Nvidia’s announcements at the 2025 Gamescom show were sprinkled with the usual minor technology refreshes across its software platforms, but the biggest updates are coming to its GeForce Now cloud gaming service.
The company upgraded its highest-end GFN server clusters («pods») used for Ultimate members with its new RTX GeForce 5080-equivalent graphics processors and is in the process of upgrading the network pipeline to deliver faster streams with less latency. In practice, that means a lot of new capabilities mostly, but not all, for the top tier subscribers.
Subscription prices are the same, for now, at least. Nvidia added Annual payment plans, though they’re not any cheaper at double the price of the six-month plan. One of the most exciting new features, available to both Premium and Ultimate members, is Install-to-Play, which essentially brings back a capability that was dropped between the original beta tests and the launch in 2020, which is the ability to play Steam games not already in GFN’s already optimized library. (If you’re not familiar with how GFN works, it’s not a game vault. It allows you to stream games you already own and the usual free-to-play games.)
Install-to-Play installs your games — still not all of them — to a virtual machine in the cloud, which is a much faster install than downloading to your device because it’s all through fat network pipes. It then streams to your device. They differ from typical GFN games since they’re not as optimized and they don’t get automatically updated the same way. I’m not quite sure how the updating will actually work.
Install-to-Play comes with 100GB of temporary storage that vaporizes once your session ends. Nvidia will offer persistent storage options for $3/month (200GB), $5/month (500GB) and $8/month (1TB). Aside from the convenience of not having to repeatedly reinstall a game and get over 2,000 more titles available, it theoretically makes it possible to stream games with local-only saves.
Nvidia also upped some of the top frame rate/resolution specs for the Ultimate plan. They rely on the Multi Frame Gen technology which is only available with the Blackwell-based 50-series GPU. Steam Deck players get a boost to a max of 90fps, and there is a new 1080/360fps option (which I believe requires an Nvidia Reflex-compatible monitor). Supporting LG TVs will be able to hit up to 4K at 120Hz and 5K for LG OLED monitors connected to Windows or Mac systems. (LG is the first partner for this feature.) Plus the company has added Logitech racing wheels to its list of supported devices.
Ultimate members also have the option for Cinematic Quality Streaming. CQS uses more bandwidth, up to 100Mbps, to deliver higher bit-depth HDR (YUV 4:4:4) and resolution (by autodetecting your screen’s native res) with optimized sharpening that better resolves text and reduces blur.
All of the above are slated to rollout starting in September. Later in the year, Nvidia plans to debut Play Instantly on Discord, starting with Fortnite. Play Instantly will let streamers invite viewers to launch the game as a one-hour trial directly from the stream for free without an account. This capability was in Google’s grand plan for Stadia, which never got off the the ground. Notable titles coming to GFN as they launch include Borderlands 4, Call of Duty: Black Ops 7 and The Outer Worlds 2.
Bits and pieces
Other announcements the company made at the show include the first implementation of its RTX Hair, in Indiana Jones and the Great Circle in an update to the game coming in September. RTX Hair uses path tracing and linear swept spheres (a fascinating new type of primitive) to produce more natural looking hair and fur with less of a performance hit than current techniques. Modding fans of RTX Remix will be able to incorporate path-traced particles into older games for more realistic rendering.
On the AI invasion front, Nvidia debuted The Oversight Bureau, a voice-driven puzzle game from Iconic that uses the company’s speech-to-text technology to feed voice commands into the game, which combines pre-recorded dialog and AI-based contextual analysis to decide which responses to use.
Nvidia made some minor optimizations to its new G-Assist AI app, reducing the amount of memory it requires and adding laptop-specific help, mostly for improving battery life. It also brought more user-requested features to the Nvidia App (which merged the GPU driver with GeForce Experience), adding DLSS global overrides and bring back some legacy 3D control panel settings.
