The card — continued
The GTX 590's expansion slot covers are pierced by three dual-link DVI ports, a mini-DisplayPort connector, and as much thermal venting as the card's designers could muster. That mini-DP output supports DisplayPort 1.1a, so it's less capable in several ways than the DisplayPort 1.2 outputs on newer Radeons. Then again, those Radeons can drive only one dual-link DVI display natively; connecting more will require expensive adapters.
When it comes to truly extreme display configurations, Nvidia and AMD have taken different paths. The GTX 590's dual-link outputs will allow it to power a trio of four-megapixel monitors at once—or three smaller (~2 MP) monitors at 120Hz for wrap-around stereoscopic gaming via Nvidia's 3D Vision scheme. That DisplayPort output enables the 590 to drive four displays simultaneously, but only for productivity; multi-monitor Surround Gaming is limited to a maximum of three displays. Meanwhile, AMD isn't nearly as far down the path of cultivating support for stereoscopic 3D, but its Eyefinity multi-monitor gaming scheme will happily support six displays at once. The 6990 can do it, too, thanks to five onboard outputs and the possibility of connecting more monitors via a DisplayPort 1.2 hub. True to this mission, the 6990 also comes with more video memory than the GTX 590—2GB per GPU and 4GB total, versus 1.5GB per and 3GB total on the 590. It's up to you to choose why you get a headache: from wearing flickery glasses, or from trying to track objects across display bezel boundaries.
If you're looking for an indication of the differences in philosophy between Nvidia and AMD for cards of this ilk, look no further than the picture above. The GTX 590 is shown sandwiched between Nvidia's best single-GPU card, the GeForce GTX 580, and the massive Radeon HD 6990. The GTX 580 is a very healthy 10.5", the 590 is a considerable 11", and the 6990 is just a smidgen shy of a full 12". Although the GTX 590's space requirements are definitely above the average, the 6990 will be problematic in all but the deepest PC enclosures. AMD has aimed for peak extremeness. Nvidia has tailored its solution to be a bit more of a good citizen in this way, among others.
Another way the GTX 590 aspires to be easier to get along with? Acoustics. Superficially, this card doesn't look too terribly different from its rival, with a centrally located fan flanked by dual heatsinks whose copper bases house vapor chambers. However, Nvidia says the GTX 590 isn't much louder than the GeForce GTX 580—and is thus substantially quieter than the howls-like-a-banshee 6990. We'll put that claim to the test, of course.
Otherwise, the 590 has all of the sophisticated bits you might expect from a dual-GPU solution of this sort, including a 10-phase power supply with digital VRMs and Nvidia's familiar NF200 PCI Express switch chip, which routes 16 PCIe 2.0 lanes to each GPU and another 16 lanes to the PCIe x16 slot.
And, yes, there is an SLI connector onboard, raising the prospect of quad SLI configurations based on dual GTX 590s. The card will do it, but Nvidia wants users to be careful about the selection of components to wrap around such a config. It recommends a motherboard with an additional expansion slot space between the PCIe x16 slots, so there's adequate room between the cards for the interior one's fan to take in air. The firm is certifying motherboards that meet its qualifications for quad SLI, along with cases and PSUs. Right now, cases are the biggest bugaboo. Only three are certified—Thermaltake's Element V, SilverStone's Raven RV02, and CoolerMaster's HAF X—although more are purportedly coming soon. You could probably build a very nice quad SLI setup with some other popular full-tower cases and the right sort of cooling. Our sense is that Nvidia is emphasizing certification simply because it wants to ensure a good user experience and adequate cooling.
As you might expect, the GTX 590 will be priced at $699.99, exactly opposite the 6990. Cards should be available at online retailers starting today. Interestingly, you'll only find GTX 590 cards from Asus and EVGA available for sale in North America. In other parts of the world, the 590 will be exclusive to other Nvidia partners. My understanding is the cards have been divvied up in this manner because they're relatively low-volume products. It may have been deemed impractical to have six or more brands offering them simultaneously in one market. How low volume? When we asked, the firm told us it would be shipping "thousands of cards" worldwide. That's noteworthy because it's not tens of thousands—just thousands. That said, Nvidia expects a "steady supply available in the market." Perhaps the $700 price tag will ensure demand doesn't exceed supply over time.
One more thing
As you may know, the Radeon HD 6990 comes with an alternative firmware, accessible via a small DIP switch, that enables a configuration dubbed "uber mode" by AMD. The switch that turns on "uber mode" is the "Antilles Unlocking Switch for Uber Mode," or AUSUM, for short. Because this config exceeds the PCIe power spec and isn't guaranteed to work properly in all systems, it's essentially overclocking, though it's tacitly approved by the GPU maker.
We tested the 6990 with the AUSUM switch enabled, and that raised an issue of fairness. Nvidia hasn't given the GTX 590 any comparable mechanism, but the card can be overclocked in software. We figured, by all rights, we should test an overclocked configuration for the GTX 590, as well. One has to be careful here, though, because the GF110 chips will definitely reach much higher clock speeds when given enough voltage—we reached 772MHz at 1025 mV, similar to the GTX 580—but you'll also definitely bump up against the GTX 590's power limiting circuitry if you push too hard. The result, as we learned, is that performance drops with the supposedly overclocked config.
We eventually decided on a more mildly overclocked config in which the GPU core was raised to 690MHz, the GPU core voltage was increased from 938 mV to 963 mV, and the memory clock was tweaked up to 900MHz (or 3.6 GT/s). This setup was easily achieved with MSI's Afterburner software, proved quite stable, and, as you'll see in the following pages, performed consistently better than stock. The only thing left to do then was give these settings a name, since they lacked one. Folks, say hello to Wasson's Intrepid Clock Konfig, Extreme Dually—or WICKED. We've put WICKED and AUSUM head to head to see which is better.