NVIDIA GTX 670 Reviewccokeman - May 10, 2012
Category: Video Cards
» Discuss this article (55)
At this point NVIDIA is pumping out video cards based on the successful Kepler GK104 core with the launch of the GTX 690, GTX 680 and now the GTX 670, which is targeting the more middle of the road gamers. We saw that the GTX 680 was a potent product from the start that handed AMD's king of the hill single GPU Southern Islands-based HD 7970 a losing hand in just about every game and resolution tested. The GTX 680 proved to be cooler running and more power efficient, and directly caused a dynamic shift in the performance playing field when the price point was lower at retail than the HD 7970 was selling for. AMD countered with price drops to make the HD 7970 more competitive from a price for performance perspective. Now NVIDIA is going to go for the jugular with the launch of the GTX 670 and really drive the price performance ratio back to the green side of the fence, with the HD 7950 square in its sights.
The mantra for NVIDIA is to run cooler, more efficiently and deliver exceptional gaming performance; things it proved with the last two launches. By using the same SMX architecture (albeit one less SMX), the GTX 670 is supposed to be between 30% to 50% faster than the GTX 570 across a wide range of games. This makes a perfect time to upgrade for the person on that traditional (or seemingly so) three year upgrade cycle. Upgrading from a GTX 470 the user should see results up to twice as fast per the press release. We have seen what the last two GK104 cards are capable of, now let's see if NVIDIA can knock this one straight out of the park with a price point that will make the GTX 670 all the more appealing.
The packaging for the GTX 670 was nowhere near as special as the GTX 690 but is more indicative of a retail box seen at big box retailers. The black box has the NVIDIA logo embossed on the cover to identify what brand is inside. Standard retail packaging follows through the interior with a box for the accessories and the obligatory anti-static bag.
The GTX 670 uses the same 28nm Kepler SMX architecture introduced on the GK104-based GTX 680. The GPU consists of a series of GPCs (Graphics Processing Clusters), four in this case, on each GK104 with two SMX units each with 192 cores for a total of 1536 CUDA cores per GPU core when used on the GTX 680. The GTX 670 has had one of the eight SMX disabled effectively dropping the CUDA core count to 1344. To more effectively manage power consumption, the traditional method of running the shader clock at twice the core clock was abandoned and now the clock speeds run at a 1:1 ratio. Each GPC has a single raster engine and dynamically share 1MB of L2 cache. The GPU core features 112 texture units and 32 ROPs on the GTX 670. A new feature with GK104 is hardware and software-based GPU Boost technology, which dynamically boosts the clock speeds of the GPU cores when there is available TDP headroom, much like the latest CPUs from Intel and AMD. The base clock speeds for the GTX 670 are 915MHz with a GPU Boost core clock speed of around 980MHz. The GTX 670 memory subsystem is the same as that used on the GTX 680 with four 64-bit (256-bit) memory controllers handling 2GB of GDDR5 memory running at 1500MHz (6000MHz effective). To put it in simple terms, the GTX 670 is a GTX 680 with only seven SMX units.
While the reference board is fully capable of handling some spirited overclocking, NVIDIA's board partners, like ASUS, Galaxy, EVGA, Zotac, Gainward and more, will have a variety of factory overclocked offerings with their own custom PCBs. Below are a few mages that show off a few of these new cards.
Even though knowing what's going to be under the skin, we need to take a look at the GTX 670 from NVIDIA.