Last updated 18 December 2024
It’s fair to say that Nvidia has dominated the graphics card landscape over the past two years, thanks to its industry-leading RTX 40-Series cards, and subsequent 40-Series SUPER refresh, but in the highly competitive graphics card industry where gamers, data scientists, AI developers and high-performance computing enthusiasts are all demanding incredible graphical performance and next-gen features, it was never going to be long before the rumours of the RTX-50 series started to circulate.
While there’s nothing concrete about the next generation of 50-series GPUs yet (including what models we will see, and no confirmed release date), there is a lot of anticipation and speculation around what the new cards will bring to the table, further fuelled by the recent Witcher IV trailer which was pre-rendered on an “unannounced NVIDIA GeForce RTX GPU”.
As soon as the cards are announced and we’re able to get them in, we’ll be looking to do some full testing and benchmarking, but while we’re waiting for an announcement from Nvidia, here’s everything we’ve heard about the 50-series.
When is the 50-Series release date?
So far there is nothing official about the release date of the 50-series, but with Nvidia’s CEO, Jensen Huang, set to deliver a keynote speech at CES 2025, a lot of the tech industry is anticipating a big announcement. Most likely, people are saying we will hear the launch of the 5080 and 5090 in January 2025, with the 5070, 5070 Ti to follow shortly after, and finally the 5060 and 5060 Ti, which would be in-line with previous release schedules.
How much will the 50-Series GPUs cost?
Any discussion around the pricing of the 50-series is pure speculation at this point, even more so than when the release date is or what the specifications might be. The original 4090 had a price point of $1,600 but high-demand soon pushed that price to over $2,000. There are rumours the 5090 could be priced between $2,000 – $2,500 making it one of Nvidia’s most expensive graphics card yet, but this isn’t currently based on anything direct from Nvidia, and even if it was, this could just be a way of Nvidia gauging the market, and not actually their intended pricing.
What is the architecture of the 50-Series?
The core details of Nvidia’s data centre Blackwell B200 GPU have mostly been provided, and whilst there will inevitably be some differences between the higher end data centre / AI chips and the consumer versions, history has shown that there are often similarities between the two, which allows us to take some educated guesses at the architecture of the new Blackwell GPUs.
Blackwel B200 is going to be using the TSMC 4NP node, which is an interesting decision given the newer, more refined N3 process technology is available. It’s not unusual for Nvidia to go with an older process node, they did the same with the 30-series, but if the consumer GPUs do follow suit, it will mean they’ll be using a refined version of the same technology that has powered their ADA Lovelace graphics cards for the last two years.
Beyond the process node, the rumour is that GDDR7 is going to be ready and in-use in the latest Blackwell cards. GDDR7 has target clock speeds of 36Gbps which is a significant speed boost over the GDDR6 and GDDR6X memory used in the current generation. The use of new GDDR7 memory could mean a significant boost to VRAM capacity for the next-generation GPUs.
We’re also expecting improvements to the ray tracing performance which has been a key focus of each new generation of RTX.
Recent leaks from Zotac and Acer suggest there may be up to five new GPUs (source) and they will feature improvements to DLSS technology, Ray Tracing and AI-enhancements, as well as the enigmatic ‘Neural Rendering Capabilities’ which are claiming to ‘revolutionise how graphics are processed and displayed’.
What are the expected specs of the 50-Series?
Some recent ‘leaks’ have hinted at the specs of the flagship 5090 and 5080 specs:
| GPU | ||
| SM Count | ||
| CUDA cores | ||
| VRAM | ||
| Memory bus | ||
| Memory clock | ||
| Memory bandwidth | ||
| Power connectors | ||
| TDP | ||
| DisplayPort version | ||
| Bus interface |
(source)
An Acer and Zotac leak seems to confirm that the RTX 5090 will have 32GB of GDDR7 memory.
| GPU | ||
| SM Count | ||
| CUDA cores | ||
| VRAM | ||
| Memory bus | ||
| Memory clock | ||
| Memory bandwidth | ||
| Power connectors | ||
| TDP | ||
| DisplayPort version | ||
| Bus interface |
The same Acer and Zotac leak suggests that the RTX 5080 will have 16GB of GDDR7 memory.
What does it mean for me?
Whilst it’s looking increasingly likely that we will see the announcement of some (if not all) of the Blackwell RTX 50-Series family at CES 2025, as it currently stands there are no official announcements from Nvidia. Although supposed leaks and industry informed best-guesses are probably going to get some things right, it’s not worth planning your system builds around what we know so far. It’s still unclear what the actual power consumption of the new cards will be or how many slots they will take up (with leaked pictures suggesting a massive heatsink may require 4-slots! Source), or what the cooling requirements may look like. What we do now is that the 5090 should be a significant step forward for graphical intensive tasks.
As soon as the new cards are announced and available, we’ll be looking to get them on our test benches and give you first-hand data on the best configurations for your application to maximise the performance of your new GPU.
To be the first to know when are test results are live, follow us on social media on X, Facebook and LinkedIn.
