Nvidia smashed it out of the park with its GeForce GTX 680, but that clearly isn’t enough graphical grunt for the green team. In a surprise unveiling at the Nvidia Game Festival 2012 in Shanghai, CEO Jen-Hsun Huang gave us its dual-GPU beast, the GeForce GTX 690.
The formula for building one of these super-powerful cards is surprisingly simple. Two of Nvidia's GTX 680 cores are soldered onto a 280mm-long piece of PCB, with a small chip between the two to allow them to work together.
As is usual in dual-GPU cards, the cores have been clocked down a little, with the 1,006MHz stock speed of the GTX 680 now standing at 915MHz. Nvidias Turbo Boost technology remains intact, though, so that core clock will adjust up or down depending on how much work the GPUs are doing. Adjustments are made every millisecond, and the GTX 690's 915MHz core will hit a peak of 967MHz when it's at maximum load.
Aside from the clock drop, little of the GTX 680 has been changed. Each core is still accompanied by 2GB of 6,008MHz GDDR5 RAM, for a total of 4GB across the board. There have been no architectural changes, either, with the eight huge clusters serving each core still packing 192 stream processors each. Across the entire GTX 690, that means there are a mighty 3,072 stream processors and just over seven billion transistors.
Thats a formidable amount of pixel-pushing power and, as expected, it translated to ridiculous benchmark results. At Full HD, the card clearly isn’t being pushed: its 73fps in our 1,920 x 1,080 Very High quality benchmark, for instance, isn't that far ahead of the HD 7970's 60fps or the GTX 680's 57fps.
Crank up the resolution and detail, though, and the two cores get to work. Running Crysis at 2,560 x 1,600 saw its score barely drop to 70fps; the GTX 680 and HD 7970, by way of contrast, ran through the same test at 42fps.
Tidak ada komentar:
Posting Komentar