notlong ago, ET editor Jason Cross took a close look at AMD's new Radeon HD 4890 graphics card. AMD's latest graphics offspring is pitted against several cards in Nvidia's lineup, including overclocked GeForce GTX Core 216 cards and the more recently released GeForce GTX 275.
When the original 4890 shipped, ATI suggested that substantial headroom for both core and memory overclocking existed. They noted that, while the architecture was the same as the older Radeon HD 4870, a lot of tweaks were made at the engineering level to more easily dissipate heat, minimize voltage leakage, and other pesky issues. In other words, they were encouraging factory overclocked cards.
Today, we look at a pair of these cards: the XFX Radeon HD 4890 XT and the Diamond Multimedia Radeon HD 4890 XOC. Both cards use the AMD-designed, stock cooler. The XFX card clocks the core frequency up to 875MHz, a paltry 25MHz, and pushes the memory clock 100MHz, to 3.9GHz QDR (as compared to the default 3.8GHz). Diamond is a little more aggressive with the core clock, pumping it to 925MHz, as well as driving the memory to 4.2GHz QDR. Which of these cards are better? That's not a simple question. Setting clock speeds is often a delicate dance between core and memory clocks. If a mismatch exists between memory and core multipliers, you could see stalls in the pipeline as the GPU waits for memory to feed it data. Then there's the issue of cost: usually, highly overclocked cards cost more
When the original 4890 shipped, ATI suggested that substantial headroom for both core and memory overclocking existed. They noted that, while the architecture was the same as the older Radeon HD 4870, a lot of tweaks were made at the engineering level to more easily dissipate heat, minimize voltage leakage, and other pesky issues. In other words, they were encouraging factory overclocked cards.
Today, we look at a pair of these cards: the XFX Radeon HD 4890 XT and the Diamond Multimedia Radeon HD 4890 XOC. Both cards use the AMD-designed, stock cooler. The XFX card clocks the core frequency up to 875MHz, a paltry 25MHz, and pushes the memory clock 100MHz, to 3.9GHz QDR (as compared to the default 3.8GHz). Diamond is a little more aggressive with the core clock, pumping it to 925MHz, as well as driving the memory to 4.2GHz QDR. Which of these cards are better? That's not a simple question. Setting clock speeds is often a delicate dance between core and memory clocks. If a mismatch exists between memory and core multipliers, you could see stalls in the pipeline as the GPU waits for memory to feed it data. Then there's the issue of cost: usually, highly overclocked cards cost more
SOURCE:www.pcmag.com