New G72 SKU: The GeForce 7200 GSTuesday 08th May 2007, 10:10:00 PM, written by Arun
NVIDIA officially announced the GeForce 7200 GS today, which remains based on the 90nm G72 core that also powers the 7300 and 7500 product families. It is incredibly similar to the OEM-only GeForce 7300 SE, but uses faster RAM. The catch? Half the pixel pipelines have been disabled, as well as one of the three vertex shaders.
The card is obviously not focused on gaming performance, as its specifications imply. It is more than enough for Windows Vista however, and could fit just fine in a low-budget HTPC if you're not willing to pay more for better acceleration and/or image quality. Also, just like the 7300 SE, its core clock runs at 450MHz; the memory is DDR2-800 instead of DDR2-667, however, with the quantity varying from 128MiB to 256MiB.
In fact, the 7200 GS has the same entry in the driver .inf file as the 7300 SE; the memory is the only difference! This also implies that it probably also supports PureVideo HD decoding acceleration, at least according to this page. As if it wasn't confusing enough that the 7200 is faster than the 7300 SE, though, it should also be noted that the 7100 GS is most likely faster in most real-world situations. No, we're not making this up.
It could be argued that it's strange NVIDIA is releasing a new 90nm-based SKU at this point in time (especially just as the 65nm shrink, G78, enters the driver's .inf!). This can be justified by the fact that they realized the importance of having such a low-end Vista SKU in the channel (for example, to replace IGPs in existing configurations where the motherboard cannot be changed), and that memory prices fell drastically in the last few months. As such, replacing the 7300 SE's 667MHz DDR2 by a 800MHz variant most likely made sense.
The bill of materials for these boards is, unsurprisingly, ultra-cheap. According to the DDR2 contract prices quoted by DRAMeXchange, the memory would cost about $5 USD for a 128MiB board, while the PCB is also extremely small (but harder to estimate, sadly). As for the chip, assuming a 90nm general-purpose wafer (price: ~$3000) and a 77mmÂ² die size for G72, more than 700 could be produced on each 300mm wafer.
It'd be surprising if less than 500 dies per wafer were usable given the amount of redundancy, and some of the chips would also go to other products with higher prices (such as the 7300 GS and LE), so you'd expect the per-chip cost for NVIDIA to be around $6 including packaging, and the cost to AIBs to not be much more than $10. These are all estimates, but they seem quite reasonable to us and are unlikely to be off by a ridiculous amount. Of course, this does not consider the AIBs and retailers' margins. In the end, the product makes sense for the market it's intended for, although we'll be more interested in what happens with G78 and G98 ourselves.
The full press release is available on NVIDIA's website, claiming immediate availability at OEMs, system integrators, retailers and e-tailers. A quick look around didn't turn out anything at e-tailers at the time of writing, however, so it will be interesting to see just how much lower than $50 the card will eventually sell for.
Related nvidia News
NVIDIA Fermi GPU and Architecture Analysis
NVIDIA's Parallel Nsight finally released
NVIDIA GeForce GTX 460 - GF104 breaks cover
PhysX87, ancient tragedy in 5 acts by RWT
So long, Chris, and thanks for all the fish
NVIDIA GF100 graphics architecture details
NVIDIA Fermi: new GPU architecture, starting with GF100
NVIDIA release OpenCL GPU drivers for Linux and Windows
NVIDIA GeForce GTX 275 at $250 to fight HD 4890