NVIDIA GeForce GTX 590 3GB Review

Ever since NVIDIA announced the Fermi architecture, there were rumors of a possible dual GPU card somewhere on the horizon but these were quickly squashed amid concerns over power consumption and heat production. Those early Fermi GPUs weren’t exactly prime candidates for use on a card that mirrored the GTX 295's initial dual PCB design and a single PCB setup also brought a number of challenges to the table. While NVIDIA pondered their situation, AMD was able to forge ahead with Hemlock; a card that became known as the HD 5970. Ever since the HD 5970 was introduced, AMD has stubbornly held onto the fastest graphics card in the world crown. That may be about to change though.

NVIDIA has been hard at work refining their Fermi architecture. The result has been a series of products that offer higher performance per watt and almost none of the rampant thermal issues that characterized their predecessors. Through the use of some careful design evolution, the GF100 core has morphed into the successful GF110 and it’s this “new” core that was picked for use in NVIDIA’s reentry into the dual GPU market: the GTX 590

Internally code named Gemini, the GTX 590 blends a pair of fully enabled GF110 cores (each with 512 CUDA cores) with 3GB of GDDR5 to produce what’s billed as the fastest DX11 card on the market. Naturally some sacrifices had to be made in terms of clock speeds – we’ll get into those later - but it makes up for any possible shortcomings with a long list of features. This is NVIDIA’s first official product that supports Surround multi-monitor setups from a single card. Also, the dual GPUs can be set up in such a way that one can process CUDA (PhysX, Folding@Home, transcoding and the like) while the other goes about rendering in-game scenes. In our eyes, the potential here is almost limitless.

Alongside all of the usual marketing points like PhysX and 3D Vision, the GTX 590’s true goal is to compete with AMD’s own Radeon HD 6990. In NVIDIA’s eyes they are doing this on several fronts and not only from a performance perspective either. One of their main intents was to offer a better overall gaming experience by decreasing the acoustical profile and slimming down dimensions. In addition, the decision has been made to offer the GTX 590 at the same price as the HD 6990: $699. If this sounds like a tall order for a dual GPU design, that’s because it is.

The GTX 590 3GB is a card which has been theorized about for the more than a year now so it will be interesting to see whether it lives up to expectations.

The GTX 590 in NVIDIA’s Current Lineup

Naturally, the GTX 590 will initially take over the flagship position in NVIDIA’s lineup and will go toe-to-toe against AMD’s HD 6990. Since NVIDIA hasn’t had a dual GPU card since the GTX 295 was discontinued, there is very little to compare it to but the price alone can almost guarantee its place among some of the most expensive cards ever released.

Speaking of the price, things really are starting to look interesting. Dual GPU cards usually come with a hefty price premium over two cards purchased individually, but at $699 this new flagship model is very comparable to a pair of GTX 570s, which can be found for a little under $680. On the flip side of the coin, a single GTX 590 consumes significantly less power and takes up less space than two GTX 570 cards.

Specifications for the GTX 590 are about what we would have expected in several areas, but many will likely be surprised NVIDIA’s chose to use fully enabled GF110 cores. Both GPUs on this card come with 512 CUDA cores and their associated 64 TMUs and 48 ROPs in addition to 384-bit memory controllers and 1.5GB of GDDR5. Unfortunately, these high-end stats come hand-in-hand with increased power consumption and heat production when compared to the cores used in GTX 570 cards. As a result, some sacrifices had to be made in terms of clock speeds, which have been whittled down to a point that nearly identical to those found on the previous generation GTX 470.

When compared to the GTX 580, the processor clock has been reduced over 200Mhz which will likely have a profound impact upon in-game performance at lower resolutions, while the lower GDDR5 speeds could have an effect upon ultra-high resolution framerates. We can also assume that most voltages have been reduced as well, which will of course limit overclocking, while NVIDIA has also implemented their power capping technology in order to keep consumption in check.

Since the GF110 doesn’t support the mixed memory configuration built into NVIDIA’s GF116, there was no way to increase the allotment past 3GB unless consumers would be willing to stomach the cost associated with 6GB of GDDR5. However, in a market where AMD’s flagship card comes equipped with 4GB of ultra fast memory, NVIDIA may have gone a bit too far in trimming down this card’s specifications.

In short, it seems like NVIDIA has been backed into a corner by the limitations of their Fermi architecture but have made sensible cuts in order to meet certain goals. They could have been a bit more aggressive in terms of clock speeds, but the 512 cores and 384-bit memory bus will definitely help to balance things out.

3DMark 11 (DX11)

3DMark 11 is the latest in a long line of synthetic benchmarking programs from the Futuremark Corporation. This is their first foray into the DX11 rendering field and the result is a program that incorporates all of the latest techniques into a stunning display of imagery. Tessellation, depth of field, HDR, OpenCL physics and many others are on display here. In the benchmarks below we have included the results (at default settings) for both the Performance and Extreme presets.

Performance Preset