OBR has just released the final specifications of both GeForce GTX 660 Ti and GTX 670 graphics card. Previous rumors were true, according this this data GTX 660 Ti will feature 6 SMX clusters, while GTX 670 will have 7 SMXs.
GeForce GTX 660 Ti is a mid-range graphics card featuring 6 SMX clusters with 1152 CUDA cores, 96 texture and 24 raster operating units. Card will be clocked at the same speed as GTX 680 — 1006 MHz. Also the memory will be clocked at 1502 MHz (6.0 GHz effective). The only difference is memory size and it’s interface. GTX 660 Ti has ‘only’ 192-bit bus width and 1536 MBs of memory buffer. Card will have a bandwidth of 144 GB/s and it’s overall single precision computing performance should be expected at 2.35 TFLOPS. This mid-range GPU will consume 150 Watts from power supply and $249 from your wallet :)
Moving forward to the most anticipated GeForce GTX 670, which lacks ‘Ti’ signature. It was also based on GK104 GPU, but it has only one SMX cluster disabled. This gives 1344 CUDA cores in total, 112 texture and 32 raster operating units. Card has slightly slower clocks than GTX 680, with a reference speed of 950 MHz. Memory is the same and is clocked at 1502 MHz, with 2048 MB of capacity and 256-bits wide bus. GTX 670 has computing performance measured around 2.7 TFLOPS. Maximum power draw should not exceed 180 Watts. Launch price was set to $399.
Both cards will likely be released simultaneously on May 10th. Today we shall expect official announcement of GeForce GTX 690, a first dual-gpu card from GeForce 600 Series.
|GeForce GTX 660 Ti||GeForce GTX 670||GeForce GTX 680|
|Transistors Count:||3.5 billions||3.5 billions||3.5 billions|
|Base Clock:||1006 MHz||950 MHz||1006 MHz|
|Boost Clock:||~1058 MHz||~1002 MHz||1058 MHz|
|Effective Memory Clock:||6008 MHz||6008 MHz||6008 MHz|
|Memory Size:||1536 MB||2048 MB||2048 MB|
|Memory Bandwidth:||144 GB/s||192 GB/s||192 GB/s|
|Performance:||2.35 TFLOPS||2.7 TFLOPS||3.1 TFLOPS|
|TDP:||150 W||180 W||195 W|