The GTX 570 is the little brother of Nvidia’s current high-end GTX 580. Like the original GTX 480, the GTX 570 has 480 shader cores active, and 32 disabled. Since it’s based on the GTX 580 GPU, however, Nvidia’s able to run the GTX 570 at a higher clock rate than the GTX 480. On the other hand, the memory controller is 320 bits wide, versus the GTX 480’s 384-bit controller. That’s still wider than the 256-bit memory bus on the Radeon HD 6970.
We’re not sure why this card isn’t faster than it is. In many ways, it’s two HD 6870s built onto one chip, but overall throughput may be hobbled by having only 32 ROPs and a 256-bit memory interface. Overall performance is nowhere near double an HD 6870, and the card beats the GTX 570 in only five of 12 benchmarks.
The Asus TOP edition of the GTX 460 adds its now-familiar DirectCU cooling engine and pushes the clock speed a full 100MHz higher than the stock GTX 460’s 675MHz. That’s a 15 percent overclock. That, plus the 11 percent memory overclock, says as much about the cooler design as the GPU.
At just less than $250, the Asus variant of AMD’s Radeon HD 6870 delivers excellent performance. While it is somewhat overshadowed by its newer cousin, the HD 6950, it costs considerably less than that part. If you do most of your gaming on a single 1080p display, this card won’t disappoint.
Call this the littlest Fermi. The Asus ENGTS450 is built on Nvidia’s GTS 450 GPU, a scaled-down version of Nvidia’s flagship architecture. The GTX 450 is its own chip, not a feature-reduced version of a large GPU. Asus’s GTS 450 card includes only a single dual-link DVI connector, plus VGA and HDMI outputs; only two can be simultaneously active. This is a newer design than the XFX HD 5770, launched in September 2010.
Nvidia’s latest GPU release, the GF110, is essentially a re-engineered version of the original Fermi chip, with the addition of a few tweaks. By re-spinning the original, the full potential of Fermi is now realized, with all 512 compute cores active. (The original GeForce GTX 480 had the same number of compute cores, but 32 of them were deactivated.) Besides that, the GF110 features other enhancements, like improved FP16 texture performance, which boosts the frame rate in scenes using high dynamic range (HDR) rendering. The new chip also clocks higher; reference cards run at 772MHz core and 1,000MHz memory.
The Asus ENGTX 560 Ti DirectCU II is that once-rare bird: a factory-overclocked card at the beginning of a GPU's life cycle. Once upon a time, you wouldn’t ever see an overclocked graphics card. Then they started to appear—usually when a particular generation of GPUs neared the end of its run. Today’s hyper-competition between AMD and Nvidia now dictates that overclocked cards come out of the woodwork as soon as a product launches. If it’s an epidemic, it’s one we like, because manually overclocking graphics cards is a headache and generally more perilous than CPU overclocking.
With its iteration of Nvidia’s GeForce GTX 560 Ti, Asus decided to revamp its DirectCU cooler. Like previous iterations, this second generation DirectCU II runs the heat pipes in direct contact with the GPU, rather than relying on a dissipation plate to transfer heat to the pipes. Asus suggests this is a more efficient way to move heat away from the chip’s hotspots. As with most modern premium graphics cards, Asus uses high quality components throughout, which increase the longevity of the card while minimizing electronic noise that can interfere with image quality.
Nvidia’s engineering teams have been pretty busy lately, reengineering and streamlining the previously inefficient Fermi architecture. We’ve seen the GTX 580 and GTX 570 released in recent months. Now it’s the sweet spot GPU, the GTX 460, getting the chip re-spin love. Dubbed the GeForce GTX 560 Ti, Nvidia is clearly hoping to recapture some of the thunder of the venerable GeForce 4400 Ti GPU from an earlier generation.
As the Maximum PC team noted earlier, the new chip, code-named GF114, now sports a full 384 shader cores, 1GB of GDDR5 and pushes the reference core clock speeds up to 822MHz. During the product briefing, Nvidia noted that the GPU has plenty of headroom for overclocking, and we’re seeing quite a few designs which push the clock speeds.
First up on the review platter is MSI’s N560GTX-Ti Twin Frozr II OC Edition. The OC Edition pushes the core clock speeds to 880MHz and the memory clock to 1050MHz. The twin fan design has been updated from the original Twin Frozr; it’s less bulky and heavy, plus it seems to be quieter than the original. MSI also claims that the board is less prone to warping when mounted in a vertical case and uses “military grade” components in its construction.
Once upon a time, factory-overclocked graphics cards with custom coolers shipped a few months after the reference cards were out. So, you’d typically end up with a pricier, slightly faster graphics card to replace the one you might already own—not a particularly cost-effective scenario.
Today, we’re seeing some customized cards ship at nearly the same time as reference cards. So it is with the Asus EAH6850 DirectCU. Asus takes its DirectCU feature, which runs the heat pipes directly across the GPU chip, pushes the clock speeds up slightly, and ships the card for roughly the same price as a stock HD 6850. Asus also bundles its voltage-tweak utility, which lets you change the voltage and push clock rates even higher, if you like.