The first product codenamed "Knights Corner" will target Intel's 22nm process and use Moore's Law to scale to more than 50 Intel cores.
Intel® Xeon® processors and Intel® Many Integrated Core architecture-based products to share common tools, software algorithms and programming techniques.
Products build upon Intel's history of many-core related research including Intel's "Larrabee" program and Single-chip Cloud Computer.
The share of the TOP500 list that features Intel processors grows to 408 systems, nearly 82 percent.
The platform is the successor to the Nehalem/Westmere architecture and integrates graphics directly onto the CPU die.
It also upgrades the Turbo Mode already seen in Core i5 and i7 processors to achieve even greater speed improvements.
The integration of GPU and CPU onto a single die (rather than separate dies within the chip package, as with Westmere processors) enables both to share on-chip cache RAM, enabling the GPU to work more efficiently on larger amounts of data
Sabre (Julian) 92.5% Stock 04 STI
Good choice putting $4,000 rims on your 1990 Honda Civic. That's like Betty White going out and getting her tits done.
Intel's current generation of Westmere parts marks the end of one era and the start of another. On the one hand, Westmere kicks off the 32nm process node, which is more important than normal because it marks the point at which you can make an x86 system on a chip (SoC) for every segment, from embedded up to servers. But Westmere marks the end of the P6 era—the architecture that has powered Intel's most successful processors since its debut with the Pentium Pro makes its curtain call with Westmere. In its place is a brand-new architecture, which Intel has codenamed Sandy Bridge.
In terms of its overall block diagram, Sandy Bridge will look familiar to any student of the P6 architecture. There's a similar mix of execution units, although the labor has been redistributed a bit (e.g., the AGUs are now general, and not specialized to load or store). And there's the same group of three main dispatch ports, with ports 0, 1, and 5 still hosting the actual scalar and vector math hardware.
But if you look a bit deeper, you can see that Sandy Bridge isn't yet another P6 derivative from Intel. It looks a bit like a fusion of the Pentium 4 architecture and the P6; or, alternately, you could say that it looks like neither. Whatever you call it, it's clear that Sandy Bridge is the first truly new microarchitecture from Intel since Atom, and it's the first new desktop microarchitecture since the Pentium 4. It's such a departure from the P6 lineage that we can say that the P6 line, which began with the Pentium Pro, officially ends with the current-generation, 32nm Westmere processors.
Sabre (Julian) 92.5% Stock 04 STI
Good choice putting $4,000 rims on your 1990 Honda Civic. That's like Betty White going out and getting her tits done.
Twin pointers with a bunch of memcpy all over the place sucks (pci bus transfers suck).
I'll be happy when you can write GPU-like massively parallel applications inside main memory space.
Should be much nicer for smaller scale mass parallelism (the kind that cares more about latency than just throughput).
If they can get main memory throughput to match GPU memory throughput, then I won't see a need for GPGPU anymore.
Without the main memory performance, depending on how much there is to compute, GPGPU still has an advantage for 'bulk' jobs.
But it's getting there. "One processor to rule them all".
