Nvidia’s next generation flagship GPU is drawing ever closer to release. Epic Games showed off a demo of their Unreal Engine 3 tech using the Kepler-based hardware during the Game Developers Conference last week, which suggested the new GPU would be faster than at least two GTX 580s.
Despite no official announcement from Nvidia, the current top of the line product, the GTX 580 has received a price cut at several retailers, which is a sure sign that the new 28nm Kepler-based GTX 680 is imminent. The price reduction is usually a method employed by retailers to skirt around non-disclosure agreements in order to shift old stock and make way for the new arrivals. So it should be a good indication that we’ll be seeing the new card released shortly.
The GPU is rumored to feature 2GB of GDDR5 memory, is powered by the GK104 core and features a 256-bit memory controller with 1536 CUDA cores. A 4GB model is also rumored for future release.
What is interesting however is VR-Zone is reporting that the new model will retain the same core frequencies as the current Fermi counterparts, with the core clocked at 705 MHz and the shaders at 1411 MHz. Despite this, the GTX 680 improves on previous memory controllers, setting a new record with the memory clock at 6 GHz, which equates to a memory controller bandwidth of 192 GB/s, identical to the current GTX 580 even though it features a narrower memory bus.
It has also been speculated that the new model includes a dynamic clocking feature which operates similar to Intel’s turbo boost as found in their processors, enabling the graphics card to adjust clock speeds in response to performance requirements and operating temperatures.
Previous rumors suggest the GTX 680 will hit store shelves around the same time as Intel’s Ivy Bridge hits retail, with a suggested price around $500. Nvidia doesn’t typically comment on speculation, and has not replied to requests for comments, but should the end product match the rumors it looks set to regain the top spot from the recently released AMD Radeon HD 7970.
The NVIDIA GeForce GTX 580 packs 512 CUDA cores with a graphics/processor clock of 772/1544MHz, 1.5GB of GDDR5 memory with a 384-bit interface and a data rate of 4.0Gbps. The 10.5-inch, dual-slot card draws a maximum of 244W over one 6-pin and one 8-pin PCIe connector and carries two DL-DVI outputs with one mini-HDMI port.
The NVIDIA GeForce GTX 590 measures just 11", the GeForce GTX 590 boasts 8 Graphic Processing Clusters, 32 Streaming Multiprocessors, and 1024 CUDA cores. There are also 96 ROP (Raster Operations) units and 128 TAU (Texture Addressing Units). The GeForce GTX 590 is paired with two banks of 1536MB of GDDR5 memory clocked at 1707MHz (3414MHz DDR) for a total memory capacity of 3072MB.
The AMD ATI Radeon HD 7970 measures 27 cm long (10.6 in), the GPU core is clocked at 925MHz, and the GDDR5 memory operates at the same frequency of 1375MHz (5.5GHz DDR). Still, pairing that frequency with a 384-bit memory bus gives the HD 7970 264GB/s of theoretical bandwidth. The HD 7970 carries 2048 SPUs, 128 TAUs and 32 ROPs, up 33% from 1536 SPUs and 96 TAUs, it's also the first PCI Express 3.0-compatible card.
Downloads and Drivers
From the Forums
Subscribe to TechSpot
Receive a weekly update of our best features and tech news you don't want to miss: