Industry Information Station

Although the release is getting closer, but NVIDIA’s official is no accurate information about the GK104, according to the VR-Zone report, named the GTX GK104 680, but specifications and has great access to the message before, has not always been rumored core/Shader 1:1 mode, but changed back to the 2:1 model that has been used before, core frequency determined as 705MHz, stream processor frequency is 1411MHz. If it is true, 2:1 mode before it seems that NVIDIA still felt in better balance between power consumption and temperature.

Others still are 1536 stream processors, matching 2GB GDDR5 memory, video memory bandwidth 256bit, frequency 1502MHz (equivalent to 6008MHz), over HD 7,970 5500MHz in one fell swoop. Also alleged that GTX 680 calculation of capacity of 4.3TFlops, around 13.5% 3.79TFlops faster than HD 7,970, 3Dcenter GK104 posted the site performance has also been broadly confirmed this view.

Of score comparison chart below reflects a total of four indicators respectively from left to right, texture fill video memory bandwidth, ROP performance as well as computing performance. First three GTX 680 are no match for HD 7,970, backward and HD video memory bandwidth, and ROP performance 7,950, performance and HD 7,950 flat texture fill. However, computing performance, eye-catching, leading HD 7,970 per cent or so . Of course, these four parameters are individual properties, does not represent a comprehensive performance.

GTX 680 the actual launch date and the price is still unknown, it is learned that NVIDIA will open in March 12 launch Kepler technology in public information to be seen.

/Power house

Thank Burberry for delivery
News source: love Live mesh
In a few hours ago, announcing their new iPad, at the same time also have the latest A5X SoC processor, it has integrated a “4-core” display chip, is the main driving force for promoting the New iPad high resolution display.

These dazzling data always proved by comparison. So, NVIDIA’s Tegra 3 processor into the guns of the unfortunate. New iPad release, Apple gave us a table, column above the Tegra 3, A5 and A5X performance. Although Apple did not give us specific data. But we can look at the scores of Tegra 3 one-fourth only A5X:

Style has always been Termagant NVIDIA will not so easily accept. NVIDIA’s spokesman said today, and will use the fastest speed on the 16th day bought new iPad to test if the results showed a difference with Apple, the NV will start the claims process, request compensation for acts of Apple for misleading consumers. NVIDIA said: “we are very pleased that Apple uses our products to compare, but the real power is the need to pass the test to answer! ”

NVIDIA if you want to test is, of course, in the 16th, “buy” to a new iPad. But the little knitted felt, in cattle queue of the growing internationalization of the party today, NV would like to buy a new iPad on the 16th day, or do you want to ask my cattle do not agree with:

Apparently leaks of the latest figure shows the NVIDIA’s next-generation Tegra 4 to achieve a qualitative leap. VR-Zone said 28 nm technology will be used on the tips, called “GPGPU”, adding CUDA,OpenCL and other ongoing processes of next-generation image processing core. Add this simple means that core complex image processing technology updates.

Might be possible using earlier this month reported that Kepler image processing core. NVIDIA CEO Jen-Hsun Huang is mentioned on Kepler was applied to a smart phone, although there is no clear will immediately apply Kepler also progressive in Tegra 3 further consolidation on the NV40.

Some people speculate that the new Tegra 4 on either 32 or 64-bit core image are stronger than ARM Cortex-A15 faster. Many businesses such as Apple hopes will ARM Cortex-A15 into their own products, but did not see any improvement ARM Cortex-A15. NVIDIA hopes through the acquisition of Icera will lead the development of future mobile core.

Updating the Tegra NVIDIA has begun trying to establish an annual plan, and are likely to be announced this year that Tegra 4.

Thanks Zm post
News source: driving home
Slightly quiet Czech OBR recently during the last few days gives us a new and explosive materials: some users look forward to is the end Kepler core GeForce video cards will be a “BER” months of listing. As regards the OBR has more specific date cards, non-eleven-twelveths months but six months, but this, in my view there is not so reliable, the most optimistic projections of Kepler launch core of Tesla/8, the first batch of GK110 most of the professional card.

GK110 core specifications, OBR said although the core area is very big but the GK110 is smaller than the GF110. Integrated memory controller will rise to 512bit, standard 4GB GDDR5. OBR are sources, or bacling core performance about Kepler’s stronger than GeForce GTX 680 20-25%.

But after the head of NVIDIA Tesla GPU Computing business Sumit Gupta used to stand, paragraph II of the Kepler framework core (hereinafter referred to as Kepler II) compared to Kepler I that is, calculation of GK104 will be participating in a large number of new features, and double precision floating point calculations with a much higher capacity.

Professional super computer news site HPCwire speculated that Kepler II with 1.3-1.5TFlops double precision floating point capabilities, compared to previous Tesla Fermi architecture M2090 about 1 time, Kepler framework Tesla cards will remain around 225W TDP, or publish a paper on the GTC in May 2012. As to the actual production according to Gupta stand to wait until the fourth quarter, by the end of a large number of listed, from this point of view sources listed in the time portion of the OBR forward than the most optimistic guess.

