Oh joy, Nvidia has gone and complicated its OEM graphics card line-up something terribly, as the company has announce three new GeForce 600-series models, yet there are no less than five new cards using three different GPUs. Talk about a slap in the face for consumers that buy branded computers, especially as three of the new SKUs are all the same model.The new models are the GeForce GT 630, GT 640 and GT 645, but two of the new cards are rebranded 40nm Fermi models whereas the other three are 28nm Kepler models. Starting from the bottom we have the GeForce GT 630 which is indeed a new card with an 875MHz GK107 Kepler GPU with no less than 384 CUDA cores, 32 texture units and 16 ROPs. The GT 630 will be available with 1 or 2GB of DDR3 memory with a clocks speed of 891MHz (1,782MHz effectively) and a 128-bit memory interface. It's a surprisingly small card and Nvidia has gone for a low-profile design with a small fan cooler and a DVI, HDMI and D-sub connector on the card. The maximum power draw of what appears to be Nvidia's baby Kepler card is an acceptable 50W.
Next up is the GeForce GT 640 and this is where things get really messy. The base model is a GK107 part with a lower GPU clock than the GT 630 at 797MHz, but everything else remains the same. This makes no sense at all, as why would anyone want a higher-end part on paper that presumably performs worse than a lower end part? The middle model is a Fermi GF116 part with 144 CUDA cores, 24 texture units and 16 ROPs. Here the GPU core clock is 720MHz with the shaders at 1440MHz. The memory clock remains at 891MHz, but this card has a wide 192-bit memory interface and as such it'll come with either 1.5 or 3GB of DDR3 memory.
The "high-end" GT 640 is the third and final Kepler GK107 card and although the CUDA core count, texture units and ROPs remain the same, the core has been clocked up to 950MHz. This card also has 1 or 2GB of GDDR5 memory which is clocked at 2.5GHz (5GHz effectively). The mid-range and high-end GT 640 models also have a higher power draw of up to 75W and the PCB is showing signs of an optional power connector. The base model sports a DisplayPort, HDMI and D-sub port whereas the middle and high-end models should come with DVI-I, HDMI and D-sub connectivity. That said, the PCB picture shows a DVI, DisplayPort and HDMI connector, so we're a little bit confused here, but we presume there's a typo on Nvidia's website where the two higher-end models should have a DisplayPort instead of a D-sub connector.Lastly and potentially the least interesting card out of the lot is the GeForce GT 645 which is a Fermi GF114 based card which appears to simply be a re-branded GTX 555 which is another OEM card. Once again it looks like a typo snuck in on the product page, as Nvidia states that the GT 645 has a 128-bit memory interface which is highly unlikely as the GTX 555 has a 192-bit memory interface and the specs are otherwise identical. Even the picture on Nvidia's website is labelled as the GTX 560 of which the OEM GTX 555 is a subset of.
It looks like Nvidia's retail models of more affordable Kepler models might not be too far away, although before that is likely to happen, we should see the GTX 690 arrive in a couple of days' time. We're not sold on this re-branding exercise by Nvidia, as it's way too confusing and the fact that the GT 630 appears to be the better option over the entry level GT 640 is just plain odd. The simple way to avoid all this confusion is to steer clear of brand name PCs and make sure you pick the graphics card you want in your computer.Source: Nvidia
Read more: Nvidia launches three new OEM GeForce 600-series models, in five SKUs by VR-Zone.com