Thank god we can finally develop new architectures and software now that Moores Law is dead.Now that memory is not a problem anymore we can get rid of RISC and fetch two 64 bit address instructions again.Wait, that does not make any sense!The terms RISC, CISC have no meaning in the 21. century, stop using them.Then what architectures is this man talking about? Specialized or general purpose? The AI chips are coming out now because their time has finally arrived, not because of new architectures. For CPUs I see no new architectures either. The ARM / mobile hype is baseless, there is no architectural revolution there.The kind of GPU architecture we have today is 15 years old.
The only way I see is to have way more threads. Not 4 but 12 doing what used to be one.
Very biassed graphic. Circa 2005+/-, processors turned away from increasing clock frequency to multicore / multiprocessing. Yet the SpecIntCPU graph completely / deliberately ignores the contribution of that revolutionary architectural change. The appropriate graph to use would be SPECIntRate ... unless perhaps one is in the "Liar's Figure" camp.
Patterson said "I can't believe RISC is still the best idea in processor architecture, but it is," he said in a brief overview of the last 50 years of computing. "Our modest goal is to become world-dominating and be the ISA you run every day."
It is the same old load, load, execute, store sequence that has been beat to death. It is the notion that an open source ISA will lead to some magical breakthrough that is creating the frenzy.
Hetereogenous MP and a CPU that does assignments, if/else, for, while, do statements will beat the socks off any RISC.
The C# syntax API already does 3/4 of the work and it is free. I have thought about starting an open source project, but since the whole world is brainwashed by RISC, why bother?
Someone please tell David Patterson that Moore's Law has not ended. "The ending of Moore's law and Denard scaling means that new innovations are needed in instruction set architectures ... I think we are entering another renaissance in computer architecture," said David Patterson.