Comment Re:built for speed (Score 1) 232
I disagree. I think what they are talking about is with unlimited gates and power and time and money to develop, then you could use 40 or 50 stages with current ISA's, etc... yawn. Who the fsck cares? We're talking about the real world (something many researchers and accademics don't get too well).
In this world, we're seeing 5% increases for throwing 25% more gates at a problem. (Or worse), or often increasing design complexity by orders of magnitude. The rationalle is that gates are cheap/free, and Intel (and others) are buying into "he with the most designers wins" mentality.
At this point, I think we've gone far enough (or too far). If I have the choice between 2 x 970 cores, or one 40-50 stage nightmare core, I'm betting the 970's would take less gates, less to design, run faster, dissipate less heat, and be easier to write compilers for, etc. But in sheer penis measuring, the "Bigger is better" crowd would go for the deeper pipes.
Look, this marketing scam makes sense for Intel. They have the most designers and most money. They're trying to lure people into more and more complex processors, because that becomes a game of "he with the most money wins". Accademics love these kind of problems, because "he who wastes the most money on research grants wins". But in the commercial world, I don't think we should buy into it. I think more parallelism and more simpler processors is better. Our return on investments aren't that great just going for deeper pipes...
In this world, we're seeing 5% increases for throwing 25% more gates at a problem. (Or worse), or often increasing design complexity by orders of magnitude. The rationalle is that gates are cheap/free, and Intel (and others) are buying into "he with the most designers wins" mentality.
At this point, I think we've gone far enough (or too far). If I have the choice between 2 x 970 cores, or one 40-50 stage nightmare core, I'm betting the 970's would take less gates, less to design, run faster, dissipate less heat, and be easier to write compilers for, etc. But in sheer penis measuring, the "Bigger is better" crowd would go for the deeper pipes.
Look, this marketing scam makes sense for Intel. They have the most designers and most money. They're trying to lure people into more and more complex processors, because that becomes a game of "he with the most money wins". Accademics love these kind of problems, because "he who wastes the most money on research grants wins". But in the commercial world, I don't think we should buy into it. I think more parallelism and more simpler processors is better. Our return on investments aren't that great just going for deeper pipes...
