GeForce GTX 580 officially introduced

Santa-Clara, California, 09.11.2010. NVIDIA officially launched its new Fermi based graphic card — GeForce GTX 580

The difference between GeForce GTX 580 and early Fermi products is enhanced performance level and reduced power drain.

GeForce GTX 580 GPU provides the best graphic card performance in the world for DirectX 11 (DX11) games as well as supports NVIDIA 3D Vision™ Surround, PhysX, SLI and CUDA™ features. A combination of innovative vapor chamber cooling system and new architectural improvement makes this cars the fastest among single-core products, thus showing 35% increase in performance per watt and up-to 30% performance gain in 3D-applications compared with GeForce GTX 480.

GeForce GTX 580 features:

• 512 Cuda Cores and 16 PolyMorph engines
• Manufacture Suggested Retail Price $499 USD.
• Performance per watt efficency enhance up-to 35%
• New vapor chamber
• New level of realistic scene building — computational power amounts up-to 2 billion of triangle per second
• Complete creativity freedom in API DX11 game development

Jason Paul, product manager:

In March, we launched the GeForce GTX 480 — our flagship DirectX 11 GPU. Since then, we’ve been hard at work building its successor. Our goal was not only to create a faster GPU, but also one that’s quieter and more power efficient. Today I am pleased to introduce our new flagship graphics processor — the GeForce GTX 580 — our fastest and most power efficient high end GPU to date.
The GeForce GTX 480 shipped with 480 CUDA cores clocked at 1.4GHz. With the GeForce GTX 580, we increased the number of CUDA cores to 512 and cranked up the processor clock to 1.54GHz. Filtering for FP16 textures (commonly used in high dynamic range rendering) is now twice as fast. Z-cull, where invisible pixels are removed to avoid wasted shading, has been designed to be even more effective. The net result is that games on the GTX 580 run 15-20% faster than the GTX 480.

We didn’t stop there. With our goal of making GTX 580 both faster and quieter, we needed to improve power efficiency. To achieve this, we literally went “down to the metal”, pouring over every transistor on the chip to find ways to reduce power consumption and leakage. We re-designed the GPU using lower power transistors on less timing sensitive processing paths, and higher speed transistors on more critical processing paths. The end result was faster clocks with slightly less power consumption than GTX 480.

With the performance and power improvements at the chip level done, we set out to design a brand new thermal solution optimized for performance and acoustics. We got quite a bit of feedback from the community on acoustics with GTX 480, and the GTX 580 design took this feedback to heart. Our first step in improving the thermal solution was to move to a vapor chamber design. A vapor chamber looks similar to a regular heat sink, except its copper base is filled with a small amount of liquid that’s constantly going through a cycle of evaporation and condensation. The heat near the GPU causes the liquid to boil and evaporate. The evaporating liquid draws away heat at a rapid pace – up to ten times faster than straight copper. As the vapor reaches the surface adjacent to the heat sink fins, it cools, condenses, and eventually flows back toward the GPU where the cycle repeats again. While we are not the first to use a vapor chamber, we have been able to implement it very effectively with the GTX 580.

Along with the new heat sink design, we spent a lot of time tuning the GPU fan. We increased the stiffness of the fins to help minimize high frequency noise (heard as whining), so the fan sounds quieter at a given RPM. We rewrote the algorithm that controls how and when the fan ramps so there’s a smoother transition between different states.
The sum total of these improvements helped us bring the noise down substantially, making it not only quieter than the GTX 480, but also its predecessor, the GTX 285.
…

Specification:

Card gallery:

Source: NVIDIA