NVIDIA GeForce GTX 480 and GTX 470 Review - GF100 and Fermi tackle gaming
The GeForce GTX 470 1.25GB Graphics Card
The GeForce GTX 470 is going to target the high-end gaming market with a $349 price point when it becomes available in April. This card uses the same GF100 GPU though with fewer of all the performance-inducing goodies enabled. The GTX 470 includes 448 CUDA cores, 56 texture units and 40 ROPs
- essentially this is the GTX 480 with another SM removed. Each group of four ROPs is associated with a single 32-bit memory bus connection so by removing 8 total raster operators NVIDIA has moved the GTX 470 from a 384-bit memory bus to a 320-bit memory bus with a 1.25GB (1280MB) GDDR5 frame buffer.
The clock speeds on the GTX 470 are noticeably slower than on the GTX 480: 607 MHz core clock, 1215 MHz shader clock and 837 MHz (3.3 GHz effective) memory clock. Memory bandwidth drops to 133.9 MB/s and the texture filtering rate goes down to 34.0 GigaTexels/s.
Power estimates (from NVIDIA) put the GTX 470 as a 215watt
graphics card with the same 105C thermal threshold. It is interesting that in NVIDIA's estimates here there is only a 35 watt difference in power consumption between the GTX 480 and the GTX 470 - our testing shows otherwise.
The GTX 470 is about an inch and half shorter in length than the new GTX 480 and seems use a less beefy heatsink. The fan speeds and noise levels were identical.
The entire outside of the GTX 470 is plastic where as the GTX 480 has a partial metal exterior for additional cooling assistance.
The back of the card looks much like the GTX 480; both new models from NVIDIA will apparently support two-way and three-way SLI.
The GTX 470 only requires a pair of 6-pin power connectors.
The external connections on the GTX 470 are also identical: a pair of dual-link DVI outputs and a single mini-HDMI port.
The heatsink on the GT 470 is smaller and lighter than that found on the GTX 480 - more so than I think a 35 watt power consumption difference would warrant.