NVIDIA GeForce RTX 5080 and RTX 5070 Ti: 2 times faster than previous models, RTX 5070 surpasses RTX 4090 for $549
In addition to the flagship GeForce RTX 5090 and the introduction of DLSS 4 technology, NVIDIA is also releasing the rest of its new series of graphics cards: RTX 5080, RTX 5070 Ti, and RTX 5070. These models significantly increase performance compared to the previous generation and offer better value.
NVIDIA GeForce RTX 5080 16 GB
RTX 5080 will be based on PCB design PG144/147-SKU45 and will use the GB203-400-A1 chip. The card will feature the full GB203 GPU with 84 SM modules and 10,752 CUDA cores, which is 51% fewer compared to RTX 5090. In comparison, RTX 4090 and RTX 4080 have a 40% difference in the number of cores, indicating a significant performance difference.
RTX 5080 will have 16 GB GDDR7 memory, operating over a 256-bit bus, and will use the fastest GDDR7 memory on the market with speeds up to 30 Gbps, providing a bandwidth of 960 GB/s. The card will have a TBP of 360W, which is 12.5% more than RTX 4080 SUPER (320W), but the actual consumption will be lower.
RTX 5080 is expected to be 2 times faster than RTX 4080 SUPER, thanks to DLSS 4, and will cost $999, which is more affordable than the original RTX 4080 ($1199).
NVIDIA GeForce RTX 5070 Ti 16 GB
RTX 5070 Ti will use the GB203 chip with reduced specifications and will feature 8960 CUDA cores. The memory will be the same as that of RTX 4070 Ti SUPER – 16 GB VRAM – but will use new GDDR7 modules, providing a 78% increase in bandwidth to 896 GB/s. The card will operate with a boost clock of 2.45 GHz.
RTX 5070 Ti will be 2 times faster than RTX 4070 Ti, will cost $749, and will support DLSS 4, making it a better choice compared to RTX 4070 Ti SUPER.
NVIDIA GeForce RTX 5070 12 GB
RTX 5070 will be based on the GB205 GPU and PG147 PCB. This card will feature up to 6144 CUDA cores and 12 GB of memory, operating over a 192-bit bus. Memory speeds will reach 28 Gbps, providing a bandwidth of 672 GB/s – a 33% increase compared to RTX 4070 and RTX 4070 SUPER. The card will have a TBP of 250W, which is 14% more than RTX 4070 SUPER (220W).
RTX 5070 is expected to be faster than RTX 4090, with a price of only $549. This positions it as an extremely competitive option in the premium segment and a serious competitor to AMD's new RX 9070 series.
RTX 5070 Ti and RTX 5070 are described as powerful cards for 1440p gaming, while RTX 5080 is aimed at 4K gaming.
Blackwell Architecture and Technical Details
The most powerful configurations of the Blackwell RTX 50 architecture are as follows:
- GB202 (FULL) -> 192 SM / 24,576 cores (100%) / RTX 5090 -> 170 SM / 21,760 cores (88.5%)
- GB203 (FULL) -> 84 SM / 10,752 cores (100%) / RTX 5080 -> 84 SM / 10,752 cores (100%)
- GB205 (FULL) -> 50 SM / 6400 cores (100%) / RTX 5070 -> TBD SM / TBD cores (TBD %)
Additionally, NVIDIA is improving the PCB drilling process for signal optimization between the GPU and GDDR7 VRAM. The new cards will use a 14-layer PCB and a 12V-2x6 power interface. Only some of the highest models RTX 5090 or RTX 5080 will have a dual 16-pin connector. The cards will be fully compatible with PCIe 5.0 and will support DP2.1a with a full bandwidth of 80 Gbps.
Release Dates
- NVIDIA GeForce RTX 5080 will be available on January 30.
- NVIDIA GeForce RTX 5070 Ti and RTX 5070 will be launched in February.
All cards will be offered in both Founders Edition and custom versions from NVIDIA partners. Prices, availability, and specific models for the Bulgarian market are still unknown. Expect information as soon as it becomes clear.
