Up to this point, the GTX 690’s story has played out like a James Bond movie. Its specifications and final design were so secret, NVIDIA employees intimately involved in the project couldn’t exchange emails about some of the project's details. Rather, many things GTX 690 had to be discussed in person or not at all, lest a stray memo get set to the wrong external contacts. Board partners (some of whom tend to leak more than a fifty year old rowboat) only had a name to go by and didn’t even know what the card looked like. In this industry, secrecy is paramount and to their credit, NVIDIA kept everyone guessing about what their next Kepler-based card would be.
This past Sunday, the rumors were put to rest as NVIDIA introduced the GTX 690. By sporting a pair of GK104 cores and more rendering power than any other card, one thing was abundantly clear: this is how NVIDIA will take back the crown from AMD after more than two years of playing second fiddle. Granted, the GTX 590 offered great framerates and blazed a new, quieter, more refined approach for dual GPU cards but it couldn’t consistently beat the HD 6990, a graphics card that has been the world’s fastest for thirteen months.
Believe it or not, the reasoning behind this lack a top end halo product was straightforward: while the previous generation Fermi-based chips were impressively powerful, they were also power hungry and could heat a small house if given the chance. This limited NVIDIA’s options when trying to incorporate a pair of relatively inefficient cores onto a single PCB and posed a challenge that AMD is surely facing now with the GCN architecture. Kepler on the other hand focuses primarily upon architectural efficiency though the use of TSMC’s 28nm manufacturing process and by cutting away certain elements that were built into Fermi but aren’t of use to gamers buying GeForce-branded graphics cards.
The GTX 690 has benefited from NVIDIA’s new engineering approach since the Kepler-based GK104 core uses a fraction of the power and produces significantly less heat than its predecessor. As a result, the GTX 690 plays host to a pair of fully enabled GK104 cores, each with 1536 CUDA cores, 128 Texture Units, 32 ROPs and is topped off by 4GB of GDDR5 (2GB per GPU) operating at 6Gbps through two 256-bit wide interfaces. These specifications should look familiar to you since they mirror those found on the GTX 680, making this new card one of the only dual GPU solutions to use fully enabled cores.
While the GTX 690 may use two GK104 cores, NVIDIA has made some minor clock speed adjustments in order to meet a reasonable TDP value. The 1006MHz Base Clock (the core’s lowest frequency when running a 3D application) on the GTX 680 is now scaled back to a more modest 915MHz and the Boost Clock is down to 1015MHz from 1058MHz. The memory speeds have remained unchanged though. However, since the cores have enough TDP overhead and most games won’t push them to the limit, they will likely operate at or above the Boost Clock in most applications, resulting in performance that closely mirrors two GTX 680s.
Power saving measures on the GTX 690 may not be all that extreme but when added to the card’s 10 phase all digital PWM and copper-infused PCB, they have a significant impact upon overall power consumption and heat production. Instead of doubling up the GTX 680’s TDP of 192W –a number that’s already quite low in today’s high end GPU market- a GTX 690 boasts a TDP of just 300W. To give you an idea of where this stands, a single GTX 295 drew 295W while the GTX 590 sucked down an impressive 365W and the HD 6990 required about 300W.
NVIDIA hasn’t stopped at GTX 680 SLI-like performance either. Magnesium alloy, a complete lack of plastic, an LED illuminated logo and other details give the GTX 690 a build quality that befits an ultra high end product.
With leading edge framerates and a design that’s bound to turn heads, this card pushes the limit in nearly every way, particularly from a pricing standpoint. At a stratospheric $999 the GTX 690 certainly isn’t an impulse buy and yet, once you see its performance, you may think twice about dismissing it based upon price alone.
This past Sunday, the rumors were put to rest as NVIDIA introduced the GTX 690. By sporting a pair of GK104 cores and more rendering power than any other card, one thing was abundantly clear: this is how NVIDIA will take back the crown from AMD after more than two years of playing second fiddle. Granted, the GTX 590 offered great framerates and blazed a new, quieter, more refined approach for dual GPU cards but it couldn’t consistently beat the HD 6990, a graphics card that has been the world’s fastest for thirteen months.
Believe it or not, the reasoning behind this lack a top end halo product was straightforward: while the previous generation Fermi-based chips were impressively powerful, they were also power hungry and could heat a small house if given the chance. This limited NVIDIA’s options when trying to incorporate a pair of relatively inefficient cores onto a single PCB and posed a challenge that AMD is surely facing now with the GCN architecture. Kepler on the other hand focuses primarily upon architectural efficiency though the use of TSMC’s 28nm manufacturing process and by cutting away certain elements that were built into Fermi but aren’t of use to gamers buying GeForce-branded graphics cards.

The GTX 690 has benefited from NVIDIA’s new engineering approach since the Kepler-based GK104 core uses a fraction of the power and produces significantly less heat than its predecessor. As a result, the GTX 690 plays host to a pair of fully enabled GK104 cores, each with 1536 CUDA cores, 128 Texture Units, 32 ROPs and is topped off by 4GB of GDDR5 (2GB per GPU) operating at 6Gbps through two 256-bit wide interfaces. These specifications should look familiar to you since they mirror those found on the GTX 680, making this new card one of the only dual GPU solutions to use fully enabled cores.
While the GTX 690 may use two GK104 cores, NVIDIA has made some minor clock speed adjustments in order to meet a reasonable TDP value. The 1006MHz Base Clock (the core’s lowest frequency when running a 3D application) on the GTX 680 is now scaled back to a more modest 915MHz and the Boost Clock is down to 1015MHz from 1058MHz. The memory speeds have remained unchanged though. However, since the cores have enough TDP overhead and most games won’t push them to the limit, they will likely operate at or above the Boost Clock in most applications, resulting in performance that closely mirrors two GTX 680s.

Power saving measures on the GTX 690 may not be all that extreme but when added to the card’s 10 phase all digital PWM and copper-infused PCB, they have a significant impact upon overall power consumption and heat production. Instead of doubling up the GTX 680’s TDP of 192W –a number that’s already quite low in today’s high end GPU market- a GTX 690 boasts a TDP of just 300W. To give you an idea of where this stands, a single GTX 295 drew 295W while the GTX 590 sucked down an impressive 365W and the HD 6990 required about 300W.
NVIDIA hasn’t stopped at GTX 680 SLI-like performance either. Magnesium alloy, a complete lack of plastic, an LED illuminated logo and other details give the GTX 690 a build quality that befits an ultra high end product.
With leading edge framerates and a design that’s bound to turn heads, this card pushes the limit in nearly every way, particularly from a pricing standpoint. At a stratospheric $999 the GTX 690 certainly isn’t an impulse buy and yet, once you see its performance, you may think twice about dismissing it based upon price alone.

Last edited: