Next generation GPU NVIDIA "Kepler" (Kepler, code-named). A figure that has yet to see dimly.
NVIDIA is already leading to the notebook PC vendors, entry-level model generation Kepler "GK107" (code name. Here, GK is the same model number) has begun sample shipments. Also, early next year, corresponding to the PCI Express Gen.3 "GK106", and it has been notified that there is a major partner for the operation verification plan providing for the desktop PC.
However, NVIDIA is not communicating the details of the Kepler and even for those partners, at present, "code name and schedule, and performance targets on, began to talk" level only.
But from there, many officials had seen major changes in the product strategy at NVIDIA. Together around this time around let it.
Kepler lineup is becoming clear
GTX 580 is the successor to the end of 2012 or 2013 (gemeint ist wohl, dass der Nachfolger der 580 erst Ende 2012/2013 kommt)
Taken together the story of a leading OEM vendors, are named to the product roadmap for NVIDIA's Kepler generation, but the following four products.
GK110: Configure GK104 × 2, supports PCI Express Gen.3
GK104: Performance GPU, supports PCI Express Gen.3
GK106: Mainstream GPU, supports PCI Express Gen.3
GK107: Entry GPU (only OEM?), Supports PCI Express Gen.2
Kepler's name is "Kepler's laws" Johannes Kepler, known for such astronomers (Johannes Kepler ,1571 -1630) derived from the name of the
Multiple parties are close to NVIDIA "Initially, GPU core architecture of Kepler is only two" are aligned with the mouth. However, among the participants are getting the same information, "GK107 GK106 is the same GPU core, but aimed to differentiate itself by disabling some of the CUDA Core" and of that, "GK106 is compatible version of PCI Express Gen.3 GK107, GK107 although the meantime be provided as needed for major PC vendor to supply early, and eventually will be integrated or not, "and so on, opinions are divided It is that reality. However, support for PCI Express Gen.3 of GK106, "The outcome of the verification operation may be postponed," and noted that some participants will want to remember.
In addition, GPU for notebook PC, and power pop from that corresponding to the PCI Express Gen.3, expected to be delayed responses. PCI Express Gen.3 story becomes a story of the GPU for the desktop PC.
So, Kepler roadmap Importantly, as can be inferred from the model number, GF100 (= GeForce GTX 480) chips like big (at least for the time being), but that does not appear. 穿 (the sale) if the view was of the AMD "sweet spot strategy" Like, is not considered to have no potential to focus resources on product shipments even NVIDIA, actually, "the Kepler Architecture characteristics, which makes it difficult to introduce because of the big chip "(An official close to NVIDIA).
As mentioned earlier, NVIDIA has not revealed the details of the Kepler architecture to partners. However, the company is "in order to greatly improve the precision floating point performance in Kepler, aimed to optimize the architecture" that is described as turning, as well as strengthen itself CUDA Core, Streaming Multi- processor (SM) Graphics Processing Cluster and (GPC) but also signs revised configuration. Thus, "I heard that the operational efficiency reaches 1.5 to 2 times compared with the Fermi architecture" is the official valve OEM.
Another official close to NVIDIA's "CUDA Core of Kepler, which also heard plans to support integer arithmetic 64bit. Whether it is achieved in the first generation of Kepler is not known, fairly CUDA Core Strengthening saw would be more natural, "that he said. Kepler generation, rather than raising the number of CUDA Core, by strengthening itself CUDA Core, but it is aimed to raise the efficiency of operations per unit of power, but officials agreed view.
GPU Roadmap as of August 2011. The latest information on the 2012 Kepler, Maxwell in 2014 and has been
On the other hand, by increased efficiency through improved operations GPU CUDA Core, sister more and more power, and many officials were seen as difficult to manufacture a big chip. OEM official said, "GK104 power exceeds 250W" also points out.
In fact, NVIDIA is in a developing stage of PCI Express Gen.3, and has sought to incorporate into standard PCI SIG power 400W. Initially, the chip had been planned as a big-generation flagship Kepler, it is possible that the development has been postponed and power consumption issues.
In GTC 2010 was held in San Jose, California in September 2010, Jen's NVIDIA-Hsun Huang CEO is "at present is to slow down the cycle and a half years to two years the evolution of process technology, the GPU will need to plan for this cycle, according to the product lineup, "had said. In short, two years of the Kepler architecture portfolio is not market must continue.
The NVIDIA that point, the GeForce GTX 580, has a track record of achieving low power consumption of GeForce GTX 480 while the same process technology. 2013 from the end of 2012, big chip low power consumption technology aimed at the optimization of the process "GK102" or "GK112" (referred to by different Source ※) It is reported that there are plans to introduce , the story would say highly likely.
NVIDIA GPU Roadmap for Desktop PC
Furthermore, according to the specifications that are currently exposed to the OEM relationship, the amount of graphics memory or 3GB GK104 is 1.5GB, the memory interface (the same GF110 and GF100) 384bit. Single-precision floating point performance is expected to greatly exceed 2TFLOPS.
GK110 GK104 is equipped with two groups, but the clock kept to the problem of low power consumption, processing performance is still coming over the 4TFLOPS. And they mentioned his name GK112 GK102 Furagushippuchipputaru or earlier, is expected to be adopted by 512bit memory interface.