Two days ago, NVIDIA has just released its flagship desktop video card GeForce GTX version 680, mark Kepler Kepler era officially opened, due to the new video card just unveiled, follow a comparative evaluation will flock to. One of the most important comparison is nothing more than an object have the fastest on the market today a flagship cards AMD Radeon HD GPU core 7,970, and many concerned friends have seen a single level of contrast, so many cards scrimmage results is what?

Recently, the Foreign Media Hardware.Info also carried out a comparative evaluation of GeForce GTX 680 and Radeon HD 7,970, however their test is not a single run of cards or dual SIM, but four composition 4-way SLI GTX 680 compared both with four HD 7970 CrossFire crossfire. Say no more, directly look at the results:

From this more before comparison, GeForce GTX single card configurations 6,801 straight ahead of HD7970, but in a four-way configuration mode, except for 3DMark 11, GTX 680 does not dominate the running test, run some benchmarks and gaming tested AMD HD7970 video card and even leading in more than 10%. In addition, the idle power consumption AMD WINS, but in the case of load lost to Nvidia. Foreign media say, pursuit of Dan Ka, you can directly select GeForce GTX 680, and four melee does not a good choice.

/Wei Feng network

Nvidia has been known for the video card, it now starts to enter the Tablet and a super computer chips. Last week Nvidia released a new computer chip architecture, committed to the framework displays on the desktop and Super performance and save energy. Nvidia CEO Huang Renxun said, will also drive down prices and the new structure of the company and makes some performance gains in the next few months.

Nvidia Tegra processors are the key weapons in the Android operating system, new products will be introduced in the spring of this year. Huang Renxun said, this summer, the commercialization of chip with enough components listed, as low as $ 199. He said: “raw materials as long as 150 dollars, such as expensive memory. Only $ 199, you can buy a Tablet from 7-11. ”

Apple iPad starting at $ 499. However, prices can just let Android Tablet iPad straight from the waist before you? Already $ 199 tablet on the market today, such as the Amazon Kindle Fire. Huang Renxun believes that Windows 8 tablet will make the situation even more different. He said: “the Andorid no as I hope the development. For many people, the PC remains the preferred. They wanted Windows to work well, you want to Outlook. A distribution has a Tegra processors Windows 8 tablet will be pretty good. ”

According to industry expectations, Windows 8 will be listed in the summer of this year. According to last week’s blog post, Sony submitted an application preparing Chromebook Vaio computer. Google Chromebook has little effect on the market, manufacturers are still working hard.

For T25 Sony Chromebook chip, this corresponds to the Tegra 2.5, which Tegra and Tegra 2 between the 3 versions. Zhiqian, Chromebook running Intel’s Atom chip. Tegra is used in the ARM architecture, it use less battery power.

Tegra 3 also has an important function is to display 3D graphics, this may be a strong response to the Apple.

Text/Sohu IT

Shopping malls such as battlefield, wanted to exterminate rival privately, but superficial to do, are also lie downin a “friends of” how, if that is the partner would have been more friendly, even contradictory rarely publicly complain, sharp criticism or accusation is even smaller. NVIDIA has not once complained about TSMC 28nm process problem, think the latter process of NVIDIA’s GPU release due to poor progress in delayed time and again, these complaints are fairly minor, but NVIDIA on TSMC is definitely “complain of deep, hate the cut.”

Privately complain that quite strongly, along with criticism of TSMC 20nm process in the future, think it’s not worth a penny (essentially worthless), they will not spend money on this process (there’s no money in 20nm), foreign media have recently published a detailed analysis of why.

These are NVIDIA ITPC meeting, held in November last year (International Trade Partner Conference) published on, should not be open to the public, because of taboo highly critical partners in the industry, such cases are very rare, XtremeTech Web site to see the content was shocked, do not know where they get the source.

NVIDIA first to ask what partners should do things, partners should understand the customer’s needs (note the underscore for emphasis), and to be open, candid and direct. (Perhaps this subtext is that TSMC early intentionally concealing the technology problems, or they think they are small problems, do not require communication with customers)

Development has always been radicals on NVIDIA video cards, per year (now once every 1.5 years to upgrade the schema) publishing graphics framework has significant growth, speed increase in the number of transistors, or even to more than Moore. In contrast GK104 of transistors only 3.54 billion, higher than previous-generation Fermi 3 billion chat a little.

(GTX performance than HD 7,970 680, but NVIDIA is also well aware of the subtle, leading the 3D performance is mainly due to the slightly higher frequencies, defeated on the GPU computing performance, in other words this is NVIDIA’s first leading competitors on the structure and size)

TSMC 28nm not in production last year, shipments of most GPU is 40&45nm technology

NVIDIA on the future technological development of TSMC to do analysis, 20nm process cost advantage compared to the previous generation of technology in reducing.

