Gainward used its own PCB, which is pretty clear once you see that its memory modules aren’t distributed as on the reference GTX 670 (picture below). Still, two 6-pin power connectors coupled with PCI Express power will be more than enough for the GTX 670. The reference card’s TDP stands at 170W, but the higher clocks on Gainward’s GTX 670 Phantom’s higher clocks should increase this number slightly. Reference GTX 670 cards are powered via two 6-pin power connectors, just like the factory overclocked Gainward GTX 670 Phantom. Namely, the PCB ends where the power connectors are placed. Things get interesting once you flip the GTX 670 over. As you can see, they more than just a bit different. GTX 670 Phantom is about 24.7cm long, which is about the same as the reference GTX 670.īelow is a picture of Gainward’s 670 2GB graphics card, which runs at reference clocks. Still, if you peek through the heatsink, you’ll see silhouettes of two fans. The fans are hidden behind the heatsink so it may trick users into thinking it’s a passive cooler. Gainwardov GTX 670 Phantom’s cooler is definitely a looker. Status monitor is a welcome addition to any tool. Info key reveals some important information about the card.
We must say it would be great if Gainward threw in a Frame Rate Target and sensor readouts for the second card in SLI chain as well.Ĭlicking Overclock G button opens up a new window, where users are free to set their OC parameters. It looks much better than the previous versions as well. ExperTool allows for overclocking Kepler based graphics, displaying sensor readouts or doing simple fan RPM control. With the launch of GTX 670, Gainward launched a new version of its ExperTool, which is now in version II. The company didn’t change memory clocks compared to the GTX 680, and it stands at 1502MHz (6008MHz effectively) on reference GTX 670 cards. Good news for Nvidia fans is that the company kept the identical memory system used on its GTX 680 card, meaning four 64-bit memory controllers (256-bit memory interface) and 2GB of GDDR5 memory. The card packs eight memory chips totaling at 2GB of GDDR5 memory. Coincidentally, reference GTX 680’s GPU runs at 1006MHz, which is the exact clock of our Gainward GTX 670 Phantom. While GTX 680 comes with eight SMX units and 1536 CUDA cores, GTX 670 has seven SMX units and 1344 CUDA cores. Note that reference clocks are 915/1502MHz for the GPU/memory. GTX 670 Phantom’s GPU ticks at 1006MHz while the memory is at 1527MHz (6108MHz effectively).
The series boasts some nice overclocks as well, not least due to quality cooling. Nonreference coolers are quite popular and Gainward’s Phantom series has always been a standout product in that respect. The other GTX 670 we mentioned is based on reference model and clocks. Today we’ll show you a special model – Gainward GTX 670 Phantom 2GB.
The actual pixel fill rate is also dependant on quite a few other factors, especially the memory bandwidth - the lower the memory bandwidth is, the lower the potential to get to the max fill rate.Upon Nvidia’s launch of GTX 670 graphics cards, Gainward already had two products to compete with. ROPs (Raster Operations Pipelines - sometimes also referred to as Render Output Units) are responsible for drawing the pixels (image) on the screen. The number is calculated by multiplying the amount of Render Output Units by the the core speed of the card.
Pixel Rate: Pixel rate is the maximum amount of pixels the video card could possibly write to its local memory in a second - measured in millions of pixels per second. It is measured in millions of texels processed in one second. The higher the texel rate, the better the graphics card will be at handling texture filtering (anisotropic filtering - AF). This number is calculated by multiplying the total amount of texture units of the card by the core clock speed of the chip. Texel Rate: Texel rate is the maximum texture map elements (texels) that can be applied per second. It especially helps with anti-aliasing, HDR and higher screen resolutions. The higher the bandwidth is, the faster the card will be in general. In the case of DDR type RAM, it must be multiplied by 2 again. The number is calculated by multiplying the card's bus width by the speed of its memory. Memory Bandwidth: Bandwidth is the max amount of data (in units of megabytes per second) that can be transferred across the external memory interface in one second.