in time to keep him alive and save the world. Share this post Link to post Share on other sites
D34N 0 Posted December 1, 2004 And one thing is for sure. AMD, INTEL, IBM and all the other Companys have much better stuff in their Lab`s as we can buy today. Intel Surely doesn't, hell the just changed from their "More GHz!!" plan, If I had to go with a processor by Intel I'd choose their 2.2Ghz Dothan, as it is much faster then 3.0GHz P4's and on par with AMDFX series for $300 Cheaper. Edit: Linkage IBM (Research), Sony (Research) and Toshiba are investing billions of dollars to develop and prepare for mass production of Cell, which is a multicore semiconductor composed of several processors that work together to handle multiple tasks at the same time. Billions?!, yea it better carry out 16teraflops per second. Share this post Link to post Share on other sites
Albert Schweitzer 10 Posted December 1, 2004 Okay, I know that the cell chip will be about 10 times faster than current CPUs, but do you seriously believe this is fast enough for OFP2? Or even to play OFP Resistance in medium high settings? I doubt that! I guess we have to wait for CELL II . Share this post Link to post Share on other sites