This is the cost of transistor size (cost-per-transistor) charts, 20nm process stages cost-per-transistor have no advantage.

There is a problem is the wafer costs are on the rise

Direct criticism of OEM partners have been very rare, and has done so many NVIDIA PPT data tell you partners can no longer the competitive production of next-generation technology, which is the creation of the first time. NVIDIA is really very angry, disappointed at the time, must find a vent out all grievances.

It has in the past have been for a long time, wonder if NVIDIA could calm, has recently released the GTX at least 680 on the performance, power consumption and cost should make him a little more easily.

Fab process upgrading is not so easy to do, not to mention TSMC will encounter problems, Samsung, IBM, GF even Intel has encountered problems, NVIDIA may wish to look at AMD is GF 32nm SOI do not temper the predicament, it didn’t feel so much better.

After this news issue, I wonder if NVIDIA rumor, after all, both sides are to cooperate, so revealing how much will hurt both sides now face, we see buzz is.

Text/Super network

NVIDIA CEO Huang Renxun (Jen-Hsun Huang) suddenly has a new idea. He said that at an NVIDIA chip in research time, Intel research and development investment was a waste of money. Everyone knows NVIDIA chips much better than Intel chips. They must consult Huang Renxun. Huang Renxun, Intel advanced semiconductor factories should put it to NVIDIA, Qualcomm, Apple and Texas Instruments chip.

Huang Renxun, Intel Mobile chip that does not require a waste of money to produce its own. As all Intel Mobile chip factory of the company is the best.

Intel spokesman spokesman qiaoen·kaweier (Jon Carvill), Intel’s manufacturing business was small. Their production focused on the Intel platform. Intel’s chip manufacturing technology has huge lead in 2012 and 2013. Intel’s current focus is to create your own products, rather than the competitor’s product.

Intel is making other companies to design chips. Earlier this year, semiconductor company Tabula said, Intel will use the most recent 22-nanometer production technology of Tabula 3D programmable logic chip.

However, Intel to enter the GlobalFoundries and TSMC dominant chip processing has a long way to go.

NVIDIA company officially releases next-generation architecture video, based on the core of the Kepler framework GK104 video card GeForce GTX 680 also released a mobile version of the Kepler framework graphics. GK104 has more than 3.5 billion transistor size, the core area of 294 square mm, just HD7970 80.5%.

The default core and memory operating at 1006/6000MHz GTX680, the default Pixel Fillrate as 32.2G/S, 128.8G/S is the default Texture Fillrate. Graphics memory bandwidth 192GB/S. 3.07T single precision floating point capability.

GTX680 video public price of 3,999 Yuan, and the previous generation GTX580 video just released the prices the same, at the same time lower than competitor AMD $ 4,299 HD7970.

Mobile GeForce 600M series video cards there are a total of ten models, the GT Series 7, GTX-series 3, they are: GT 610M/620M/630M/635M/640M LE/640M/650M and GTX 660M/670M and 675M, of course, this is only one of the first models, the future will certainly have more models to launch.

GT640M the following products all of which are based on the GT 500M replace the vests, more than GT640M products are based on the Kepler framework. 600M series thinner, lower heat, and can even be used in Super, greatly enhanced the Super game performance.

Anyway, enter after March NVIDIA next-generation 28nm process of the Kepler framework graphics is really close to the, not surprisingly legend of GK104 will be in the end of space travel at the end of this month, officially comes to Earth. A few days ago, VR-Zone foreign counterparts “accidentally” got a GTX 680 exam results table, which also joined the GTX 580, HD 7,970, and HD 7,950 as compared.

From looking at the test results, frequency-1050/5700MHz GK104 (GTX 680) to X3001 score in 3DMark 11 tests, comparing HD 7,970 default frequencies able to achieve a lead of around 8%, in the test of other games is also capable of delivering 5% per cent lead.

It is not clear whether the 1050/5700MHz frequency for GTX default frequency of 680, but from the results under homodyne GTX 680 is likely to reach even with HD 7,970 . GTX more than 680 default frequencies up to 1GHz to lead the HD 7,970 nor a thorny problem for AMD, HD 7,970 overclocking potential is obvious, after all, enhance the default frequency and GTX 680 tied again. Finally sing the words: not all claimed to be the clouds before publishing.

Another news VR-Zone enough to be regarded as The Rock,NVIDIA GeForce GTX Logo of long established Kepler published when synchronous will be replaced.

Brand new GeForce Logo

Currently the GeForce Logo

New GeForce GTX Logo font design somewhat square, especially the “NVIDIA” part of the difference with the past is clear, where “n” character from lowercase-became the capital of the fonts in the past. Does any of this is true? Time will tell us the answer.

/Power house