GeForce GTX 690 Review: Dual NVIDIA GK104 GPUs
On Saturday night at the GeForce LAN that took place in Shanghai, NVIDIA’s CEO Jen Hsun Huang unveiled the company's latest dual-GPU powered, flagship graphics card, the GeForce GTX 690.
As you've probably read by now, the GeForce GTX 690 features a pair of fully-functional GK104 “Kepler” GPUs. If you recall, the GK104 is the same chip that powers the excellent GeForce GTX 680, which debuted just last month. On the GeForce GTX 690, each of the GK104 GPUs is also be paired to its own 2GB of memory (4GB total) via a 256-bit interface, resulting in what is essentially GeForce GTX 680 SLI on a single card.
On GeForce GTX 680 cards, the GK104 GPU has a base clock speed of 1006MHz, with a Boost clock of 1058MHz. The GeForce GTX 690 has somewhat a lower base GPU clock of 915MHz with a boost clock of 1019MHz. The memory clock on the GeForce GTX 690 is unchanged from the GTX 680, however, and runs at an effective 6Gb/s. If you're keeping count, dual GK104 GPUs equates to 3,072 CUDA Cores (1,536 per GPU), 256 total Texture Units, and 64 total ROP Units. Total Memory Bandwidth for the card is 384.4 GB/s (192.2 GB/s per GPU) with a peak Texture Filtering Rate (Bilinear) of 234.2 GigaTexels/sec.
With those specifications, the GeForce GTX 690 should offer about 90% (or more) of the performance of a GeForce GTX 680 SLI setup, give or take a couple of percentage points depending on the application. We'll tell you for sure on the pages ahead.
The GPUs on the GTX 690 are linked to eachother via a 48-lane PCI Express 3.0 switch from PLX, with a full 16 lanes of electrical connectivity between each GPU and the PEG slot. Previous dual-GPU powered cards from NVIDIA relied on the company’s own NF200, but that chip lacks support for PCI Express 3.0, so NVIDIA opted for a third party solution this time around. For those paying attention, AMD has used PCIe switches from PLX on their dual-GPU powered cards.
With the kind of horsepower likely lurking under the GeForce GTX 690’s hood, it will obviously have some heavy-duty power requirements. But considering the GK104’s power efficiency and the GTX 690’s somewhat lower clocks they won’t be as high as previous-gen dual-GPU cards. NVIDIA is claiming a 300W TDP for the GeForce GTX 690 and the card a dual-8-pin power connector requirement. If you're keeping track, that’s over 20% lower than the 365W peak of the GeForce GTX 590.
The GeForce GTX 690 isn't just about speeds and feeds, though. NVIDIA has also done some innovating on the cooling and aesthetic fronts with the GeForce GTX 690 as well. Some of the new design features of the GeForce GTX 690 include an exterior frame made from trivalent chromium-plated aluminum, and a fan housing made from a thixomolded magnesium alloy, which offers excellent heat dissipation and vibration dampening. The GeForce GTX 690 also has highly-efficient 10-phase, heavy-duty power supply with a 10-layer, two-ounce copper printed circuit board.
The cooler on the GeForce GTX 690 consists of a pair of vapor chambers (one for each GPU), each with a nickel-plated finstack, and large center-mounted axial fan with optimized fin pitch and air entry angles. The GeForce GTX 690 also uses low-profile components and a flat, ducted baseplate for unobstructed airflow, which reportedly minimizes turbulence and help quiet down the card.
Just by looking at the GeForce GTX 690, it’s obvious NVIDIA has set out to create a powerful graphics card that also happens to look the part. We should also point out that the cooling hardware is designed in such a way that the fan blows air through the shroud, where half is directed towards the front GPU and ultimately expelled from the system, and the other half cools the rear GPU and exhausts within the case.
The GeForce GTX 690 should have more than enough muscle to push multiple displays, and as such, the card is outfitted with a trio of DVI outputs and a single mini-DP output, all of which can be powered simultaneously. Of course, the card supports NVIDIA's 3D Vision Surround technology as well.