Though official product pages have yet to be created, NVidia announced at its Tuesday press conference its latest RTX graphics cards for the mid-priced and extreme performance gaming PC markets. The presentation began with the usual self congratulation and delved a bit into underlying technologies before jumping straight into the RTX 3050, which with a sub-$250 MSRP was touted as the cheapest solution to support ray traced details in modern 3D games. No “Founders Edition” was shown, but instead a slew of manufacturer-specific designs and some basic specs, which include 8GB of GDDR6 memory.
Desktop gamers willing to sit through another 15 minutes of talk about NVidia Studios software, Omniverse API for 3D modeling suites, Nucleus Cloud simplified scene sharing for Ominverse, Ai for NVidia Canvass, 4th Gen Max-Q for notebooks, and even a new 3080 Ti notebook GPU would be rewarded with the big announcement that everyone was waiting for and insiders were denying: The RTX 3090 Ti.
The larger a piece of silicon the more likely it is to have defects, and NVidia has some of the largest chips in the consumer market. The binning process disables any defective portions of the GPU and sells what’s left as the model that corresponds to remaining features, but even the firm’s highest model spec tend to have at least some portion of the silicon disabled. The RTX 3090 for example came with only 10496 of its 10752 “Cuda” cores enabled, but there was certain to be some perfect parts produced, and Tuesday’s announcement lets us know what would become of those. The RTX 3090 Ti has the GA 102’s full 10752 cores, along with its full 336 texture mapping units (TMUs). It’s GDDR6 bandwidth is increased from 19.5 to 21 Gb/s, and it’s equipped with a 16-pin (!) power input to provide up to 450W!
See the full presentation at NVidia.