With 2,688 stream processors and 7.1 billion transistors, the GeForce GTX Titan set new standards for GPU performance last March. The Titan’s 384-bit memory bus offers an astonishing memory bandwidth of 288.4 GB / s – 50% more than the GTX 680. The GTX Titan is in a class of its own and while a niche product, it is a niche product, the other top -Can handle single GPU cards and that is enough to get our attention.
A few months after releasing its $ 1,000 heavyweight, Nvidia unveiled a slightly scaled-down version as the GTX 780, which still features 7.1 billion transistors and a 384-bit bus with fewer stream processors and TAUs – changes that the Card 14% slower and 35% cheaper than the GTX Titan.
Traditionally, AMD and Nvidia board partners have overclocked the performance of most GPUs by 5 to 10%. So we fully expected one of them to try to fill the meager 14% gap between the GTX 780 and Titan. Palit answered the answer with his GTX 780 Super JetStream, which is clocked with 980 MHz and 1033 MHz Boost of 863 MHz and 902 MHz, 14% higher than usual.
So far we've only seen water-cooled GTX 780 cards pushed this far, while most air-cooled solutions maxed out at around 960 MHz, at least as far as factory overclocking goes. The GTX 780 Super JetStream isn't your ordinary graphics card, however, as its massive heat sink and three large fans keep its core cool under stress – a solution that allows the card to outperform the Titan, says Palit.
GTX 780 Super JetStream
The GTX 780 reference board is 10 inches long and the GTX 780 Super JetStream is slightly longer due to the improved cooler that extends to 11 inches. The display outputs remain standard, including two dual-link DVIs, an HDMI and a DisplayPort connector.
With 2304 CUDA cores, the GTX 780 has 50% more cores than the GTX 680, while the 3 GB memory buffer standard is also 50% higher than that of the GTX 680. Six 64-bit memory controllers form a 384-bit bus The specification from Nvidia combines the GPU with a GDDR5 memory that is clocked at 6008 MHz and offers a maximum memory bandwidth of up to 288.4 GB / s. Palit has updated this specification to 6200 MHz and enables a memory bandwidth of 297.6 GB / s, which corresponds to a slight increase of 3%.
The 12 SMX units with 2304 CUDA cores are clocked at 863 MHz as standard. With Boost 2.0, however, they can be clocked at up to 902 MHz in certain scenarios. The second generation GPU Boost technology works in the background and dynamically adjusts the graphics clock rate of the GPU to the operating conditions.
This is where Palit made the biggest changes in performance and increased the base clock to 980 MHz, while the boost clock exceeded 1 GHz at 1033 MHz. This 14% overclock should be enough to meet Palit's claims for titanium-like performance. The massive triple fan cooler from Palit keeps the GTX 780 cool at these frequencies. It has three fans in an 80mm, 90mm and 80mm configuration to pump as much air over the heat sink while making as little noise as possible.
Palit used its TurboFan Blade technology, which is inspired by the performance of the jet engines, which improves cooling performance by creating a strong airflow and air pressure.
Three fans are required because Palit used a gigantic heat sink or heat sink. A series of heat pipes are connected to a huge copper base to quickly transfer heat up from the floor and up through a series of fins. There are two large heat sinks measuring 20 x 9 x 2.5 cm which provide a huge area for heat dissipation.
There is also a large aluminum heat spreader on the front of the card that keeps the GDDR5 memory chip and 8-phase PWM cool. According to Palit, a unique 8-phase PWM design is used that reduces the maximum current load for each phase to stabilize the voltage level and improve overclocking ability. This design also provides an additional 30% power capacity which undoubtedly aids the overclocking process.
The external power supply configuration corresponds to that of a standard GTX 780, ie you will find a single 8-pin and a 6-pin connector. The GTX 780 Super JetStream has a TDP power of 250 watts, which is 28% more than the GTX 680. Therefore, Nvidia recommends using a 600 W power adapter.