Significant changes are expected to accompany NVIDIA’s RTX 50 Series compared to the current generation, in particular the use of new GDDR7 memory, changes in the width of the memory bus, as well as an increase in the amount of VRAM. With perhaps also some unpleasant surprises.
NVIDIA is probably putting the finishing touches on its next generation of graphics cards, preparing the official launch of the first GeForce RTX 50 Series for CES 2025. As the fateful date approaches, the unofficial technical characteristics of a majority A growing number of models are gradually appearing on the web: today it is the video memory on board these new cards that interests us.
GDDR7: a major development
One of the most anticipated changes to this new range is the widespread adoption of memory GDDR7with the promise of higher transfer speeds than those allowed by the GDDR6/GDDR6X of the current RTX 40 Series. We are therefore talking about a speed of around 28 Gbps, or even up to 32 Gbps for the RTX 5080. Enough to effectively reduce bottlenecks and improve overall performance, particularly for high-definition games and bandwidth-intensive applications.
GDDR7 also introduces improvements in energy efficiency: consequently, reducing consumption would make it possible to maintain these new GeForce RTX 50 Series at reasonable operating temperatures, while providing increased performance.
Configurations between compromises and ambitions
The specific memory configurations of the different models of the RTX 50 series are still subject to debate, but NVIDIA would opt as usual for variable memory configurations depending on the target market. At the very high end, the RTX 5090 should be equipped with 32 GB of GDDR7 memory on a 512-bit bus, offering a gargantuan memory bandwidth of 1792 GB/s, no less. The RTX 5080 should retain the 16 GB of VRAM of the previous generation, but with the new GDDR7 and a 256-bit bus: enough to achieve in practice 1024 GB/s of bandwidth.
More affordable models in the range, like the RTX 5070 and RTX 5060, are the subject of conflicting rumors. Some sources claim that the RTX 5070 could have 16 GB of GDDR7 on a 256-bit bus (perhaps a Ti version?), while others point to the possibility of 12 GB on a 192-bit bus.
The RTX 5060 should be available in two versions: a Ti model with 16 GB of GDDR7 and a standard version with only 8 GB, both on a 128-bit bus. Such a low amount of VRAM on a new generation graphics card, even at the bottom of the catalog, still leaves us quite doubtful: restricted in this way, an RTX 5060 would have all the trouble in the world with the ever heavier textures of games AAA coming soon.
| GeForce | RTX 5090 | RTX 5080 | RTX 5070 Ti | RTX 5070 | RTX 5060 (TI ?) |
|---|---|---|---|---|---|
| GPU | Blackwell GB202-300 | Blackwell GB203-400 | Blackwell GB203-300 | Blackwell GB205 | Blackwell GB206? |
| SMs | 170 | 84 | 70 | 50 | ? |
| Hearts | 21760 | 10752 | 8960 | 6400 | ? |
| VRAM quantity | 32 Go GDDR7 | 16 Go GDDR7 | 16 Go GDDR7 | 12 Go GDDR7 | ? |
| Memory bus | 512-bit | 256-bit | 256-bit | 192-bit | 128-bit? |
| Vitesse VRAM | 28 Gbps | 32 Gbps | 28 Gbps | 28 Gbps | 28 Gbps? |
| Bandwidth | 1792 Go/s | 1024 Go/s | 896 Go/s | 672 Go/s | 448 Go/s |
| TBP | 600W | 400W | 350W | 250W | ? |
???? To not miss any news from 01net, follow us on Google News and WhatsApp.
Source :
WCCFtech
