NVIDIA's next-gen GPUs sure took their sweet time arriving, but the first of the Kepler crew is finally available in stores and its 28nm silicon
is just itching to show off what it can do. You may be wondering what
the 2GB GeForce GTX 680 brings to the gaming table, and whether it'll
put an end to AMD's free run at the top of the food chain. Well, NVIDIA
now claims it has "the fastest GPU in the world", with both lower power
consumption and a 10-40 percent performance advantage over AMD's single-GPU rival, the Radeon HD 7970,
at 1920 x 1080. How can it back up such a boast? Ultimately, everything
hinges on independent benchmarks (coming soon in our review round-up),
but in the meantime we need to look at NVIDIA's new architecture for
clues. Intrigued? Then head on past the break.
First, the GTX 680 has a GPU Boost feature that automatically increases the clock speed when the GPU isn't fully taxed. You see, GPUs are designed with a thermal power limit (TPL), and they often operate well beneath that limit -- the new boost technology simply leverages that TPL headroom to increase the clock speed, while still keeping the card running within its tolerances.
As explained in the video below, this is all done with real-time hardware monitoring and it means that NVIDIA no longer has to factory-set its cards to a fixed, 'safe' clock speed that will work even in poorly cooled environments. For example, the GTX 680 has a base clock speed of 1GHz, but GPU Boost can increase that by 10 percent when needed, while packaged overclocking utilities will let you notch up the clock / voltage curve even further to readily achieve 1.2GHz, so long as your over-sized fans or water pipes are doing their job.
Kepler also utilizes NVIDIA's new Adaptive VSync technology to keep your fragging sessions as smooth as possible. The tech enables the GPU to dynamically switch on VSync when frame rates are above 60fps -- to keep rendering in lock step with your monitor's refresh rate -- and switch VSync off when frame rates drop below 60fps in order to reduce the image stuttering that would normally occur were it left on. Lastly, Kepler can utilize an improved anti-aliasing algorithm, called TXAA, to smooth out onscreen polygons better than ever before without sacrificing gaming performance. See the results for yourself in the gallery below.
What does all that tech do in real world situations? Well, during GDC 2012, the company showed one GeForce GTX 680 card running a stunning demo of the game Samaritan that just a year ago demanded three GTX 580 cards just to play. With three times as many CUDA cores as the Fermi-based GTX 580 but greater efficiency due to the 28nm process, the Kepler hurtled through the demo with almost a quarter of the power and heat radiation.
Now, we're talking about a top-level GPU here, with sufficient guts to power four displays at once (twice as many as Fermi) and a hefty price tag to match -- a £429 confirmed MSRP in the UK and likely in the same $500 ballpark as the GTX 580 in the States. But of course, it won't be the only card in the new stack. NVIDIA's got plenty of capable cousins in the Kepler family tree that include not only other desktop GPUs awaiting their own no-doubt imminent launches, but mobile models too.
The company says we'll see Kepler in many laptops this year, and not just in purpose-built gaming rigs, either -- Jen-hsun's crew promises to put the 600 series in regular laptops and even Ultrabooks, and let us play Battlefield 3 on them. To that end, NVIDIA has been working with OEMs to design the most thermally efficient laptop chassis designs possible to best take advantage of Kepler's GPU boost capabilities. We're not convinced that a true Ultrabook can deliver a good gaming experience, but we're currently reviewing Acer's Timeline Ultra M3 with a GeForce GT 640M onboard, so we'll know soon enough if Kepler can keep that promise in a thin and light laptop. Stay tuned.
Source :endgadget
First, the GTX 680 has a GPU Boost feature that automatically increases the clock speed when the GPU isn't fully taxed. You see, GPUs are designed with a thermal power limit (TPL), and they often operate well beneath that limit -- the new boost technology simply leverages that TPL headroom to increase the clock speed, while still keeping the card running within its tolerances.
As explained in the video below, this is all done with real-time hardware monitoring and it means that NVIDIA no longer has to factory-set its cards to a fixed, 'safe' clock speed that will work even in poorly cooled environments. For example, the GTX 680 has a base clock speed of 1GHz, but GPU Boost can increase that by 10 percent when needed, while packaged overclocking utilities will let you notch up the clock / voltage curve even further to readily achieve 1.2GHz, so long as your over-sized fans or water pipes are doing their job.
Kepler also utilizes NVIDIA's new Adaptive VSync technology to keep your fragging sessions as smooth as possible. The tech enables the GPU to dynamically switch on VSync when frame rates are above 60fps -- to keep rendering in lock step with your monitor's refresh rate -- and switch VSync off when frame rates drop below 60fps in order to reduce the image stuttering that would normally occur were it left on. Lastly, Kepler can utilize an improved anti-aliasing algorithm, called TXAA, to smooth out onscreen polygons better than ever before without sacrificing gaming performance. See the results for yourself in the gallery below.
What does all that tech do in real world situations? Well, during GDC 2012, the company showed one GeForce GTX 680 card running a stunning demo of the game Samaritan that just a year ago demanded three GTX 580 cards just to play. With three times as many CUDA cores as the Fermi-based GTX 580 but greater efficiency due to the 28nm process, the Kepler hurtled through the demo with almost a quarter of the power and heat radiation.
Now, we're talking about a top-level GPU here, with sufficient guts to power four displays at once (twice as many as Fermi) and a hefty price tag to match -- a £429 confirmed MSRP in the UK and likely in the same $500 ballpark as the GTX 580 in the States. But of course, it won't be the only card in the new stack. NVIDIA's got plenty of capable cousins in the Kepler family tree that include not only other desktop GPUs awaiting their own no-doubt imminent launches, but mobile models too.
The company says we'll see Kepler in many laptops this year, and not just in purpose-built gaming rigs, either -- Jen-hsun's crew promises to put the 600 series in regular laptops and even Ultrabooks, and let us play Battlefield 3 on them. To that end, NVIDIA has been working with OEMs to design the most thermally efficient laptop chassis designs possible to best take advantage of Kepler's GPU boost capabilities. We're not convinced that a true Ultrabook can deliver a good gaming experience, but we're currently reviewing Acer's Timeline Ultra M3 with a GeForce GT 640M onboard, so we'll know soon enough if Kepler can keep that promise in a thin and light laptop. Stay tuned.
Source :endgadget
No comments:
Post a Comment