GeForce Now gets RTX-5080 equivalent and new AMD Ryzen Threadripper

The Ultimate class of Nvidia's GeForce Now is getting a major hardware upgrade. Users will soon be able to install games themselves.

listen Print view
Platforms with various play platforms

(Image: Nvidia)

4 min. read
Contents

Nvidia is introducing new so-called Superpod servers for its game streaming service GeForce Now. Those who have subscribed to the most expensive Ultimate subscription level starting at 22 euros per month will receive the performance of a GeForce RTX 5080 instead of an RTX 4080 in the cloud from September. Nvidia is not asking for a price increase for this.

At the same time, Nvidia is updating the processors from AMD's Ryzen Threadripper Pro 5000 (Zen 3) to Ryzen Threadripper Pro 9000. Although there are still eight CPU cores per gaming instance, these should now be significantly faster. Nvidia speaks of double the amount of RAM, but does not reveal the previous initial value.

The GeForce Now servers do not contain real GeForce graphics cards, but server models that are supposed to achieve similar frame rates. So far, these have been L40G models (Ada Lovelace) with an almost full expansion of the AD102 GPU – with even more shader cores than the GeForce RTX 4090.

The use of the RTX Pro 6000 (Blackwell) in the new server edition is an obvious upgrade: two game instances with RTX 5080 performance could run on one GPU. The 48 GB of graphics memory per instance – an RTX Pro 6000 has 96 GB of GDDR7 memory.

On average, a GeForce RTX 5080 is around 10 to 15 percent faster than an RTX 4080. Nvidia's comparison with a significantly higher gap is only possible with multi-frame generation (MFG), which inserts several artificial frames between the actually rendered ones.

Videos by heise

Meanwhile, Nvidia promises improvements in streaming latency, including new network processors in the servers (ConnectX-7). The 5G+ function L4S (Low Latency, Low Loss, Scalable Throughput) for latency reduction only works in the Deutsche Telekom mobile network in this country.

Nvidia aims to improve image quality with an efficient AV1 stream, an increase in the maximum transfer rate to 100 Mbit/s and YUV 4:4:4 scanning, among other things. Nvidia calls this comprehensive package of measures Cinematic Quality Streaming (CQS). In the future, GeForce Now will render the native resolution and suitable frame rates for the maximum refresh rate on more devices, such as 5K monitors and the Steam Deck. Racing simulation fans can use steering wheels for the first time, but initially only from Logitech.

Overview of Nvidia's quality improvements for GeForce Now.

(Image: Nvidia)

Around 2300 games are currently running via GeForce Now. Nvidia is expanding the portfolio with the "Install to Play" option. If the publisher allows it, subscribers to the Performance and Ultimate tiers can install games in the cloud themselves. This should be possible with 2200 titles at launch.

However, there is a catch: a maximum of 100 GB of storage space is available at no extra charge. The latest Call of Duty, for example, won't fit. In addition, users have to download and install their games in the GeForce Now Cloud every time they play. Even from cloud to cloud, this is likely to take a few minutes and is moderately energy-intensive.

Alternatively, Nvidia offers permanently usable storage quotas in which all installations and game saves are retained as long as users pay. 200 GBytes cost 3 euros per month, 500 GBytes 5 euros, and one TByte 8 euros.

Meanwhile, GeForce Now receives new games that also run without individual installation, including "Borderlands 4", "Call of Duty: Black Ops 7" and "The Outer Worlds 2". In the case of "Fortnite", Nvidia, Epic and Discord are testing game entry via the Discord app: if a friend is playing "Fortnite", you can join a game directly from Discord without having to start the game manually.

(mma)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.