In the HD 6990, each Cayman is configured to use all 1536 of its VLIW4 stream processors. The GPU core is clocked at 830 MHz, and memory at 1250 MHz (5.00 GHz GDDR5 effective). The 256-bit wide memory interface of each GPU is populated with 2 GB of memory, setting the total board memory to 4 GB. Display IO includes five connectors, including one DVI and four mini-DisplayPort 1.2. While the HD 6990 is said to outperform GeForce GTX 580, AMD is also bracing itself for NVIDIA's dual-GPU GTX 590, with HD 6990 OC variants following an official specification of 880 MHz core, and the same 5.00 GHz memory.
Two power two GPUs with over 2.64 billion transistors, you need a lot of power. AMD has engineered its PCB to deliver up to 450W of power. This should give you a rough idea of what the HD 6990 will draw with its PowerTune feature disabled. AMD is known for using high-quality digital PWM components on its high-end reference boards, the trend is kept up with. The HD 6990 uses two 4+2 phase PWM circuitry per GPU. The inductors and PWM chips are hand-picked to offer the least leakage and highest efficiency. The PowerTune technology throttles power to ensure the best energy efficiency. This is both a boon and a bane for enthusiasts.
Lastly, the business-end of the presentation. AMD will have relaxed CCC overdrive limits to let you crank up the clock speeds. With a typical max board power of 375W and a PCB designed for 450W, there should be some a decent overclocking headroom. In its comparison with NVIDIA products, AMD used the GeForce GTX 580, the green team's fastest graphics card. When put through batteries of DirectX 9, DirectX 10/10.1, DirectX 11, and OpenGL game tests, AMD claims its HD 6990 to be faster than GTX 580 by 67% average, and up to 110% faster in some cases. The HD 6990 OC variant is claimed to push that a little higher, with up to 8% increase in performance over the HD 6990 reference.
Source: DonanimHaber
No comments:
Post a Comment